(Oxford Moral Theory) Guy Fletcher, Michael Ridge - Having It Both Ways - Hybrid Theories and Modern Metaethics (2014, Oxford University Press)

Having It Both Ways
OXFORD MORAL THEORY

Series Editor
David Copp, University of California, Davis
Drawing Morals
Essays in Ethical Theory
Thomas Hurka
Commonsense Consequentialism
Wherein Morality Meets Rationality
Douglas W. Portmore
Against Absolute Goodness

Richard Kraut
The Lewd, the Rude and the Nasty

Pekka Väyrynen
In Praise of Desire
Nomy Arpaly and Timothy Schroeder
Confusion of Tongues
A Theory of Normative Language
Stephen Finlay
The Virtues of Happiness

A Theory of the Good Life
Paul Bloomfield
Having It Both Ways

Hybrid Theories and Modern Metaethics
Edited by Guy Fletcher and Michael Ridge
Having It Both Ways
Hybrid Theories and Modern Metaethics
EDITED BY GUY FLETCHER
AND
MICHAEL RIDGE
1
1
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.
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trademark of Oxford University Press
in the UK and certain other countries.
Published in the United States of America by
Oxford University Press
198 Madison Avenue, New York, NY 10016
© Oxford University Press 2014
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,
by license, or under terms agreed with the appropriate reproduction rights organization.
Inquiries 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 work in any other form
and you must impose this same condition on any acquirer.
CIP data is on file at the Library of Congress
ISBN 978–0–19–934758–2
9 8 7 6 5 4 3 2 1
Printed in the United States of America
on acid-free paper
CONTENTS
Contributors vii
Introduction ix
PART ONE Optimism about Hybrid Theories

1. How to Insult a Philosopher 3
Michael Ridge
2. Expressivism, Nondeclaratives, and Success-Conditional Semantics 22
Daniel R. Boisvert
3. Can a Hybrid Theory Have It Both Ways? Moral Thought, Open
Questions, and Moral Motivation 51
David Copp
4. Attitudinal Requirements for Moral Thought and Language: Noncognitive
Type-Generality 75
Ryan Hay
5. Diachronic Hybrid Moral Realism 95
Jon Tresan
6. The Pragmatics of Normative Disagreement 124
Stephen Finlay
7. Hybrid Expressivism: How to Think about Meaning 149
John Eriksson
vi Contents
PART TWO Pessimism about Hybrid Theories

8. Moral Utterances, Attitude Expression, and Implicature 173
Guy Fletcher
9. Pure versus Hybrid Expressivism and the Enigma of Conventional
Implicature 199
Stephen Barker
10. (How) Is Ethical Neo-Expressivism a Hybrid View? 223
Dorit Bar-On, Matthew Chrisman, and James Sias
11. Why Go Hybrid? A Cognitivist Alternative to Hybrid Theories of
Normative Judgment 248
Laura Schroeter and François Schroeter
12. The Truth in Hybrid Semantics 273
Mark Schroeder
Index 295
CONTRIBUTORS
Stephen Barker is Associate Professor and Reader in Philosophy at the

University of Nottingham
Dorit Bar-On is Zachary Smith Distinguished Professor of Philosophy at the
University of North Carolina, Chapel Hill
Daniel R. Boisvert is Lecturer in Philosophy at the University of North
Carolina, Charlotte
Matthew Chrisman is Reader in Philosophy at the University of Edinburgh
David Copp is Distinguished Professor and Chair of the Department of
Philosophy at the University of California, Davis
John Eriksson is researcher in Philosophy at University of Gothenburg
Stephen Finlay is Associate Professor of Philosophy at the University of
Southern California
Guy Fletcher is British Academy Postdoctoral Fellow and Lecturer at the
University of Edinburgh
Ryan Hay is Lecturer in Philosophy at the University of Wisconsin, Stevens
Point
Michael Ridge is Professor of Philosophy at the University of Edinburgh
Mark Schroeder is Professor of Philosophy at the University of Southern
California
François Schroeter is Senior Lecturer in Philosophy at the University of
Melbourne
Laura Schroeter is Lecturer in Philosophy at the University of Melbourne
James Sias is Assistant Professor of Philosophy at Dickinson College
Jon Tresan is Visiting Assistant Professor at the University of Rochester
Introduction
GUY FLETCHER AND MICHAEL RIDGEn
OV E R V I E W
The title of our volume is Having It Both Ways because the essays collected
here all discuss metanormative theories that try to accommodate both the
belief-like and the desire-like features of moral and other normative judg-
ments without abandoning a broadly Humean view in the philosophy of
mind. Such theories are perhaps most often called “hybrid theories,” and
we shall stick with that terminology here. By “metanormative theory” we
simply mean what one gets when one takes the more familiar category
“metaethical” and expands it to include all normative discourse and judg-
ment (e.g., prudential, aesthetic, epistemological). A metanormative theory
counts as a hybrid theory in our sense just in case it satisfies one of the fol-
lowing two criteria:
(i) Moral (/aesthetic/prudential/epistemological) claims express both

belief-like and desire-like mental states.
(ii) Moral (/aesthetic/prudential/epistemological) judgments are
constituted by both belief-like and desire-like components.
The new millennium has seen a blossoming of interest in hybrid theories

in this sense. This is perhaps no surprise, since the existing debate between
pure cognitivists and noncognitivists can easily appear intractable. Each
side of this debate can be seen to go in for more and more epicycles to
accommodate what the other side accommodates so easily or instead to
try to debunk the pretheoretical intuitions so cherished by the other side
of the debate. Moreover, the availability of hybrid theories reveals that this
apparently calcified traditional way of framing the debate rests on a false
dichotomy.
x Introduction
Hybrid theories themselves can come in both cognitivist and noncognitiv-

ist forms, though it is not entirely trivial to explain how this distinction is best
drawn within the hybrid framework. Because there are no clear and uncontro-
versial views on how the distinction is best drawn, we shall not offer a canonical
view of the matter in this introduction. (For one attempt to draw such a distinc-
tion, see c hapter 1).
The discussion of hybrid theories has moved beyond the stage at which parti-
sans developed and defended novel theories. These theories have now been sub-
jected to some criticism, and the possibility of comparing and contrasting them
along various dimensions has enriched the debate, leading naturally to some
fairly specific problems and questions. The contributions collected here attempt
to make progress on these problems and questions. Here we simply mention
a few of the most salient of these to give a flavor of where the debate is going.
For those hybrid theorists who emphasize the expression of desire-like states,
there is a very lively debate over what the best model is for understanding such
expression. Some hybrid theorists (e.g., Daniel Boisvert and David Copp) put
a lot of weight on the analogy with slurs and/or other pejoratives, but others
find this emphasis misguided (e.g., Mark Schroeder). Other hybrid theorists
instead understand the expression of desire-like states in terms of some form of
Gricean implicatures that the speaker has the relevant desire-like state. Among
these theorists there is a lively further debate about which form of implicatures
provides the best model—conventional, particularized conversational, or gen-
eralized conversational implicatures (here see Michael Ridge, Guy Fletcher, and
Stephen Barker for useful discussion). Yet other theorists appeal to a broadly
Gricean metasemantic framework to explain the relevant form of expression
(John Eriksson).
One attraction of hybrid theories has been their ability to accommodate
what seems plausible about pure forms of expressivism without having such
trouble with the dreaded “Frege-Geach” problem. With a number of initial
models having been developed to explain how such theories can avoid or solve
the Frege-Geach problem, an important challenge has arisen for this line of
argument. Schroeder has argued that, while many of these theories do well at
capturing some of what we want from a theory of validity, they do not explain
why valid arguments have what he calls the “inference-licensing property.”
Several contributions discuss this problem, its scope, and how it might be
solved (Copp, Ryan Hay, Schroeder). This challenge is closely related to the
question of whether hybrid theorists should embrace what is sometimes called
“the Big Hypothesis” about the target discourse (Copp, Schroeder). According
to the Big Hypothesis, “If ‘P’ is a sentence expressing mental states M1-Mn then
the descriptive content of ‘S believes that P’ is that S is in each of mental states
M1-Mn” (Schroeder 2009: 301).
Introduction xi
Another issue that has arisen is precisely what commitments in the theory of
meaning more generally are incurred by various hybrid theories. This issue is
touched on in several of the contributions here, with some theorists arguing for an
alternative to orthodox truth-conditional semantics (Boisvert, Schroeder, Barker),
while others argue that their preferred hybrid theory is best understood as a
metasemantic thesis that is compatible with the more traditional truth-conditional
approach (Eriksson). Another contribution (Dorit Bar-On, Matthew Chrisman,
and James Sias) puts a lot of weight on the distinction between semantic theory
and pragmatics to put pressure on a wide range of hybrid theories.
Another issue in the debate over hybrid theories is the extent to which they
gain plausibility from their ability to accommodate or explain away so-called
internalist intuitions—intuitions that there is an essential connection of some
kind between the relevant judgments and suitable motivation. This issue arises
in many of the contributions here in one way or another but is more central in
some than in others (Bar-On, Chrisman, and Sias; Copp; Eriksson; Fletcher;
Hay; Schroeter and Schroeter; and Tresan).
Other topics discussed include how certain hybrid theories should under-
stand normative disagreement (especially Stephen Finlay) and whether the
hybrid approach can usefully be extended to discourse about rationality (Ridge).
Another is whether hybrid theories can give a satisfying account of discourse
about normative truth without abandoning what made them attractive in the
first place (Schroeder). The danger that certain putatively expressivist forms of
hybrid theory ultimately collapse into a subjectivist form of cognitivism also
gets some discussion (Eriksson).
We have divided the volume as a whole into two parts. The first includes
contributions that defend hybrid theories (or some specific hybrid theory). The
second includes contributions that in some way argue against “going hybrid.”
C H A P T E R GU I DE
Part 1
(1) Michael Ridge, “How to Insult a Philosopher.” Ridge argues for a cog-
nitivist form of hybrid theory for discourse about rationality. This is a
departure from Ridge’s (2006a, 2006b, forthcoming) treatment in other
work of what he considers more robustly normative discourse (e.g., moral
discourse), where he defends a self-consciously expressivist form of hybrid
theory. In his contribution to this volume Ridge spends some time explain-
ing why a more cognitivist treatment is more apt in the case of thought
and discourse about rationality. The theory developed is a hybrid form of
xii Introduction
cognitivism because of the role of a certain kind of generalized conversa-

tional implicatures that Ridge argues is associated with discourse about
rationality. Ridge further argues that this theory can elegantly explain both
how “irrational” and cognates often function as terms of abuse and why
they sometimes do not. The contribution also defends a purely instrumen-
talist conception of the truth-conditional/descriptive meaning of “ratio-
nal” and cognates. The argument for this conception should be of interest
independently of its relation to the hybrid element of the theory.
(2) Daniel R. Boisvert, “Expressivism, Nondeclaratives, and Success-

Conditional Semantics.” Boisvert builds on previous work (Boisvert 2008) in
which he defends a form of hybrid theory that puts a lot of weight on an
analogy between moral predicates and slur terms and on some ideas from
speech-act theory. In particular he argues that moral discourse is like dis-
course in which slur terms are used as in each case declaratives are typically
used to make two direct speech acts. In this contribution he provides a more
systematic framework for hybrid theories by developing a novel approach
to semantic theory that he calls “success-conditional semantics.” He lays out
this framework in some detail and argues that it generalizes elegantly and
provides a very useful framework for hybrid theories of a certain sort. Pure
forms of expressivism do not, he argues, fit as plausibly into this framework.
(3) David Copp, “Can a Hybrid Theory Have It Both Ways? Moral Thought,
Open Questions, and Moral Motivation.” Like Boisvert, Copp draws on
an analogy between moral predicates and pejoratives. He takes a certain
hybrid account of the meanings of pejoratives as a model for the meanings
of moral predicates. In light of this model, in previous work (Copp 2001,
2009) he has defended a cognitivist hybrid theory according to which cer-
tain moral utterances express a commitment to a policy of avoiding moral
wrongdoing. He builds on this previous work here by responding to two
objections. First is the objection that his theory cannot account for the
element of endorsement that distinguishes moral thought from thought
about nonnormative matters. Second is the objection (from Schroeder)
that his theory (a cognitivist and realist one) cannot accommodate the
so-called “inference-licensing property” of valid arguments and that,
like other cognitivist hybrid theories, it is committed to the implausible
so-called Big Hypothesis.
(4) Ryan Hay, “Attitudinal Requirements for Moral Thought and

Language: Noncognitive Type-Generality.” In this contribution Hay wres-
tles with what he calls “the entangling problem.” The problem arises out
Introduction xiii
of the fact that certain hybrid theories seem committed to the implausible
thesis that a sincere nihilist who says “torturing cats is not morally wrong”
thereby expresses or has some motivation to not act wrongly or to not
torture cats. Hay argues that this problem can be avoided if we understand
the general attitudes expressed by moral claims as toward types rather than
tokens. He further argues that this allows the hybrid theory to accommo-
date a form of internalism that is not implausibly strong.
(5) Jon Tresan, “Diachronic Hybrid Moral Realism.” Some hybrid theo-
ries say that moral terms have two meanings. In this contribution Tresan
describes a “diachronic” version of this view, according to which moral
terms exhibit the two meanings on different occasions, going back and
forth between them, rather than exhibiting both on the same occa-
sions. The crucial distinction is between their use in moralizing and in
“metamoralizing,” thinking and talking about moral attitudes and prac-
tices, where they are alleged to have a social-functional meaning. Tresan
argues that such a theory draws dialectical dividends by breaking the link
between moral realism and the orthodox realist assumption that moral
judgments are individuated by their contents. By doing this it can combine
nonrelativistic moral realism with a robust kind of motivational internal-
ism and significant metaphysical and anthropological frugality.
(6) Stephen Finlay, “The Pragmatics of Normative Disagreement.” In

previous work Finlay (e.g., 2004, 2014) has developed a cognitivist
theory that makes the contents of normative claims end relative. The
hybrid element of this theory consists in its position that at least some
normative claims also express associated proattitudes through strictly
pragmatic mechanisms akin to conversational implicatures. A prima
facie problem for such relativistic theories is that they seem to imply that
people do not disagree when intuitively they do. Finlay develops a novel
“quasi-expressivist” theory of normative disagreement that he combines
with his end-relational theory to meet this worry about disappear-
ing disagreement. This builds on earlier work with Gunnar Björnsson
(Björnsson and Finlay 2010) by extending the quasi-expressivist account
into a general theory of normative disagreement and arguing that it
enjoys important advantages over rival accounts of disagreement that
are purely cognitivist or purely expressivist.
(7) John Eriksson, “Hybrid Expressivism: How to Think about Meaning.”

Unlike the other contributions in part 1, Eriksson’s contribution defends
an expressivist form of hybrid theory (heavily inspired by the work of
xiv Introduction
R. M. Hare) rather than a cognitivist one (see also Eriksson 2009). Indeed
one of the main tasks of the contribution is to explain why the theory on
offer counts as a form of expressivism rather than as a form of subjec-
tivism. The contribution also argues that expressivists would do well to
understand their theory in terms of the ideationalist framework devel-
oped by Wayne Davis.
Part 2
(8) Guy Fletcher, “Moral Utterances, Attitude Expression, and Implicature.”

Fletcher’s contribution examines the plausibility of implicaturist hybrid
theories by examining how closely attitude expression by moral utter-
ances fits with the varieties of implicature (conventional, particular con-
versational, generalized conversational). He argues that conventional
implicature is a clear nonstarter as a model of attitude expression by moral
utterances and that generalized conversational implicature yields the most
plausible implicaturist hybrid. He closes by considering a challenge that
remains even for such generalized conversational implicaturist views, one
that motivates a nonimplicaturist and nonhybrid alternative.
(9) Stephen Barker, “Pure versus Hybrid Expressivism and the Enigma of
Conventional Implicature.” Barker’s contribution builds on his previous
work (Barker 2000) on the role of conventional implicatures in hybrid the-
ories. Here he in effect retracts his previous enthusiasm for such theories
and argues for the superiority of a kind of pure expressivism. The form of
the argument is that to make sense of conventional implicatures itself we
already need to embrace a kind of pure expressivism about implicatures.
Once we have gone this far with pure expressivism, though, we might as
well be pure expressivists about normative discourse too. The motivations
for a specifically hybrid theory are dialectically elusive on this way of fram-
ing the debate—in particular given its emphasis on specifically conven-
tional implicature.
(10) Dorit Bar-On, Matthew Chrisman, and James Sias, “(How) Is Ethical
Neo-Expressivism a Hybrid View?” This contribution further develops
and defends the neo-expressivist theory that Bar-On and Chrisman (2009)
have defended elsewhere. The core idea is that moral claims express prop-
ositions but also motivational states of mind. They argue, however, that
this is not because moral sentences express states of mind with belief-like
and desire-like components but rather because there are two notions of
Introduction xv
expression that are theoretically relevant to this debate. The first semantic
notion of expression is the way a sentence stands in relation to its semantic
content, for example, a declarative sentence expresses a proposition. The
second action notion of expression is the way an action can stand toward
a state of mind when it communicates or shows it, for example, an act of
assertion expresses a belief. They suggest that several of the reasons for
thinking that moral sentences express hybrid states of mind can be better
accommodated by the distinction between these two notions of expres-
sion. Their contribution also responds to a range of interesting objections
that have been lodged against their positive alternative.
(11) Laura Schroeter and François Schroeter, “Why Go Hybrid?

A Cognitivist Alternative to Hybrid Theories of Normative Judgment.”
Schroeter and Schroeter argue that a primary motivation for “going
hybrid” (in their view), namely the appeal of some kind of motivational
internalism, can be accommodated by a pure form of cognitivism if one
embraces the original theory of concepts they develop here. The proposed
theory understands concepts as individuated relationally: sharing the
same concept is determined in part by quasi-anaphoric relations between
token thoughts. They emphasize that their theory is offered at the level of
thought rather than at the level of talk, in the first instance, and argue that
this gives their pure cognitivist theory some advantages over at least some
forms of hybrid theory.
(12) Mark Schroeder, “Truth in Hybrid Semantics.” In previous work

Schroeder (2009) has argued that cognitivist hybrid theorists need to
embrace the so-called Big Hypothesis and that this tells against the anal-
ogy with slur terms. In this contribution Schroeder builds on this previ-
ous work and explores what a theory that embraced the Big Hypothesis
would look like. He argues that although such a theory is not hard to
construct and describes a possible realm of discourse, it does not fit well
with our actual normative discourse. In particular he argues that it does
not fit well with our use of the truth predicate in normative contexts.
He allows that some forms of hybrid theory might avoid this worry but
only at the price of making the attraction of their hybrid theory very
subtle and perhaps elusive. For broader context, it is worth noting that
Schroeder’s critique here applies only to cognitivist forms of hybrid the-
ory. He himself has developed a form of hybrid expressivism (in one
useful sense of that phrase), which he does not endorse, all things con-
sidered, even while arguing that it has several very large advantages (see
Schroeder 2013).
xvi Introduction
Acknowledgments
The editors would like to thank David Copp and Peter Ohlin for proposing
and developing this collection. This work arose from a conference generously
funded by the Arts and Humanities Research Council; the Analysis Trust; the
School of Philosophy, Psychology, and Language Sciences at the University of
Edinburgh; and the Mind Association. Guy Fletcher was supported during this
time by a British Academy postdoctoral fellowship.
References
Barker, S. 2000. “Is Value Content a Component of Conventional Implicature?” Analysis
60: 268–279.
Bar-On, D., and M. Chrisman. 2009. “Ethical Neo-Expressivism.” In Oxford Studies in
Metaethics, ed. R. Shafer-Landau, vol. 4, 133–166. Oxford: Oxford University Press.
Björnsson, G., and S. Finlay. 2010. “Metaethical Contextualism Defended.” Ethics
121: 7–36.
Boisvert, D. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly
89: 169–203.
Copp, D. 2001. “Realist Expressivism: A Neglected Option for Moral Realism.” Social
Philosophy and Policy 18: 1–43.
———. 2009. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies
in Metaethics, ed. R. Shafer-Landau, vol. 4, 167–202. Oxford: Oxford University
Press.
Eriksson, J. 2009. “Homage to Hare: Ecumenism and the Frege-Geach Problem.” Ethics
120: 8–35.
Finlay, S. 2004. “The Conversational Practicality of Value Judgement.” Journal of Ethics
8: 205–223.
———. 2014. A Confusion of Tongues. Oxford: Oxford University Press.
Ridge, M. 2006a. “Ecumenical Expressivism: The Best of Both Worlds?” In Oxford
Studies in Metaethics, ed. R. Shafer-Landau, vol. 2, 51–76. Oxford: Oxford University
Press.
———. 2006b. “Ecumenical Expressivism: Finessing Frege.” Ethics 116:302–336.
———. 2014. Impassioned Belief. Oxford: Oxford University Press.
Schroeder, M. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119: 257–309.
———. 2013. “Tempered Expressivism.” In Oxford Studies in Metaethics, ed. R. Shafer-
Landau, vol. 8, 283–314. Oxford: Oxford University Press.
PART ONE
Optimism about Hybrid

Theories
1
How to Insult a Philosopher

MICHAEL RIDGE n
Reason is, and ought only to be, the slave of the passions.
—David Hume, A Treatise of Human Nature
Let there be no crude suspense: one of the best ways to insult a philosopher

is to call him or her irrational. While philosophers are especially prone to feel
the force of the epitaph, “irrational” does more generally function as a term of
abuse. It is also characteristically used to offer advice or express disapproval.
Yet all of these functions of “irrational” depend on context. An adequate theory
of the meanings of “rational” and “irrational” should help explain why “irratio-
nal” is well suited to play these roles but at the same time should help explain
why it plays them only in certain contexts.
In this chapter I focus primarily on talk of practical rationality. I defend a view
of practical rationality on which it amounts to nothing more or less than a kind
of internal coherence. Very roughly, I defend a broadly instrumentalist concep-
tion. My main argument for this view (developed in sections 1 and 2) is that it
best fits with our most strongly held convictions about rationality and at the
same time preserves the right kind of linkage between “rational” in what I shall
call the “capacity sense” and “rational” in what I shall call the “success sense.”
At the same time, I try to accommodate the fact that “irrational” can char-
acteristically function as a term of abuse, as a vehicle for offering advice and
expressing disapproval. To this end, I build on some of the insights of a certain
sort of hybrid theory in metanormative theory I have elsewhere called “ecu-
menical cognitivism.”1 In particular, I defend a species of what I call “implicative
1. See Ridge 2006, forthcoming.

4 H aving I t B oth Ways
ecumenical cognitivism” about rationality thought and discourse (in section 3).
The main idea is to appeal to a kind of Gricean implicature to explain the dis-
tinctive and context-sensitive features of discourse about rationality.
I argue that this account can not only explain why “irrational” and cognates
play the roles canvassed above when they do. I further argue that it can help
explain why these terms seem normative in a richer sense than (on my view)
they actually are. On the proposed account, certain uses of these terms implicate
robustly normative thoughts even though they do not have normative semantic
contents. The idea is to debunk the idea that these terms have robustly norma-
tive semantic contents in a way that would undermine the sparser and broadly
Humean analysis of “rational” and cognates defended here.
RAT I ON A L C A PACI TI E S
We must not suppose that, because a man is a rational animal, he will,

therefore, always act rationally.
—Philip Dormer Stanhope, 4th Earl Chesterfield, letter, Dec. 19, 1749,
Letters Written by the Late Right Honourable Philip Dormer Stanhope, to
his son, 5th ed. Vol. II, p. 297. London.
The 4th Earl of Chesterfield reminds us that being a rational animal does not
ensure universally rational action. This in turn should remind us that “rational” is
said in many ways. In one sense, to say that someone is rational is to ascribe to him
or her a set of distinctive capacities—capacities that we have and that cockroaches
and mushrooms lack. In another sense, to say that someone is rational is to say
that he or she is succeeding in some important sense. However, this is not brute
ambiguity—it is very plausible to suppose that the capacity sense and the success
sense of “rational” are semantically linked in some important way. The hypothesis
guiding this chapter is that to say that someone is rational in the success sense just
is to say that he or she successfully adheres to those norms to which anyone who
counts as rational in the capacity sense must, for the most part, adhere.2
I begin my defense of this hypothesis by listing what I take to be some of the
platitudes about rational agents in the capacity sense. Because I take these prop-
ositions to be intuitively plausible, I do not argue for them but just lay them out:
(1) Rational agents as such can set and abandon ends.

(2) To set an end, a creature must have a will; to set an end is to
commit one’s will to it or, more briefly, “to will it.” Mere stimulus/
response creatures and creatures that act directly on instincts or
2. For a similar form of argument for a Humean theory of rationality, see Savavarsdóttir 2006.
How to Insult a Philosopher 5
innate desires cannot set ends, though in some sense they may
“have ends” set for them by nature.
(3) To have a will, a creature must be capable of asking itself “What
shall I do?” and answering this question and adopting a suitable
end on the basis of its answer.
(4) In asking “What shall I do?” a creature acts “under the idea of
freedom”—that is, it acts under the idea that it is in some sense “up
to him or her” which of a set of options to select and pursue.
(5) For creatures with immediate impulses, such a creature can count
as rational only if it is capable of delaying or even rejecting the
gratification of its immediate impulses for the sake of an end that
it wills.
(6) Rational agents as such can take what they believe to be the
essential means to their ends because they believe them to be an
essential means to their ends.
(7) Rational agents as such can will whatever they take to be a means
to their ends.
(8) Rational agents as such can revise their ends when they take those
ends to conflict—in the sense that, by the agent’s lights anyway,
both ends cannot both be realized.
(9) Rational agents as such can form new ends as a way of specifying
otherwise more abstract and unspecified ends. For example, a
rational agent who adopted the more abstract end of learning
a foreign language can in turn adopt the more specific end of
learning French.
(10) Rational agents as such are capable of making and acting on
normative judgments—for example, judgments of the form “I
ought to Φ in circumstances C.”
I am not tempted to make the bold claim that these platitudes entirely
exhaust our (largely implicit) pretheoretical concept of practical rationality in
the capacity sense. Instead, I make the following more modest claim. These
platitudes “hang together” in that they describe a kind of entity that is capable
of willing and pursuing a coherent set of ends as well as revising those ends over
time in the light of new information and reflection. This striking coherence
among the platitudes is no accident.
One might object that at least one of the platitudes listed above does not
“hang together” with the others. In particular, platitude number 10, the the-
sis that a rational agent as such is capable of making and acting on normative
judgments does not obviously cohere well with a conception of rational agents
setting ends, revising them to make them into a coherent set, and pursuing
them. In fact, though, this platitude does turn out to fit very well with the others
given a form of expressivism I argue for elsewhere—ecumenical expressivism.3
On my preferred version of ecumenical expressivism, normative judgment is
partially constituted by a certain sort of commitment of the will. In this way
ecumenical expressivism can explain why making normative judgment itself
presupposes many of the other features on the preceding list.
According to ecumenical expressivism, to judge (in the first person) that one
must perform a given action just is to occupy a normative perspective that is
constituted by (i) an intention (a commitment of the will) to do whatever any
standard of a certain sort would require and (ii) a belief that any such standard
would require performing the action. Someone with this combination of prac-
tical commitment and belief about what is necessary to fulfill that commitment
is rationally committed to acting accordingly. To fail to do so would be to fail to
take a constitutive means to one’s end and hence to be instrumentally irrational.
The theory of rationality developed here vindicates the idea that this failure is
a form of irrationality.
In fact this represents an independent, albeit indirect, argument for ecu-
menical expressivism. Ecumenical expressivism has the virtue of being able to
explain how the pretheoretical platitude linking rational agency to normative
judgment coheres well with many of our other pretheoretical platitudes about
rational agency. Many other theories are unable to offer such a direct and ele-
gant explanation of this connection between normative judgment and the will.
Typically, realist theories of normative judgment are driven to stipulate and
take it as semantically axiomatic that acting against what one takes to be the
balance of reasons, or failing to intend to do what one thinks is best, is sufficient
for irrationality (see, e.g., Scanlon 1998, 25).4 In effect such theorists are forced
to hold that this is just part of what “irrational” means as a kind of primitive
semantic axiom alongside seemingly independent axioms about coherence. On
these views, the notion of rationality becomes something more of a “hodge-
podge” notion, bringing together seemingly conceptually unrelated notions of
acting on the balance of reasons, on the one hand, and taking the necessary
means to one’s ends, not willing incompatible ends, not believing contradic-
tions, and so on, on the other. At least realists are driven to such a hodgepodge
view unless they can explain how recognizing that an action instantiates the
property privileged by the realist theory as being the property of being what one
must do, yet not being motivated accordingly is a genuine form of incoherence
broadly akin to intending incompatible actions or believing a contradiction.
3. See especially Ridge2014 but also Ridge 2006.

4. A notable example is Michael Smith (1994). Smith’s view faces other problems, though it
would take me too far afield to delve into the problems specific to his view here.
This is no easy task, particularly for antireductive and nonnaturalist realists.

Some forms of reductionism may do better but face other familiar problems.
By contrast, on the account developed here, rationality (in both the practical
and the theoretical domains as well as across these) is a more unified notion.
Realist theories instead require fairly ad hoc and disjointed analyses of “irratio-
nal” and cognate terms or must tell some very fancy story about rationality and
normative judgment that can link the two without such a disjointed account—
as, for example, Michael Smith tries to do and as some Kantians also try to do.
I am skeptical about the latter, but a full discussion of such theories would go
well beyond the present context.
In fact in many interesting respects the account of practical rationality devel-
oped on the basis of the capacities listed above is itself recognizably Kantian.
Rational agents are, on my account, capable of adopting ends, which in Kantian
terms is to incorporate an “incentive” into their maxim (which might be
understood as a policy, plan, or intention) as an end. To count as being capa-
ble of adopting ends, in turn rational agents must have wills. Moreover, like
Immanuel Kant, I hold that rational agents take it that their wills are free in a
way that would be incompatible with their behavior being entirely determined
by instincts or desires—they act “under the idea of freedom.”
In spite of this Kantian flavor, in another important respect the account of
practical rationality developed here is deeply unKantian. Kant famously held
that a rational agent as such is committed to a substantive necessary end—
humanity. The commitment to this end is, in Kantian terms, a “categorical
imperative” in that its rational authority is not contingent on the agent’s adop-
tion of any contingent end. Indeed it is not too far off the mark to characterize
my view of practical rationality as what one gets when one takes Kant’s view
but subtracts “the” categorical imperative—the moral law understood as an
imperative of reason.
Given the stark contrast between my proposed account and Kant’s much
more ambitious one, how might we test their relative plausibility? Consider the
following thought experiment. Suppose we are on board the USS Enterprise (as
depicted in the TV show Star Trek) seeking “new life and new civilizations.”
Part of our mission is to classify new life-forms as rational or otherwise—in
the capacity sense. Suppose we come upon a species of creatures that satisfies
all of the platitudes listed above. However, further suppose that these creatures
are not committed to whatever end the Kantian suggests is a necessary end—
humanity, respect for the moral law, or whatever. We might suppose that the
creatures we find are completely amoral, and their ends may vary greatly from
one agent to the next. They have no moral code, no sense of moral conscience,
guilt, or shame. By hypothesis, though, they fulfill all of the platitudes listed
above. They have wills, adopt ends, deliberate about what to do, act under the
idea of freedom, make normative (but nonmoral) judgments about what to do,
act on those judgments, and so on. Nonetheless, they are deeply amoral.
My suggestions about this possibility are twofold. First, it is intuitively per-
fectly coherent. The burden of proof is firmly on the Kantian to demonstrate
that there is some incoherence lurking in it. Kantians have of course tried
to discharge this burden of proof but, in my view, unsuccessfully.5 So-called
Kantian constitutivists try to derive all practical normativity from the norms
that, on their account, we must accept simply to count as rational in the capac-
ity sense. I shall not discuss the wide variety of Kantian arguments that have
been given in this chapter; that would take us too far afield. I have discussed
one of the most influential arguments of this kind at length elsewhere, though.6
Second, ordinary speakers would not hesitate to call such an amoral spe-
cies of creatures rational (in the capacity sense). Indeed popular conceptions of
psychopaths fit this profile, and ordinary speakers do not hesitate in character-
izing such agents as rational in the capacity sense. Given my proposed linkage
between “rationality” in the success and capacity senses, though, Kantians seem
committed to arguing that the creatures I have described are not truly rational
even in the capacity sense. If being rational in the success sense is somehow
succeeding in terms fixed by the relevant capacities and being irrational if fail-
ing in terms likewise fixed by those capacities, then it is hard to see why such
creatures’ failure to treat humanity as an end needs be irrational. To be sure,
their failure to do so could be irrational, depending on what contingent ends
they happen to have. My point is simply that they need not be. Furthermore, the
creatures as I am imagining them have no concept of treating humanity with
respect. Nor do they have any concept of moral duty or obligation. Yet for the
Kantian these concepts are essential to being a rational agent.
A Kantian could allow that the case is coherent but deny that the creatures
in question are genuinely rational even in the capacity sense. This, though, just
seems pretheoretically very counterintuitive. Captain Kirk could certainly be
excused for characterizing such a species of creatures as rational (in the capac-
ity sense) but amoral. We would certainly want to mark an important distinc-
tion between such a species and, say, earth cats. My hypothesis is that in English
the word we would most naturally use to capture this contrast is indeed “ratio-
nal” as used in the capacity sense.
5. See, perhaps most famously in the recent literature, Korsgaard 1996. See also Tubert 2010
and Bagnoli forthcoming. The general constitutivist framework need not come with such
ambitious Kantian assumptions, though; compare Street 2008.
6. See Ridge 2005.
Consider a more down-to-earth illustration of the central point. Tom Ripley

is the main character in Patricia Highsmith’s novel The Talented Mr. Ripley (later
made into a film by the same name). In the story Ripley does not hesitate to kill
someone with whom he seemed to be friends in order to take over his iden-
tity—Ripley is also a master of disguise and forgery. Ripley is lacking in all signs
of guilt or remorse and does not hesitate to commit homicide for personal gain.
All the same, he does not seem to be arational in the way that a cockroach or a
cat is arational. He is capable of planning, indeed elaborate planning, carrying
out his plans, resisting temptation, and the like. He presumably takes himself to
be free to choose what to do, he has views about what he ought to do, all things
considered, and he can and does act on those views. Once again, rationality in
the capacity sense seems too thin to ground robust norms.
R AT I ON A L I T Y AS A SUCCE SS NO TI ON
Recall my opening methodological maxim: “rational” in the capacity sense and

“rational” in the success sense are semantically linked. Our use of “rational”
across these contexts is not a brute ambiguity. I propose specifically to under-
stand rationality in the success sense in terms of the successful deployment of
the capacities essential to rationality in the capacity sense. More specifically
still, I propose the following definitions of “rationally permissible,” “rationally
impermissible,” and “rationally required,” respectively:
A rational agent’s action is rationally permissible just in case it is compat-

ible with the agent’s adherence to those norms that an agent must be
taken to adhere, for the most part, and absent some special explanation,
in order for the idea that the agent is rational (in the capacity sense) to
be intelligible.
A rational agent’s action is rationally impermissible just in case it is not
rationally permissible.
A rational agent’s action is rationally required just in case its omission is
not rationally permissible.
The idea that these notions (permissibility, impermissibility, and require-

ment) can be defined in terms of one another in this way is familiar. What is
distinctive about the proposal is the idea of defining these notions in terms
of those norms an agent must be interpreted as following, for the most part,
in order for the idea that he or she is a rational agent in the capacity sense to
be intelligible. Crucially, the norms are ones to which an agent must adhere
only for the most part and absent special explanation to count as rational (in the
capacity sense) in the first place. These caveats are essential, because it must also
be possible to flout these norms some of the time and still count as a rational
agent. Otherwise the view would entail the absurd conclusion that irrationality
is impossible.
This account preserves an elegant conceptual connection between “ratio-
nal” in the capacity sense and “rational” in the success sense. It also promises
to explain why convincing someone that his or her proposed course of action
would be irrational can motivate him or her not to pursue that course of action.
For, plausibly, rational agents manage to adhere to the relevant norms for the
most part, at least in part, because they are sensitive to violations of those norms
and are reliably disposed to resist such violations.
What norms are in this sense constitutive of rational agency? First and fore-
most, what Thomas Hill has called “the Hypothetical Imperative” is constitutive
of practically rational agency in this sense. Hill suggests that implicit in Kant’s
discussion of specific hypothetical imperatives is the idea of the hypothetical
imperative. As Hill defines it, the hypothetical imperative does not straitjacket
us, so that once we have adopted an end we must follow through on it come hell
or high water. All the hypothetical imperative requires is that we always either
(a) take the necessary means or (b) give up the end (Hill 1992, 24). Someone
who continues to will an end but refuses to take what he or she believes to be
an indispensable means to that end flouts the hypothetical imperative and is
therefore irrational.7
My point here is simply that we cannot make sense of a rational agent who
was not disposed, for the most part and absent some special story, to take what
he or she believed to be a necessary means to his or her end. Such special stories
are of course not hard to come by, and people do often act irrationally. I intend
to get some work done today, but I am depressed about the loss of a loved one,
so I do not even get as far as the office. I intend to go skydiving, but as I reach
the door of the plane I am paralyzed with fear. I intend to pick up some soy milk
on the way home, but I simply forget; I am absentminded.
Why are such special stories needed to make sense of instrumental irratio-
nality? To count as deploying the capacities outlined in the previous section
at all, an agent must be disposed, special stories notwithstanding, to take the
necessary means to his or her ends. As Kantians sometimes put the point, to
will an end just is to commit oneself to taking the necessary means to that
end. Such a commitment should be enough to explain your taking such means
unless some special feature of the situation prevents your commitment from
issuing in an action.
7. Interestingly, the hypothetical imperative is itself a categorical imperative—it applies to us

regardless of our ends.
Do any other practical norms have the same special status as the hypotheti-
cal imperative? Perhaps; one candidate is a norm that forbids the adoption of
logically incompatible ends, though that might be thought to be derivable as a
sort of theorem from the hypothetical imperative. For presumably a necessary
means to my achieving a given end E will, at least typically, be that I do not also
will not-E (or anything that entails not-E). In any event, practical rationality
will, on this account, turn out to be understood entirely in terms of internal
coherence—coherence of ends with one another and coherence of means to
ends. Apart from logically contradictory ends, such as making it my end to
make a round square, there will be no intrinsically irrational ends. Nor will
there be any substantive ends that a rational agent must will.
This of course is where my own account of rationality departs sharply from
robustly Kantian ones. Why might one accept such a thin conception of ratio-
nality? Crucially, this conception preserves the right sort of linkage between
“rational” in the capacity sense and “rational” in the success sense in an obvious
and intuitive way. I take this to be an important point in favor of the proposal.
This conception also fits well with many of our most strongly held intuitions
about practical rationality. It is not without reason that generations of philoso-
phers have taken a failure to take the necessary (by one’s lights) means to one’s ends
to be a clear and paradigmatic instance of practical irrationality. Likewise, the idea
that willing logically incompatible ends is irrational is a very plausible one.
Finally, this conception holds out the promise of continuity with a plausible
account of theoretical rationality. For in both cases the idea of a kind of consis-
tency or internal coherence looks essential to our notion of rationality. After all,
what clearer instance of theoretical irrationality could there be than someone
holding contradictory beliefs?
I do not pretend that the proposal can capture absolutely all of our intuitions
about practical rationality. For at least some ordinary speakers are disposed
to characterize pointlessly self-destructive actions as irrational even when they
advance the agent’s most cherished ends. What Allan Gibbard at one point
called the “fully coherent anorexic,” someone who wholeheartedly wills to
starve himself or herself to death, might, I admit, naturally enough be charac-
terized as irrational (see Gibbard 1992, 166). Yet such an agent need not violate
the hypothetical imperative, nor need he or she thereby will incompatible ends.
Is this a decisive counterexample?
For a start, is not entirely clear how many people find this characterization
apt. I expect that most ordinary speakers would say that the coherent anorexic
“is crazy,” “is mental,” “is highly self-destructive,” or just “needs help.” I expect
ordinary speakers would find the accusation of irrationality odd at best and sim-
ply false at worst. These expectations are admittedly just anecdotally informed
hunches, though.
More importantly, in reaching a reflective equilibrium, sometimes some intu-

itions—even some strongly held ones—must be abandoned. Insofar as it really
is intuitive that the fully coherent anorexic is irrational, this intuition might be
an acceptable casualty of systematic theorizing. One reason to think this would
be an acceptable intuition to jettison is that any account that can preserve our
intuitions about such cases (e.g., that the ideally coherent anorexic is irrational)
will thereby sever the right sort of linkage between rationality in the capacity
sense and rationality in the success sense or presuppose an implausibly strong
conception of the capacities essential for rational agency. The thin capacities
that, on my account, define rationality in the capacity sense are far too formal
to ground a commitment to something as substantive as not starving or, more
generally, not harming oneself. Insofar as we should preserve that linkage, that
provides a good reason to reject the only sorts of theories that could vindicate
the intuition that the fully coherent anorexic is irrational (for example).
However, I do not have to insist that the intuition that “irrational” is an apt
term in such cases is entirely mistaken. For I allow that “rational” and its cog-
nates are sometimes used in a broader and more prescriptive sense than the one
I have outlined here. In my view, “irrational” is sometimes used simply to mean
something like “self-destructive without good reason.”8 In other contexts “irra-
tional” is perhaps used simply to mean something like “obviously stupid” or
“obviously not the thing to do.” In my view, “irrational” in these senses should
be understood in terms of ecumenical expressivism, given that they are prop-
erly defined in terms of justifying reasons for action. However, these senses of
“irrational” are also less central than the instrumentalist one laid out here in
that they fail to preserve the right linkage between “rational” in the capacity
sense and “rational” in the success sense.
Moreover, it is not terribly hard to see how “rational” and its cognates could
have come to have these sorts of more richly normative meanings as well, given
their more central internal coherence-based meanings. For we generally and
quite safely presume that agents intend not to seriously sacrifice their own over-
all welfare without good reason. So we might safely presume that someone, for
example, an anorexic, who did sacrifice his or her own welfare without good
reason would thereby fail to take a necessary means to this (presumed) end
and would therefore be irrational on the instrumentalist account too. This gen-
eralization holds up so well that it might have often made sense, simply as a
heuristic, to use “irrational” to describe anyone who does sacrifice his or her
8. This definition is similar, at least in spirit, to the sorts of definitions proposed by both
Bernard Gert (1966) and Joshua Gert (2004). I differ from them on this point only in that
I take this usage to be derivative and less central than the one I defend as the primary sense
in the text.
own welfare without good reason. This heuristic could then have taken on a
“life of its own,” and “irrational” and cognate terms could come in this sense
to apply to anyone who sacrifices his or her own interest without good reason
even if the agent is not in fact irrational in the instrumentalist sense that I have
argued represents our most central notion of practical rationality. Even more
clearly, we presume that people will recognize and act on obvious and obviously
decisive reasons.
Why think the instrumentalist analysis is primary? First, it is much better
placed to provide the right kind of linkage between the capacity and success
senses of “rational” and cognate terms. That linkage is likely to be most at home
in the original and primary sense of the term; derivative uses often are idiom-
atic or depart in other ways from the original semantic content.
Second, as I noted above, the proposed account of “rational” in its most cen-
tral sense provides a nice symmetry with what I take to be a plausible view of
theoretical rationality couched in terms of consistency. In both cases, rational-
ity amounts to a kind of internal coherence, while irrationality is the corre-
sponding sort of incoherence.9
Third, rationality in the instrumentalist sense does some very important
theoretical work that the other, more robustly normative notion cannot do. For
rationality in the instrumentalist sense represents a sort of limit on our ability
to make sense of some entity even counting as a reflective agent. Indeed ratio-
nality in this sense is widely assumed to be necessary for interpreting someone
as a reflective agent in the first place. This does not seem to be true of rationality
if defined in terms of self-interest (for example). Plausibly, we can at least con-
ceive of agents who are not so motivated—for example, agents who will sacri-
fice their own welfare for very poor reasons or for no reason at all. Granted, this
stretches the imagination, because human beings are our only clear paradigms
of rational agency, and human beings do tend to be strongly self-interested.
Somewhat more controversially, we can make sense of rational agents who are
regularly not motivated by what are obviously good reasons or who lack the
concept of welfare.
In fact I would go so far as to claim that the primary function of judgments
of rationality is to make a certain kind of sense of agents—to interpret them as
acting for reasons. Deploying the concept of rationality is a way of taking up
what Daniel Dennett (1987) memorably calls the “intentional stance.” By taking
this stance we are able to predict and explain behavior in rich and systematic
ways that we could not manage from the “physical stance” or even the “design
stance.” On my preferred account of normativity, a class of judgments is norma-
tive only if its primary function is to settle the thing to do (or the thing to think,
9. Compare Davidson 1980.

or the thing to feel).10 So the hypothesis I have just floated, insofar as it is prima
facie plausible, provides an independent argument for the thesis that discourse
about rationality is not normative in my preferred sense.
I admit, though, that there is a sense in which judgments of irrationality
function to settle the immediate thing not to do as well as the thing not to think
or feel. Insofar as I am rational in the capacity sense, I will be disposed not to
do things I take to be irrational here and now. On the one hand, my judgment
that some action of mine in the future would be irrational, given my ends at
that time, might not motivate me to avoid so acting. I might view the likely ends
of my future self as banal, worthless, or perverse. For example, I might now
endorse various liberal causes but be savvy enough to anticipate that in my old
age I will likely become jaded and more conservative. That need not motivate
me here and now to ensure that my future, more conservative self rationally fol-
lows through on his various conservative ends.11 However, my judgment here
and now that what I am presently doing is irrational given my current ends will
dispose me either to stop doing it or to give up the relevant end(s).
Note that this is much more anemic than the sense in which a judgment
about what I must do, all things considered, settles the thing to do. First, these
judgments settle the thing to do by my lights not only for my immediate actions
here and now, given my current ends. They settle the thing to do for any cir-
cumstance for which I make the judgment.
Second, judgments about what I must do, all things considered, can settle
that I am to do such and such specific action. Judgments of irrationality can
only settle that here and now I must not both act in a given way and hold onto
some relevant end, given my beliefs. That is a much weaker sense of “settling”
the thing to do, since it leaves it entirely open that I might perform the very
action, here and now, and just give up the end. It is in this sense that what is
settled is only something negative—do not both hold onto the relevant end and
perform this action (which by your lights makes achieving the end impossible).
Third, it is plausible that the primary function of rationality thought and dis-
course, given its thick descriptive sense and role in our practice of interpreta-
tion, is plausibly the function of making a certain kind of explanatory/predictive
sense out of patterns of behavior, whereas the primary function of genuinely
normative judgments is to settle the thing to do (or think or feel). Note that
similar contrasts apply in the case of judgments of theoretical rationality and
judgments of truth. For theoretical rationality, on the account proposed here, is
10. I defend this at length in Ridge 2014.

11. Compare Derek Parfit’s (1984, 327–328) nice discussion of the nineteenth-century
Russian nobleman.
also always a relative matter and indexed to an agent at a specific context, where
the context includes the agent’s other beliefs.
For a useful comparison, note that judgments of the form “I here and now
believe I must perform such and such action” also arguably settle “the thing to
do” in whatever sense first-person judgments of rationality, on my account, do.
It is, however, only in a very strained sense that these judgments are normative
ones rather than purely psychological ones that happen to have considerable
normative significance for the person making the judgment at the time of judg-
ment. Insofar as we do not consider beliefs about one’s normative judgments to
themselves be normative judgments, we should not take first-person judgments
of rationality to be normative judgments.
Some readers will remain dissatisfied with this proposal, insisting that even
the more central sense of “rational” is more robustly normative than I have
allowed. To some extent I do not want to fight over the word “normative,” which
is anyway a term of art. I am happy to allow that “rational” even in its primary
sense might be normative in a sense close to the one I have proposed. In that
case, I would simply insist that it differs in ways from other normative notions
that makes an expressivist analysis less compelling. In the following section
I try to accommodate what I find insightful in the idea that thought and dis-
course about rationality are normative while at the same time explaining why
I favor a form of ecumenical cognitivism about such thought and discourse.
E C U ME N I C AL CO G NI TI V I SM R E V I SI TED
To meet the worry that thought and discourse about practical rationality are
more robustly normative than I have allowed, even in the most central senses
of “rational” and cognate terms, I must first elaborate my account further. For
I hold not only a cognitivist theory of “rational” and cognate terms but a form
of ecumenical cognitivism.
Claims about which actions are rational or irrational can be action guiding. In
Stevensonian terms, “rational” and cognate words have a kind of practical “mag-
netism.” My instrumentalist proposal can explain why this should be. For, given
my analysis, to tell someone that his or her proposed course of action is irrational
exerts a kind of pressure on the agent to either abandon the end(s) that make the
action irrational or abandon the proposed course of action. For from the agent’s
point of view, if a contemplated action really is irrational, then it follows that it
frustrates at least one end that he or she cherishes. The concern for that end alone
should in that case be enough to motivate him or her, at least to some extent, to
choose between those ends and revise his or her commitments accordingly.
The fact that claims about what would be irrational exert this sort of pressure
can already partly explain why such claims can seem to be normative in a more
robust sense. However, as Niko Kolodny (2005) has independently argued, this
is an illusion. Someone can, from the third-person point of view, simultane-
ously and coherently maintain that someone’s action is rational but insist that
he or she ought not perform it all the same. Ripley may not have been irrational
to commit homicide, but he ought not to have killed his friend all the same.
Likewise, someone can from the third-person point of view simultaneously and
coherently hold that a person’s action is irrational but judge that it was the right
thing to do all the same. We can coherently think, for example, that it would
be irrational, given his or her ends, for the jewel thief not to break into the
unguarded diamond store but judge that not breaking in was right.
So claims about what would or would not be rational do not settle how one
ought to act. This is not, however, to maintain that claims about rationality
are in no sense normative. They do not settle or even entail what to do, so they
are not normative in the primary sense of “normative” introduced above—
they are not sufficiently directly relevant to how a rational agent decides what
to do to count as robustly practically normative. However, such claims are
normative in a number of weaker but still quite interesting and useful senses
of “normative.” It is no accident that metanormative theorists have been inter-
ested in the concept of rationality.
First, plausibly at least part of the function of rational agency is to ensure
certain patterns of willing and behavior and prevent others. In which case, to
call someone irrational is in effect to announce that he or she is malfunctioning
qua rational agent. This is normativity in one useful sense of the word, even if
it is not normativity in a sense that settles “the thing to do.” We might contrast
this species of normativity with practical normativity by calling it “functional”
normativity. It is important to emphasize both the modesty and the ubiquity
of functional normativity. Any entity with a function can be characterized in
normative terms in this sense. Clocks, pencil sharpeners, cars, hearts, eyes, and
spectacles, just to take a few examples, are all functional kinds.
Second, as noted above, telling someone that his or her proposed course of
action is irrational can exert a sort of pressure on one’s interlocutor not to act in
that way. For if you are right that my action would be irrational, I can infer that
it would frustrate at least one of my ends. This should motivate me to recon-
sider my course of action, try to figure out (perhaps through further conversa-
tion) in what ways it might be irrational, and contemplate alternative courses of
action that might not frustrate any of my ends. So rationality talk has the sort
of Stevensonian “magnetism” that is often taken as a mark of the normative.
Third, at least some uses of “rational” and cognate terms implicate norma-
tive contents. In my view, the best model of this normative implicature is what
H. Paul Grice (1975) called a “generalized conversational implicature” or GCI.12

Like particularized conversational implicature, GCI can be canceled and does
not (simply by virtue of being an implicature anyway) figure in the truth condi-
tions or semantic content of the sentence whose utterance implicates it. Like
conventional implicature, we need no special stage setting for the implicature
to arise—it is, as it were, the default setting for the relevant sentences. Standard
examples of GCI include (a) “I got into a car” as implicating that the car was
not the speaker’s car and (b) the use of disjunction to implicate that for all the
speaker knows only one of the disjuncts is true without his or her knowing
which one.
The clearest cases in which “rational” and cognate terms are used with
normative implicatures are ones in which one party is explicitly deliberat-
ing about what to do (or believe). If in that context I tell my interlocutor
that the option (or belief) under consideration would be irrational, I thereby
implicate that he or she ought not do choose that option (or form that belief).
Intuitively, in saying that the option in question would be irrational without
any further caveats or commentary, I am thereby advising my interlocutor
against choosing that option. However, it would be insincere for me to advise
you to do something if I did not think it was the thing to do—that is, what
one ought to do in those circumstances, all things considered. So insofar as
I am being a cooperative interlocutor, one can derive from standard Gricean
maxims of relevance and the like that I do indeed believe that my interlocutor
ought to act accordingly.
This in turn can explain why such an utterance expresses a proattitude.
Suppose for these purposes we define expression as follows: a speech act
expresses a state of mind S if and only if the speaker would be either insincere,
deluded, or irrational in making that utterance if he or she were not in S. Insofar
as I am sincere and undeluded, I will believe that my interlocutor ought to act
accordingly, given the normative implicature of my speech act. Insofar as I am
myself rational, if I believe he or she ought to perform the action, then I will be
motivated to perform such an action if I find myself in his or her circumstances.
Note that in this sense it is speech acts, and not sentences, that express
thoughts and other states of mind. This is important, because the sense of
“express” that is relevant for understanding theories like ecumenical expres-
sivism is quite different. This reflects the fact that ecumenical expressivism is a
view in metasemantics, whereas the phenomena discussed here is pragmatic.
This account puts a lot of weight on the idea that certain uses of “rational”
and cognates function (in their default setting anyway) to offer advice. Why
should that be? Precisely because of the “magnetism” sketched above. It is
12. For a similar view in the moral case, see Strandberg 2012.
after all not only true but common knowledge that convincing someone that a
contemplated course of action (or belief) would be irrational will tend to lead
him or her not to perform the action (or form the belief). Given this common
knowledge, it would be uncooperative to tell someone without qualification
that a given option (or belief) would be irrational unless one thought it was
not the thing to do. That, though, is already enough to explain why such utter-
ances are in their default mode used to offer advice. Because telling someone
his or her action would be irrational exerts this sort of pressure, and transpar-
ently does, it is reasonable for one’s interlocutor to interpret the claim that an
action would be irrational as not only describing the action as a kind of rational
“malfunction” but as also prescribing against it. This point is important, because
even GCIs must be calculable from Gricean maxims in principle, even though
speakers do not normally need to calculate them.
I have focused on cases in which one’s interlocutor is explicitly deliberating,
but arguably the normative implicature I have sketched will arise in other cases.
After all, there is such a thing as general advice, as when I say things like “I
would advise anyone under thirty to avoid being head of department.” Perhaps
general claims about rationality can function to provide quite general advice
and thereby express more high-level normative judgments and associated pro-
attitudes. For example, an utterance of “It is irrational to inflict revenge for its
own sake” without any qualifications might express a quite general normative
thought and associated practical stance in the sense laid out above.
Unlike claims about what one ought to do, all things considered, rationality
claims’ expression of disapproval can be canceled. Precisely because irrational-
ity is such a normatively thin notion, it is not incoherent to say “It would be
irrational for you, given your ends, to do that, but you really should, all the
same, do it—your ends are morally perverse and self-destructive.” Such an
utterance cancels any suggestion that the speaker disapproves of the action in
question. To make this even more explicit, a speaker could, without a hint of
semantic or conceptual incoherence, add “nor do I in any way disapprove of
your doing it.”
What about backward-looking uses of “rational” and cognate terms? Well,
“irrational” in these contexts is plausibly a term of abuse, at least in its default
setting—again, this can be canceled. Ceteris paribus, to call someone or his or
her behavior irrational is to insult or at least to criticize him or her. Similarly,
ceteris paribus, to call someone rational (in the success sense) is to praise him
or her or at least to indicate that he or she is succeeding in one respect. Praise
and criticism are intimately associated with certain pro- and conattitudes—
praise with admiration, pride, and the like, and criticism with contempt, pity,
and various other negative attitudes. A similar story can therefore be told about
why we find the expression of attitude in these cases too.
On this account, not all uses of “rational” and cognates carry such norma-
tive/attitudinal implicatures. Speech acts in which these terms are embedded in
unasserted contexts will not necessarily have any such implicature, though this
will depend on the specific content of the claim. For example, a normal use of
“If Jones acted irrationally, then so did Smith” does not seem as such to impli-
cate any specific normative thought. However, a normal use of “If anything is
irrational, then starving yourself to death is” will have a normative implicature.
It is not hard to see why these cases differ in this way.
Even more clearly, speech acts in which “rational” and cognates are used in
propositional attitude ascriptions will not typically have such GCIs. My claim
that Jones thinks Smith was irrational does not itself indicate my own view
of what is irrational. Note the contrast with racial slurs, which are sometimes
used as models for ecumenical cognitivst theories. Plausibly the use of a racial
slur in an embedded context does still typically express a racist attitude, as
does the use of such slurs in propositional attitude ascriptions. This is one
reason that a GCI model is a better one for the form of ecumenical cognitiv-
ism that is true of rationality discourse than the more commonly invoked
model of racial slurs.13
In any event, the fact that discourse about practical rationality implicates
normative thoughts, has a practical and dynamic aspect, and can be used to
express various forms of approval and disapproval can explain why it might
well seem to have a literal normative content. This in turn provides a further
rejoinder to the worry that the account is not robustly normative enough
in its account of “rational.” The fact that the literal content of such claims
invokes a more low-grade “functional” form of normativity may also help
meet this worry.
CONCLUSION
In this chapter I have defended a form of what I call implicative ecumenical

cognitivism about thought and discourse about practical rationality. The view
13. So-called general pejoratives are perhaps a better model—words like “jerk” and “asshole.”
See Hay 2011. However, it is not clear whether the negative attitudes expressed by general
pejoratives can be canceled in the way I have suggested those expressed by characteristic uses
of “irrational” can. This in turn may also mean that the negative attitude is part of what must
be true of someone to count as believing that (for example) someone is a jerk; again, this is
not true in the case of “rational” and cognate terms. I can believe you are irrational but not
thereby disapprove of you. Also, on Ryan Hay’s account, there is no clear descriptive content
that can be “detached” from a general pejorative, but my reductionist account of “rational”
and cognate terms departs from this model.
on offer is broadly instrumentalist. Unlike its more Kantian rivals, it preserves

the right kind of link between “rational” in the capacity sense and “rational” in
the success sense, or at least so I have argued. I have also allowed that there are
derivative uses of “rational” that are more richly normative in their semantic
content but that these uses are less central and more amenable to an expressiv-
ist treatment.
A worry about this account is that it downplays the normativity of thought
and discourse about rationality. I have addressed this worry in several ways,
apart from allowing for a separate and fully normative sense of “rational.” For
a start, rationality thought is normative in the low-grade sense of “norma-
tive” associated with merely functional normativity. Furthermore, discourse
about practical rationality can in some contexts express a speaker’s normative
thoughts qua Gricean generalized conversational implicature. This is enough
to explain what is plausible in the worry about my account. I therefore con-
clude that this worry is misguided and that the proposed ecumenical cognitiv-
ist treatment of “rational” and cognates is sound.
References
Bagnoli, C. Forthcoming. “Starting Points: Kantian Constructivism Reassessed.” Ratio
Juris.
Davidson, D. 1980. Essays on Actions and Events. Oxford: Clarendon.
Dennett, D. 1987. The Intentional Stance. Cambridge, MA: MIT Press.
Gert, B. 1966. The Moral Rules. New York: Harper and Row.
Gert, J. 2004. Brute Rationality. Cambridge: Cambridge University Press.
Gibbard, A. 1992. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press.
Grice, H. 1975. “Logic and Conversation.” In: Syntax and Semantics, Vol. 3, Speech Acts,
ed. by Peter Cole and Jerry L. Morgan, 41–58. New York: Academic Press; here 45–47.
Hay, R. 2011. “Hybrid Expressivism and the Analogy between Pejoratives and Moral
Language.” European Journal of Philosophy 21: 1–25.
Hill, T. 1992. Dignity and Practical Reason in Kant’s Moral Theory. Ithaca, NY: Cornell
University Press.
Kolodny, N. 2005. “Why Be Rational?” Mind 114:509–563.
Korsgaard, C. 1996. The Sources of Normativity. Cambridge: Cambridge University
Press.
Parfit, D. 1984. Reasons and Persons. Oxford: Clarendon.
Ridge, M. 2005. “Why Must We Treat Humanity with Respect? Evaluating the Regress
Argument.” European Journal of Analytic Philosophy 1:57–74.
———. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116:302–336.
Savavarsdóttir, S. 2006. “Evaluations of Rationality.” In Metaethics after Moore, ed. T.
Horgan and M. Timmons, 61–78. Oxford: Oxford University Press.
Scanlon, T. 1998. What We Owe to Each Other. Cambridge, MA: Harvard University
Press.
Smith, M. 1994. The Moral Problem. Oxford: Blackwell.

Strandberg, C. 2012. “A Dual Aspect Account of Moral Language.” Philosophy and
Phenomenological Research 84:87–122.
Street, S. 2008. “Constructivism about Reasons.” In Oxford Studies in Metaethics, vol. 3,
ed. R. Shafer-Landau, 207–245. Oxford: Oxford University Press.
Tubert, A. 2010. “Constitutive Arguments.” Philosophy Compass 5:656–666.
2
Expressivism, Nondeclaratives,
and Success-Conditional
Semantics
DANIEL R. BOISVERT n
INT R OD U C T I O N
Theories of meaning explain how we understand the complex expressions of a lan-

guage by understanding their parts and the way those parts are combined.1 Some
theories of meaning accomplish this task for a language’s declarative sentences
(“Rocks are hard,” “Today is Tuesday”) by using machinery found in a suitably
constrained theory of truth. The assumption guiding this strategy is that if we con-
struct for an object language a theory of truth from which one can derive theorems
that specify the truth conditions of the object language’s sentences, but specify
them in a way that also interprets those object language sentences, we at the same
time exhibit how we can understand declaratives on the basis of understanding
their parts and they way they are combined.2 For example, if we can construct a
suitably constrained truth theory for Turkish from which one can derive (1) that
1. Earlier versions of this paper were presented to the Expressivism Reading Group at
the University of North Carolina, Chapel Hill (April 2012), and at the Hybrid Theories in
Meta-Ethics Conference in Edinburgh (July 2012). Thank you to those audiences for espe-
cially helpful comments, questions, and suggestions. For pressing me on several important
issues, I also thank Stephen Barker, Jamie Dreier, Mark Schroeder, Jon Tresan, and Peter
Vranas.
2. The idea traces to Davidson 1967. A different kind of truth-conditional theory of meaning
aims to accomplish this task by assigning entities to sentences (e.g., functions, intensions,
Expressivism, Nondeclaratives, and Success-Conditional Semantics 23
derivation would exhibit how we can understand the Turkish sentence (2) on the
basis of understanding its parts (kaya, serttir) and the semantic significance of the
way those parts are combined (e.g., Kaya serttir rather than Serttir kaya):
(1) Kaya serttir is true in Turkish if and only if rocks are hard.
(2) Kaya serttir.
We could then justifiably replace the predicate “is true in Turkish” and the con-
nective “if and only if ” with the predicate “means in Turkish that” to ultimately
derive a statement of what (2) means:
(3) Kaya serttir means in Turkish that rocks are hard.
Nondeclarative sentences (“Go home,” “Thank you!” “What time is it?”) and
more complex sentences containing them (“If you cleaned up the room, thank
you!”) frustrate a full-blown use of truth-conditional theories of meaning for
a natural language. For neither nondeclaratives nor sentences containing them
are truth evaluable; consequently sentences of neither type have truth condi-
tions; consequently constructing a truth theory to help explain our understand-
ing of these types of sentences appears bound to fail. Since nondeclaratives and
sentences containing them constitute a very large portion of natural languages,
the success of truth-conditional theories is limited.
Which brings us to the field of metaethics, a field infected by this limitation
of truth-conditional semantics. The carrier of this infection into metaethics is
expressivism, which holds that moral sentences (“Insulting others is wrong”)—
declarative though they may be—work in important ways like nondeclaratives
and so, like nondeclaratives, fail to have truth conditions in any sense sufficient
to adopt a truth-conditional account of their complete meanings. Expressivists
have thus been forced to explain our understanding of moral sentences by
adopting some other kind of meaning theory. For example, Simon Blackburn
and Alan Gibbard have adopted “mentalist” semantic theories, which explain
our understanding of sentences in terms of the mental states sentences express.3
Earlier expressivists, such as A. J. Ayer, appear to have assumed a “direct
structured composites) and their semantically significant subsentential components (e.g.,

properties, relations, functions) as their meanings. The theory of meaning I outline below
is Davidsonian and therefore does not assign entities as meanings. However the challenges
nondeclaratives pose for truth-conditional theories of meaning, which I outline presently,
arise for both types of truth-conditional theories.
3. See Blackburn 1984, 1988; Gibbard 1990, 2003.
illocutionary force” semantic theory, which explains our understanding of

sentences in terms of the direct illocutionary acts (e.g., directing, expressing,
asserting) that sentences are used to perform.4 At least when used to explain
our understanding of moral sentences, mentalist and direct illocutionary act
theories have faced some widely recognized difficulties of their own.5
Expressivists should adopt a new semantic strategy—or in their case perhaps
readopt an older one. For like earlier expressivists, current expressivists might
be better guided by taking seriously their analogy between moral and nonde-
clarative sentences. Prima facie, doing so is promising. For nondeclaratives are
certainly meaningful—both (4) and (5) are meaningful and differ in meaning,
a difference that can be explained only by a difference in the meanings of their
respective exclamative and declarative consequents.
(4) “If you ate all of your dinner, terrific!”

(5) “If you ate all of your dinner, I am pleased.”
Furthermore, nondeclaratives are embeddable into a wide variety of linguis-

tic contexts—into conjunctions (“Good night and good luck”), disjunctions
(“Congratulations, or did someone else win the race?”), consequents of con-
ditionals (“If you won the race, congratulations!”), and quantifiers (“What a
race!”) for example—and so there must be a compositional story that explains
how some sentences that lack truth conditions nevertheless embed mean-
ingfully into these contexts. Thus reflecting on an adequate meaning theory
for nondeclaratives may provide expressivists with much needed semantic
resources.
This chapter follows just this kind of strategy by probing the extent to which
expressivists can adopt what I now call success-conditional semantics (SCS).6
SCS exploits the intimate connections among sentence type (declarative,
imperative, interrogative, exclamative, or optative), speech-act type (asser-
tive, directive, expressive), and the types of conversational goals we usually
4. See Ayer 1936, chap. 6.

5. See, for example, Hale 1993; Schroeder 2008a, 2008b; Unwin 1999, 2001; van Roojen 1996.
6. The central insights of SCS are most fully developed and advanced under the names
“fulfillment-conditional” or “aptness-conditional” semantics in Ludwig 1997. The theory is
slightly modified and extended in Boisvert 1999; Boisvert and Ludwig 2006; and Lepore
and Ludwig 2007, chap. 12. Note that Lepore distances himself from this semantic account
of nondeclaratives. The second section of this chapter further extends SCS by clarifying the
nature of success (fulfillment, aptness) conditions and using them to explain the logical
relations among sentences of different syntactic moods. For earlier versions of this type of
semantic theory, see Harnish 1994, 431–437; McGinn 1977; Segal 1990–1991.
have when performing those types of speech acts—and explains our under-
standing of sentences in terms of the conditions that would make those typical
conversational goals successful. Here is the rough idea: “Yes!” is an exclama-
tive sentence, and exclamative sentences are conventional devices that allow
speakers to directly express noncognitive, affective attitudes. But speakers do
not typically want to express any old noncognitive attitude; rather speakers
typically want to express noncognitive attitudes they actually have—that is
the purpose of the exclamative mood. Thus an adequate success-conditional
theory of meaning for English would explain our understanding of “Yes!” in
terms of what would make the sentence’s conventional use sincere, namely,
that the speaker be excited at the time of utterance. Likewise, “Go home” is an
imperative sentence, and imperative sentences are conventional devices that
allow speakers to (directly) direct people’s behavior. But speakers do not typi-
cally want to direct people to do any old thing; rather speakers typically want
to direct people to do what is actually directed—that is the purpose of the
imperative mood. Thus an adequate success-conditional theory of meaning
for English would explain our understanding of “Go home” in terms of what
an addressee must do to comply with the sentence’s conventional use, namely,
that the addressee goes home later than the time of utterance (perhaps with
the intention of satisfying the directive issued).
Can expressivists adopt a success-conditional theory of meaning? I will
argue that hybrid expressivism and pure expressivism can indeed adopt SCS
but that only the former can do so without raising additional difficulties.7 For
unlike hybrid expressivism, pure expressivism can adopt SCS only by accepting
an objectionable relativism and increasing an already hefty explanatory burden,
and even then it may still be unable to avoid the full force of the Frege-Geach
problem. The result is an argument that if one is going to be an expressivist at
all, one should be a hybrid, rather than a pure, expressivist.8
All of this will be explained in due course, but I begin by laying some ground-
work. In the first part I use expressivism’s perennial challenge, the Frege-Geach
problem, to articulate two minimal requirements that any adequate meaning
7. Roughly, hybrid expressivism holds that moral sentences both express (affective states)
and assert, or have both expressive and descriptive meaning, or that moral thoughts are
complex mental states constituted by both an affective state and an ordinary belief. Pure
expressivism holds that moral sentences only express (affective states), or have only expres-
sive meaning, or that moral thoughts are unitary mental states constituted by only an affec-
tive state. More is said below about the differences between hybrid and pure expressivism.
8. Obviously, this argument is defeasible. For example, since it is hybrid, hybrid expressiv-
ism requires an ontology of moral properties, a difficulty that may be sufficient for one to
ultimately reject hybrid expressivism.
theory for a natural language must satisfy, which I call the “Contribution”
and “Logical Preservation” requirements. In the second part I explain why
success-conditional meaning theories satisfy these constraints better than
truth-conditional theories: the former can, while the latter cannot, explain how
we understand nondeclaratives or more complex sentences containing them. In
the third part I return to the central issue, the extent to which expressivists can
adopt a success-conditional theory of meaning.
E X P R E S S I V I S M AND THE FR E G E - G E ACH P ROBLEM
The Frege-Geach problem is the perennial problem for expressivism.9 What is

often overlooked, however, is that Frege-Geach is a problem for every meaning
theory for a natural language, since natural languages contain nondeclaratives.
The heart of the Frege-Geach problem is that there are (at least) two minimal
requirements for any adequate meaning theory for a language, and meaning
theories, including metaethical expressivist theories, have had a difficult time
jointly satisfying these.
Contribution Requirement. Natural languages are compositional in that we
understand complex sentences by understanding their parts and the syntactic
combination of those parts. Consider (6)–(11):
(6) The light is green.

(7) The light is not green.
(8) The light is green, or I am color blind.
(9) If the light is green, you may proceed.
(10) Jill believes that the light is green.
(11) It is possible that the light is green.
Sentences (7)–(11) embed (6). We understand (7)–(11) by understanding their

parts and the way those parts are combined. And we understand the same thing
by (6) when it occurs in any of (7)–(11). Since we understand (7)–(11) by under-
standing their parts, and since we understand the same thing by (6) as it occurs
in (7)–(11), an adequate meaning theory must tell us what we understand when
we understand “The light is green”—and what we so understand is therefore
what is contributed to our understanding of (7)–(11) and, for that matter, to
our understanding of any complex sentence into which (6) can embed. Thus
a minimal requirement on any adequate semantic theory for a natural language
9. See Geach 1958, 1960, 1965. See also Frege 1997; Ross 1939, chap. 2.
is the Contribution requirement. Contribution: an adequate meaning theory for

a language must specify, for any sentence of that language, the contribution that
sentence makes to our understanding of more complex sentences into which it
may embed.
Logical Preservation Requirement. A second minimal requirement on an ade-
quate meaning theory is what I will call the Logical Preservation requirement.
Logical Preservation: an adequate meaning theory for a language must respect the
intuitive logical relations among which the sentences of that language stand. (This
requirement will be slightly modified below.)
Sentences (12) and (13) are inconsistent; an argument from (12) and (14) to
(15) is valid; an argument from (13) and (14) to (15) is invalid.
(12) The light is green.

(13) The light is not green.
(14) If the light is green, you may proceed.
(15) You may proceed.
There are other kinds of intuitive logical relations that obtain among some sen-
tences of a language, although admittedly these latter relations are more difficult
to specify precisely. For example, some important logical relation obtains between
(17) and (18) as well as among (12), (16), and (17).
(16) If the light is green, go.

(17) Go.
(18) Don’t go.
The relation exemplified by (17) and (18) I will take to be inconsistency and that
exemplified by (12), (16), and (17) to be validity. Those who want to reserve “con-
sistency” and “validity” for relations grounded in truth preservation may of course
just use other names for these logical relations. The important point is that an
adequate meaning theory must respect these logical relations.10 A meaning theory
10. Peter Vranas, who for the past decade has been doing important work on imperative
logic, would hold that a transition from (12) and (16) to (17) is invalid. Vranas’s arguments
for this conclusion deserve a full response, which can be provided only in a full paper of
its own. I can say, however, that his arguments appear to rest on his claim that “conditional
imperatives,” such as (16), have not two but three kinds of semantic evaluation. For example,
(16) would be satisfied if the light is green and the addressee goes, violated if the light is green
and addressee does not go, avoided altogether if the light is not green. I think such evaluative
intuitions are insufficient to warrant a three-valued logic for conditional imperatives just as
similar intuitions are insufficient to warrant a three-valued logic (true, false, neither true nor
minimally respects these intuitive logical relations by not entailing their violation.
It ideally respects these intuitive logical relations by explaining them.
To see that these constraints are at the heart of the Frege-Geach problem,
consider a variant of P. T. Geach’s own objection to direct illocutionary force
theories of meaning:11
(19) Insulting others is wrong.

(20) If insulting others is wrong, then getting one’s little brother to
insult others is wrong.
(21) Therefore getting one’s little brother to insult others is wrong.
Geach observes that even assuming (19) and (21) are conventionally used to
directly express a speaker’s mental states or to directly prescribe behavior, they
are not conventionally used to perform these acts when they are embedded
in (20). Therefore direct illocutionary force cannot be what sentences contrib-
ute to our understanding of more complex sentences. One might suggest that
(19) and (21) do contribute direct illocutionary force to (20) but a force differ-
ent than they have when unembedded.12 On a direct illocutionary act theory
of meaning, however, that suggestion amounts to the claim that (19) and (21)
have different meanings when embedded than they do when unembedded and,
as Geach notes, therefore implies that (19)–(21) would violate modus ponens.
A direct illocutionary force meaning theory fails to jointly satisfy Contribution
and Logical Preservation.
Likewise for the more recent mentalist semantic theories. For example, Mark
van Roojen’s underlying challenge to higher order attitude theories is that while
such theories may satisfy Contribution, they violate Logical Preservation by
entailing the validity of some intuitively invalid arguments.13 Mark Schroeder’s
underlying challenge to a family of semantic theories like Gibbard’s is that while
constructed in part to satisfy Logical Preservation, they violate Contribution by
failing to specify the mental states that sentences express and thereby failing
to specify what the sentences of a language even mean.14 And since such theo-
ries fail to specify what the sentences of a language even mean, they fail to tell
false) for “conditional declaratives,” such as (14), whose antecedents are false. But see Vranas
2008, 534–535.
11. Geach 1965.
12. I do not know what force that would be, for intuitively it seems to have no force at all.
13. Van Roojen 1996.
14. Schroeder 2008a, 2008b.
us what they contribute to our understanding of more complex sentences into

which they embed.
So the Frege-Geach problem is a serious problem for expressivism. And
actually this should never have been surprising. For expressivism holds that
moral declaratives work in most important ways like nondeclaratives, and non-
declaratives also run head-long into the Frege-Geach problem. For example, a
meaning theory for a language that includes (22)–(27) will be adequate only if it
specifies what (22) and (23) contribute to our understanding of (24), (26), and
(27) and only if it respects, for the example, the inconsistency of (23) and (24)
and the validity of an inference from (25) and (26) to (23).
(22) Thank you.

(23) Come home.
(24) Don’t come home.
(25) The street lights are on.
(26) If the street lights are on, come home.
(27) If you cleaned up the room, thank you.
Still, if expressivism holds that moral declaratives work in most important

ways like nondeclaratives, one strategy for expressivism is to examine the more
promising meaning theories for nondeclaratives and determine to what extent
they may be able to adopt those theories to account for the meanings of moral
declaratives. One of the more promising theories for nondeclaratives is SCS, so
in the third part I examine the extent to which expressivism can adopt such a
theory. But first I will say more about SCS and how it satisfies the Contribution
and Logical Preservation requirements for nondeclaratives.
Direct Illocution, Sentential Mood, and SCS
An illocutionary act is, roughly, an act we perform in using a sentence for a spe-
cific purpose or point. We can use sentences to warn, congratulate, complain,
predict, command, apologize, inquire, explain, describe, request, bet, marry,
and adjourn, to list just a few specific kinds of illocutionary act. Several tax-
onomies place the variety of illocutionary acts into a small number of basic
types. Three of these basic types have been especially important in the meta-
ethical literature: assertives, in which we represent the way things are, as we
typically do when using declarative sentences such as “The street lights are on”;
directives, in which we direct people to do something, as we typically do when
using imperative sentences such as “Come home”; and expressives, in which we
express our noncognitive attitudes, as we typically do when using exclamatives
such as “Thank you!” Illocutionary acts can be direct or indirect. An indirect

illocutionary act is one we perform by performing another illocutionary act;
a direct illocutionary act is one we perform that is not indirect. For example,
we might use “It’s cold in here” on occasion to direct our audience to close the
window; in that case we would be doing so indirectly by directly asserting that
the room is cold. Similarly, we might use “Brrrr!” on occasion to assert that the
room is cold; in that case we would be doing so indirectly by directly expressing
the phenomenological state we are in when we are cold. Indirect illocution-
ary acts are irrelevant to semantics, but direct illocutionary acts are central.
For since sentences are conventional devices that we use to perform particular
types of direct illocutionary acts and since these, like any act, can succeed or fail
in various ways, understanding a sentence is to a great extent understanding the
various ways the direct illocutionary acts can succeed or fail.
If allowed to speak loosely, the most basic way a direct illocutionary act
can fail is by failing to be performed at all. A direct illocutionary act is per-
formed when a speaker utters a certain meaningful sentence (correctly, liter-
ally, etc.) with certain intentions. For example, I directly assert that rocks are
hard when I use the English sentence “Rocks are hard” with the intention of
describing rocks as being hard. I directly direct my addressee to come home
when I use the English sentence “Come home” with the intention of direct-
ing my addressee to come home at a time later than the time of utterance.
Thus a direct illocutionary act is “constitutionally” successful if its respec-
tive constitution conditions obtain and is constitutionally defective other-
wise. Another way an illocutionary act can be defective is by resting on a
false presupposition. My act of congratulating you for winning the race is
defective if you have not won the race. Likewise, my act of asserting that
Jill is going home after the party is defective if Jill is homeless or there is no
party. Thus a direct illocutionary act is “presuppositionally” successful if its
respective presupposition conditions obtain and is presuppositionally defec-
tive otherwise. A third way an illocutionary act may be defective is by being
performed insincerely. My expressing excitement when I use the sentence
“What a car!” is defective in this sense when your car does not actually excite
me; my act of asserting that rocks are hard is defective in this sense when
I do not actually believe that rocks are hard. Thus a direct illocutionary act is
“sincerity” successful when its respective sincerity conditions obtain and is
sincerity defective otherwise. And there are other ways direct illocutionary
acts could plausibly be evaluated as successful or defective.
For reasons to be explained shortly and from the point of view of SCS, the
most important way a direct illocutionary act may be defective is by failing to
achieve its typical conversational purpose. Typically, it is important to notice,
we do not just want to assert, direct, or express any old thing. That is:
typically, we do not just want to describe the world any old way; we want
to describe the world as it actually is;
typically, we do not just want to direct others to do any old thing; we want
to direct others to do that which they (the speakers) have actually directed;
typically, we do not just want to express any old attitude; we want to
express attitudes we actually have.
A direct illocutionary act, then, can be said to be “conversationally” suc-

cessful when its respective typical conversational purpose is achieved—that
is, when its conversational success conditions obtain—and conversationally
defective otherwise.15 Conversational success conditions for illocutionary acts
thus come in three varieties: truth (for assertives), compliance (for directives),
and sincerity (for expressives).
Of course theories of meaning are not designed to explain the pragmatic
phenomenon of direct illocution but rather to explain the semantic phenom-
enon of what sentences mean or, more precisely, how the meanings of sentences
depend on the meanings of their parts and the ways those parts are combined.
Still, a theory of meaning can exploit the close connection between sentence
types and the direct illocutionary acts types they are conventionally used to
perform. That is just what SCS does.
SCS for the Sentential Moods
Consider how SCS works for the various sentential moods, which SCS was
originally designed to explain. The sentences of English come in one of several
types, or moods. The three major sentential moods are the declarative (“The
street lights are on,” “Rocks are hard”), imperative (“Come home,” “Take two
aspirin”), and interrogative (“What time is it?” “Where are you going?”). The
two minor sentential moods are the exclamative (“Yes!” “Thank you!”) and the
optative (“How I wish you were here!” “If only I could be with you!”). There
are also complex sentences containing more than one mood (e.g., “If the street
lights are on, come home,” “Congratulations, or did someone else win the
race?”), which we will call “complex moods.”
15. Obviously, “conversational success” should not be taken to imply that a meaning theory
employing this kind of success can only explain our understanding of languages used in
conversation among more than one individual. A sentence can be used by a person when
conversing with or thinking to oneself, as, for example, I do when I “utter” or think “What
a great day!”
Declarative sentences are apt for performing direct assertives.16 SCS there-
fore aims to construct a theory of success for a language from which one can
derive theorems specifying its declaratives’ truth conditions, which they inherit
from the direct assertives they are apt for performing. Taking the declarative
“The street lights are on” as an example and relativizing to a speaker and time
of utterance, which we will abbreviate using the subscript “[s,t],” an adequate
success-conditional meaning theory for English would permit the following
abridged derivation:17
(i) “The street lights are on” is successful[s,t] in English iff “The street
lights are on” is true[s,t] in English;
(ii) “The street lights are on” is true[s,t] in English iff the street lights are on
at the time of utterance; (therefore)
(iii) “The street lights are on” is successful[s,t] in English iff the street lights
are on at the time of utterance.
Imperatives and interrogatives are apt for performing direct directives.

SCS therefore aims to construct a success theory for a language from which
one can derive theorems specifying the imperatives’ and interrogatives’
compliance conditions, which they inherit from the direct directives they
are apt for performing. Taking the imperative “Come home” as an exam-
ple and, again, relativizing to a speaker and time of utterance (and taking
obedience conditions to be a particular kind of compliance condition), an
adequate success-conditional theory would permit the following abridged
derivation:
16. A speaker can obviously misuse words or sentences, use words or sentences while staging
a play or telling a story or hypothesizing a scenario, and use words or sentences to perform
indirect illocutionary acts. Thus to say that a particular type of sentence is “apt” for perform-
ing a particular type of direct illocutionary act is roughly to say that a speaker will perform
a direct illocutionary act of that type if she or he uses the sentence correctly, literally, and
intentionally to perform all and only the direct illocutionary acts which that sentence is in
the language to help us perform. For more precise clarification, see Boisvert 2008, 174–177.
17. In what follows I offer the “chains of reasoning,” indicated with Roman numerals, as
abridged proofs leading to interpretative, metalanguage “S-sentences” of the form “ ‘S’ is
successful[s,t,] in L iff p,” where “p” is metalanguage sentence that interprets the object lan-
guage sentence mentioned or described by “S.” Like the proofs arising from interpretative
truth theories, these “S-sentences” are to be taken not only as specifying the success condi-
tions of the object language sentences. More importantly, their derivation is to be taken as
explaining (by exhibiting) how we can understand the respective object language sentences
by understanding their parts and the significance of the way those parts are combined.
(i) “Come home” is successful[s,t] in English iff “Come home” is obeyed[s,t]

in English;
(ii) “Come home” is obeyed[s,t] in English iff the speaker’s addressee
comes home at a time later than the time of utterance;18 (therefore)
(iii) “Come home” is successful[s,t] in English iff the speaker’s addressee
comes home at a time later than the time of utterance.
Likewise for exclamatives and optatives. Taking “Yes!” as our example, an ade-
quate success-conditional meaning theory would allow permit this derivation:
(i) “Yes!” is successful[s,t] in English iff “Yes!” is sincere[s,t] in English;

(ii) “Yes!” is sincere[s,t] in English iff the speaker is excited at the time of
utterance; (therefore)
(iii) “Yes!” is successful[s,t] in English iff the speaker is excited at the time
of utterance.
For atomic sentences, then, the SCS story can be summarized as follows.
Atomic sentences inherit their respective success conditions from the direct
illocutionary acts that they are particularly apt for performing. Since declara-
tive sentences are particularly apt for performing direct assertives, whose typi-
cal conversational purpose is to describe the world as it actually is, the success
conditions for declaratives are truth conditions. Since imperative (and inter-
rogative) sentences are particularly apt for performing direct directives, whose
typical conversational purpose is to direct others to do that which has actually
been directed, the success conditions for imperatives (and interrogatives) are
compliance conditions. And since exclamative (and optative) sentences are
particularly apt for performing direct expressives, whose typical conversa-
tional purpose is to express attitudes one actually has, the success conditions
for exclamatives (and optatives) are sincerity conditions. The foregoing con-
siderations thus help explain what might otherwise appear an odd disparity,
namely, that although sentences of any mood may be evaluated as sincere or
insincere only in the case of exclamatives (and optatives) does sincerity con-
stitute semantic success.
Let us turn to sentences containing logical connectives. Success conditions
for these are assigned recursively, and thus SCS nicely handles complex moods.
For example, consider an abridged derivation of the success conditions for “If
the street lights are on, come home”:
18. Arguably, the directive, and hence the imperative, will be obeyed only when the person
to whom the directive is issued carries out the directive with the intention of obeying it. For
brevity, I omit this detail in what follows.
(i) “If the street lights are on, come home” is successful[s,t] in English iff if
“the street lights are on” is successful[s,t] in English then “come home”
is successful[s,t] in English.
Given the success conditions of the respective atomic sentences as derived

above, then:
(ii) “If the street lights are on, come home” is successful[s,t] in English iff
if the street lights are on at the time of utterance then the speaker’s
addressee comes home at a time later than the time of utterance.
And similarly for other complex moods.

SCS thus holds that sentences contribute their success conditions to the
more complex sentences into which they can embed and thereby satisfies the
Contribution requirement. It also ideally respects Logical Preservation by
implying that success conditions explain the various logical relations among
sentences. For example, consider the following:

(29) The street lights are not on.
(30) Come home.
(31) Don’t come home.
(32) Thank you for cleaning the room!
(33) Shame on you for cleaning the room!
Sentences (28) and (29) are logically inconsistent. Intuitively, sentences (30)
and (31) and sentences (32) and (33) are also logically inconsistent in some
important sense. SCS explains the logical inconsistency of these sets as the
impossibility of joint success. Relative to a speaker and a time of utterance, it
is impossible for (28) and (29) both to be true[s,t], impossible for (30) and (31)
both to be obeyed[s,t], and impossible for (32) and (33) both to be sincere[s,t].19 By
grounding logical inconsistency in the more general notion of the impossibility
of joint success and assuming that sentences of different moods can be logically
inconsistent, then SCS also explains the logical inconsistency of many other
types of sentence sets, such as between (34) and (35), an inconsistency between
an interrogative and imperative, and between (36) and (37), and inconsistency
between a declarative and exclamative:
19. I am assuming it is not possible for a person to be both grateful to and disappointed in a
person for performing a particular act. If both are possible, then (32) and (33) are not logi-
cally inconsistent.
(34) What time is it?

(35) Don’t tell me what time it is.
(36) I am excited about you winning the race.
(37) Shame on you for winning the race!
One might think that a very specific type of sentence set prohibits explain-
ing logical inconsistency across moods as the impossibility of joint success.20
The sets consisting of (38) and (39) and of (41) and (42) are instances of this
particular type. Each consists of an imperative whose obedience is ruled out
by the truth of the declarative (and so it is impossible that they be jointly suc-
cessful), yet intuitively the imperative and the declarative are not logically
inconsistent, as suggested by respective utterance of (40) and (43), which seem
conversationally unproblematic:
(38) Don’t insult her.

(39) You will insult her.
(40) Well, you will do so anyway, but don’t insult her.
(41) Compliment her.
(42) You won’t compliment her.
(43) Well, you won’t do so, but compliment her.
Thus explaining logical inconsistency across moods by appealing to the impos-

sibility of joint success appears to overproduce cases of logical inconsistency.
My own sense is that such sets of sentences are in fact logically odd—I want
to say logically inconsistent—although my intuitions are by this point no doubt
theoretically influenced. But in any case, the objection rests on the assump-
tion that sentences are not logically inconsistent if uttering the sentences (for
example) as a conjunction is conversationally understandable. This assump-
tion is false; it is not uncommon for a speaker to utter logically inconsistent
sentences (for example) as a conjunction for an understandable conversational
purpose. For example, consider a manager who, having asked one of her or his
employees to take more initiative, utters (44):
(44) Well, I am both glad and not glad you contacted the client without
consulting me first.
20. See also Vranas (2008, 545), who argues that the conditional imperatives “If he loves
you, marry him” and “If he doesn’t love you, don’t marry him” cannot both be successful
even though “there is not even a hint of conflict between them.” However, Vranas’s conclu-
sion appears to rest on accepting a three-valued logic for conditional imperatives, which, as
I mentioned in note 9, I do not accept.
(45) <Gloss> I am glad you contacted the client without consulting me
first because doing so demonstrated initiative, but I am not glad
you contacted the client without consulting me first because you
did not have all of the requisite information you needed.
Or consider a person responding to a friend who has asked whether a particu-

lar shirt is red:
(46) Well, it’s red and it isn’t.

(47) <Gloss> It looks red when no light is directly hitting it, it does not
look red (it looks orange) when light is directly hitting it.
And so on. Similarly, consider the following conversational glosses of (40) and
(43), respectively:
(43*) <Gloss> I acknowledge that you will insult her no matter what
I say, but I feel obliged (in my role as a friend, parent, adviser, etc.)
to so direct you anyway: do not insult her.
(46*) <Gloss> I acknowledge that you will not compliment her no
matter what I say, but I feel obliged (in my role as a friend, parent,
adviser, etc.) to so direct you anyway: compliment her.
In other words, there is often a conversational purpose to uttering logically

contradictory sentences that is not only understandable but relatively easy to
work out. The rather understandable conversational purpose for uttering sen-
tences that constitute the putative counterexamples is that one feels obliged to
issue a particular directive in virtue of fulfilling some advisory role while also
acknowledging the apparent futility of issuing that directive. So I do not think
this very particular type of sentence set is sufficient to set aside the claim that
logical inconsistency is best explained by the impossibility of joint success.
Turn now to the next three sets of sentences:
(48) If the street lights are on, it is dark outside.

(50) It’s dark outside.
(51) If the street lights are on, come home.
(53) Come home.
(54) If you won the race, congratulations!
(55) You won the race.
(56) Congratulations!
A transition from (48) and (49) to (50) is valid. Transitions from (51) and (52)
to (53) and from (54) and (55) to (56) are also valid in an important sense. SCS
grounds the validity of these transitions in the preservation of success condi-
tions. Relative to a speaker and a time, (50) must be successful[s,t] (true[s,t]) if
(51) and (52) are successful[s,t,] (true[s,t]), sentence (53) must also be successful[s,t]
(obeyed[s,t]) if (51) is successful[s,t,] and (52) is successful[s,t,] (true[s,t]), and (56)
must be successful[s,t] (sincere[s,t]) if (54) is successful[s,t,] and (55) is successful[s,t,]
(true[s,t]). Thus SCS can, while truth-conditional semantics cannot, explain the
intuitive validity and invalidity of a variety of different kinds of logical transi-
tions, including those across moods.
SCS, then, provides a principled, unified, compositional theory of mean-
ing of the major, minor, and complex moods of the sentences of a language.21
It satisfies Contribution by holding that sentences contribute their respective
success conditions to the more complex sentences that embed them. It satis-
fies Logical Preservation by explaining, and thereby ideally respecting, logi-
cal inconsistency as the impossibility of joint success and by explaining, and
thereby ideally respecting, validity as the preservation of success conditions.
And it does these things in a Davidsonian spirit, holding that we gain insight
into our understanding of complex expressions by showing how to construct an
interpretative success theory without assigning entities to words and sentences
as their meanings.
This last point is worth repeating: SCS does not assign entities to expressions
to serve as their meanings. Stressing this point forestalls a possible objection to
SCS, namely, that it implies that some intuitively nonsynonymous sentences are
nevertheless synonymous. For example, consider (57) and (58):
(57) Thank you!

(58) I am grateful.
Assuming that “Thank you!” expresses gratitude, these two English sentences
have the same success condition: “Thank you!” is sincere[s,t] in English and “I
am grateful” is true[s,t] in English if and only if the speaker is grateful at the time
of utterance. These two sentences therefore have the same success condition,
namely, the speaker being grateful at the time of utterance. Now if a semantic
theory assigns as the meaning of a sentence, say, a function from possible world
states to the values “successful” or “unsuccessful,” then (57) and (58) must be
assigned the same function, since they will always have the same “success value”
21. Technically, it explains complex moods containing only the traditional logical connec-
tives. It does not yet touch on those containing nonlogical connectives, such as “because,”
etc.
for any possible world state (relative of course to the various speakers in those
possible world states). But if the theory must assign the same function to (57)
and (58) and if such functions are their meanings, then (57) and (58) are syn-
onymous, contrary to fact. This implausible implication, that (57) and (58) are
synonymous, is even more obvious if one thinks that success conditions are
meanings and consequently assigns success conditions themselves as the mean-
ings of sentences. For in that case there is nothing more to their meaning than
the speaker being grateful at the time of utterance. SCS might therefore be in
trouble were it to construe success conditions as entities of certain kinds. But
SCS avoids every objection of this sort, since, like Donald Davidson, it holds
that we gain insight into sentence meaning not by constructing entities that are
assigned to sentences as their meanings but by using a certain kind of deriva-
tion to exhibit how we understand a complex sentence on the basis of its parts
and their mode of combination.
SCS for Pure Expressivism
Can expressivists adopt SCS? First to pure expressivism, a family of views that
hold that a typical utterance of an ethical sentence is the performance of only
one direct illocutionary act where that one act is something other than an asser-
tive—usually either a direct expressive (e.g., Ayer, Blackburn, early Gibbard) or
a direct directive of a certain kind (e.g., J. J. C. Smart and, if taken to be a pure
prescriptivist, R. M. Hare).22 Pure expressivists can certainly use SCS to satisfy
Contribution. Consider the ethical sentence “Insulting others is wrong” and
assume an expressivist view that holds such sentences to work like exclama-
tives. Adopting SCS, pure expressivism would imply something like the follow-
ing abridged derivation:23
(i) “Insulting others is wrong” is successful[s,t] in English iff “insulting

others is wrong” is sincere[s,t] in English;
22. Although Blackburn and Gibbard would likely not categorize themselves in this way.
Rather, they would likely hold a different view of assertive acts as expressions of belief and
claim that the typical use of moral claims is the expression of a moral belief, where a moral
belief is not a representation-like belief. See also Hare 1952, 1963; Smart 1984.
23. Similar considerations would apply for pure expressivist views holding that ethical sen-
tences work very much like imperatives. I say “something like the following,” since it is open
to expressivists to hold different views about what specific noncognitive attitude is expressed
by utterances of ethical sentences.
(ii) “Insulting others is wrong” is sincere[s,t] in English iff the speaker

disapproves of insulting others at the time of utterance; (therefore)
(iii) “Insulting others is wrong” is successful[s,t] in English iff the speaker
disapproves of insulting others at the time of utterance.
The success conditions identified in (iii) are thus what would be contrib-
uted to more complex sentences that embed “Insulting others is wrong.” For
example:
(i) “If insulting others is wrong, I won’t do it” is successful[s,t] in English

iff if “insulting others is wrong” is successful[s,t] in English then “I
won’t do it” is successful[s,t] in English;
(ii) “Insulting others is wrong” is successful[s,t] in English iff “Insulting
others is wrong” is sincere[s,t] in English;
(iii) “Insulting others is wrong” is sincere[s,t] in English iff the speaker
disapproves of insulting others at the time of utterance;
(iv) “I won’t do it” is successful[s,t] in English iff “I won’t do it” is true[s,t] in
English;
(v) “I won’t do it” is true[s,t] in English iff the speaker will not insult others
at times later than the time of utterance; (therefore)
(vi) “If insulting others is wrong, I won’t do it” is successful[s,t] in English
iff if the speaker disapproves of insulting others at the time of
utterance, then the speaker will not insult others at times later than
the time of utterance.
And similarly for other more complex constructions.

Whether pure expressivism’s adoption of SCS can satisfy Logical Preservation
is more delicate, for its adoption entails the validity of intuitively invalid argu-
ments. For example, consider the transition from (59) to (60), a simplified
version of an intuitively invalid transition originally suggested to me by Jamie
Dreier:
(59) I disapprove of insulting others; (therefore)

We should set aside for the moment questions about relativism, to which we
will return, and notice that (60) must be successful[s,t,] (i.e., sincere[s,t]) if (59) is
successful[s,t,] (i.e., true[s,t]). That is, even if (59) and (60) are not synonymous,
the transition from (59) to (60) is at least success preserving and hence valid,
contrary to intuition. And of course once one gets the hang of these puta-
tive counterexamples, one can generate a number of others that appear more
disturbing. For example, the counterexample Dreier actually suggested to me

was the following:
(61) Everyone raised in the suburbs disapproves of lying;

(62) I was raised in the suburbs; (therefore)
(63) Lying is wrong.
If (61) and (62) were true[s,t,], then it must also be true that the speaker disap-
proves of lying and hence that (63) must also be sincere[s,t,]. This transition is
thus success preserving and hence valid, contrary to intuition.
Pure expressivism might be able to defend itself from this objection, but
doing so will depend on theoretical considerations about the correct logical
form of attitude attribution sentences, such as (59) and (61), which attri-
bute disapproval to certain groups of people. For notice that this objection
assumes, as we have also thus far been assuming, that an adequate seman-
tic theory must respect all intuitive logical relations. But really—this is the
slight modification of Logical Preservation—it must respect only all of the
intuitive logical relations that obtain in virtue of logical form. To clarify, con-
sider that several centuries ago the following transition would have been
intuitively invalid:
(64) Water is falling from the sky; (therefore)

(65) H2O is falling from the sky.
Although this transition would have been intuitively invalid, it is truth pre-
serving and hence valid. But it is not valid because of its logical form but rather
because of something else, in this case because of the way the world actually is.
Importantly, that this truth-preserving transition would have been intuitively
invalid is not, I take it, an indictment of any truth-conditional semantic theory
for declaratives. The point more generally is that the validity of a transition may
sometimes turn out to be a surprise, especially if the validity is a result of some-
thing other than logical form. But our surprise by itself should not indict pure
expressivism’s adoption of SCS, just as “surprise” in the water-H2O case should
not indict adoption of truth-conditional semantics. The issue for pure expres-
sivism, then, is whether these putative Dreier-like counterexamples would be
valid in virtue of their logical form.
What is their correct logical form? The answer depends in turn on the
correct logical form of attitude attribution sentences, such as (59) and (61).
Simplifying and using the argument from (59) to (60) as our example, one
possibility is the following:
(59’) I disapprove of φ-ing; (therefore)

(60’) φ-ing is wrong.
Given our pure expressivist construal that instances of (60’) express dis-
approval, all instances of this argument form would indeed preserve success
and hence be valid in virtue of their logical form. In that case expressivism’s
adoption of SCS would indeed fall foul of the (now slightly modified) Logical
Preservation requirement. On the other hand, and again simplifying, a different
possible logical form is the following:
(59*) I ψ φ-ing; (therefore)

(60*) φ-ing is wrong.
But not every instance of this form would preserve success. For example:
(59#) I approve of lying; (therefore)

(60#) Lying is wrong.
Thus whether these Dreier-like putative counterexamples would preserve

success because of their logical form—and consequently whether pure expres-
sivism’s adoption of SCS would fail to minimally respect Logical Preservation—
depends on what is the correct logical form of attitude attribution sentences.
This question need not be resolved here, for pure expressivists face other chal-
lenges should they want to adopt SCS.
The most obvious concern, which has already reared its head while consider-
ing the Dreier-like putative counterexamples, is that pure expressivism’s adop-
tion of SCS would commit pure expressivism to a very strong kind of relativism
that most, including its most ardent supporters, such as Blackburn and Gibbard,
would want to reject. For in that case the truth of (59) or of (61) and (62) would
entail that lying is wrong and hence lying being wrong would “depend,” in a
very strong sense of “depend,” on of what a certain group of people disapprove.
Again, this is something that most pure expressivists would reject.
Of course pure expressivists who are not troubled by such relativism might
still want to adopt SCS. But even these pure expressivists face a third roadblock,
one that does not so much arise from adopting SCS as it is made worse by
doing so. For in adopting SCS it aligns the semantics of ethical sentences even
more closely to that of exclamatives or imperatives. The problem is that ethical
sentences, on the one hand, and exclamatives and imperatives, on the other, are
not everywhere equally embeddable, a disparity that becomes even more dif-
ficult to explain the more their respective semantics are taken to be similar. For
example, declaratives, including moral declaratives, but neither exclamatives
nor imperatives, are embeddable as antecedents of certain conditionals (“If go

home, then I’m tired”),24 within scopes of negations (“It is not the case that
yes!”), and as complements of modal operators (“It is possible that yes!”) and
attitude verbs (“Jill believes that yes!”). Thus pure expressivists who adopt SCS
increase the already hefty burden of providing a principled explanation of why
ethical sentences, but neither exclamatives nor imperatives, are embeddable in
these linguistic contexts.25
Thus even if pure expressivism respects Logical Preservation and even if pure
expressivism accepts a very strong form of moral relativism, it likely would
not want to adopt SCS, since it is unclear how it would bear the increase of
an already hefty explanatory burden. Pure expressivism remains a theory in
semantic trouble.
SCS for Hybrid Expressivism
Hybrid expressivism is a family of views that hold at least one of three theses,
one about speech acts, a second about meaning, a third about thought:
(H1) The direct illocutionary acts performed in correctly and literally

using an ethical sentence are at least two: at least one assertive and
at least one nonassertive.
(H2) The meaning of an ethical sentence is both descriptive (truth
conditional) and expressive; that is, its meaning is not exhausted by
its truth conditions.
(H3) Moral thoughts are constituted by both a desire-like attitude and a
representation-like belief; that is, they are logically complex.
I have argued for a hybrid view, which I call “expressive-assertivism,” that

accepts all of (H1)–(H3) and holds that moral language works in other impor-
tant ways, like slurs and pejoratives plausibly do.26 For example, a slur claim,
such as “Dillon is a ***,” plausibly both asserts that Dillon is of a certain group
R and expresses contempt for Rs. Moreover, a slur claim plausibly expresses
24. Some conditional constructions take imperative antecedents, such as “Go home only if
you’re tired.” See Vranas 2008, 544.
25. It is sometimes suggested that the principled explanation is an appeal to sentential mood,
that is, an appeal to the fact that these constructions take only declarative sentences. The
response simply restates the problem: Why do these linguistic contexts embed only declara-
tive sentences if they are so semantically similar to exclamatives or imperatives?
26. Boisvert 2008, forthcoming.
contempt when it occurs in any extensional context. For example, “If Dillon
is a ***, he won’t be invited” plausibly expresses contempt for Rs even though
“***” appears in the antecedent of the conditional. Thus “Dillon is a ***” has
both descriptive (truth-conditional) meaning and expressive meaning, and the
contemptuous thought that Dillon is a *** is constituted by both the belief that
Dillon is R and contempt for Rs. In the remainder I will assume a similar hybrid
view of moral language and thought.27
Can adopting SCS help this kind of hybrid view jointly satisfy Contribution
and Logical Preservation? And even if so, might it still run into challenges with
relativism and an increasing explanatory burden? Adopting SCS will certainly
help hybrid expressivism satisfy Contribution. Because the resulting theories
would hold that ethical sentences are conventional devices used to perform
two direct illocutionary acts, one assertive and one expressive, they would also
hold that any sentence in which an ethical predicate appears in an extensional
context has at least two kinds of success conditions—truth conditions and sin-
cerity conditions. For example, on such a view the following even-numbered
sentences would be rendered as their respective odd-numbered sentences, and
thus these sentences will all have at least two success conditions:

(67) Insulting others has the wrong-making property F; down with
F-things!
(68) Insulting others is wrong, or I’ve been raised on a lie.
(69) Insulting others is F, or I’ve been raised on a lie; down with
F-things.
(70) If insulting others is wrong, I won’t insult others.
(71) If insulting others is F, then I won’t insult others; down with
F-things!
As (67) suggests, (66) would be true[s,t] just in case insulting others has the
wrong-making property F and sincere[s,t] just in case the speaker disapproves
of insulting others. As (69) suggests, (68) is true[s,t] just in case either insulting
others is F or the speaker has been raised on a lie and is sincere[s,t] just in case
the speaker disapproves of F-things. As (71) suggests, (70) is true[s,t] just in case
if insulting others is F, then the speaker will not insult others and is sincere[s,t]
just in case the speaker disapproves of F-things.
27. David Copp (2001, 2009) holds (H1) and (H2) but explicitly rejects (H3). Michael Ridge
(e.g., 2006, 2007) holds at least (H3). Stephen Barker (2000) and Ryan Hay (2011) each hold
at least (H2).
Since on this view moral declaratives have two success conditions, they
contribute these two conditions to more complex sentences that embed them.
However, as (69) clarifies, the sincerity conditions of (68) are contributed to
(68) as a whole and not more specifically to its first disjunct; likewise, as (71)
clarifies, the sincerity conditions of (70) are contributed to (70) as a whole and
not more specifically to its antecedent. It is worth pausing over this point, for it
might seem that we are treating the respective expressive components of these
sentences inappropriately by treating them as semantically and logically dis-
tinct. I think reflection on other kinds of complex natural language construc-
tions, for example, conditionals, justifies our treating the expressive component
as distinct semantic and logical components. In fact I think reflection on many
other natural language constructions shows that the practice is ubiquitous. For
example, consider (72), which contains the appositive “which is common”:
(72) If insulting others, which is common, causes emotional harm,

I won’t do it.
Intuitively, (72) conditions the speaker’s restraint only on whether insulting

others causes emotional harm, not on both its causing emotional harm and its
being common. That is, intuitively an utterance of (72) is the performance of
two distinct, direct assertives, as if it were an utterance of (73); it is not the per-
formance of one direct assertive with a conjunctive antecedent, and so it is not
as if it were an utterance of (74):
(73) If insulting others causes emotional harm then I won’t do it;

insulting others is common.
(74) If insulting others causes emotional harm and is common, then
I won’t do it.
The appositive in (72), “which is common,” is thus treated appropriately as a

distinct semantic and logical component of (72) as a whole, not as a semantic
and logical component of its antecedent. Likewise for expressive adjectives, for
example, as “damn” occurs in (75):
(75) If I have to wash the damn dishes, I’ll be late to the party.
Intuitively, (75) conditions the speaker’s being late to the party only on her or
his having to wash the dishes, not on her or his having to wash the dishes and
being frustrated at having to do so. That is, intuitively, an utterance of (75) is the
performance of two distinct, direct illocutionary acts, one (complex) assertive
and one expressive, as if it were an utterance of (76); it is not the performance
of one direct (complex) illocutionary act with a conjunctive antecedent, and so

it is not as if it were an utterance of (77):
(76) If I have to wash the damn dishes, I’ll be late to the party; damn
those dishes!
(77) If I have to wash the dishes and damn those dishes, I’ll be late to
the party.
Again, the expressive adjective in (75) is thus treated appropriately as a dis-

tinct semantic and logical component of (75) as a whole, not as a semantic and
logical component of its antecedent. Likewise also for honorifics, such as “Your
Honor” as used in (78) and slur terms as used in a construction like (79):
(78) If Your Honor is ignoring relevant case law, your decision will be
overturned on appeal.
(79) If Dillon is a ***, then he won’t be invited.
The expression of respect in an utterance of (78) and the expression of con-

tempt in an utterance of (79) would each be treated appropriately as a distinct
semantic and logical component of the whole sentence, not a semantic and log-
ical component of their respective antecedents. These phenomena even appear
across syntactic moods. For example, (80) and (82) should be treated semanti-
cally and logically as (81) and (83), respectively:
(80) Does Tom, who is tall, have enough leg room?

(81) Does Tom have enough leg room; Tom is tall.
(82) Wash the damn dishes.
(83) Wash the dishes; damn it!
Thus assuming that use of moral terms are in part the performance of a distinct
direct expressive illocutionary act, then there is nothing extraordinary, ad hoc, or
otherwise inappropriate about treating their expressive component as a distinct
semantic and logical component of the entire sentence in which they occur.28
Let us now return to (70), the conditional “If insulting others is wrong, then
I won’t insult others,” which, according to the suggested hybrid theory on offer,
we understand as (71):
28. At least when such words appear in an extensional context. Things might be otherwise
when they appear in intensional contexts, such as when embedded within attitude-attribution
verbs like “believes that.” See Boisvert 2008.
(71) If insulting others is F, I won’t insult others; down with F-things!
According to SCS, then, this sentence would be successful[s,t,] in English if and

only if (71) is successful[s,t,] in English and (71) is successful[s,t,] in English iff and
only both of its components are successful[s,t,] in English. Adopting SCS, then,
would permit the following abridged derivation:
(i) “If insulting others is wrong, then I won’t insult others” is

successful[s,t,] in English iff “If insulting others is F, then I won’t insult
others” is successful[s,t,] in English and “Down with F-things!” is
successful[s,t,] in English;
(ii) “If insulting others is F, then I won’t insult others” is successful[s,t,] in
English iff if “insulting others is F” is successful[s,t,] in English then “I
won’t insult others” is successful[s,t,] in English;
(iii) “Insulting others is F” is successful[s,t,] in English iff “Insulting others
is F” is true[s,t] in English;
(iv) “Insulting others is F” is true[s,t] in English iff insulting others is F;
(v) “I won’t insult others” is successful[s,t,] in English iff “I won’t insult
others” is true[s,t] in English;
(vi) “I won’t insult others” is true[s,t] in English iff the speaker does not
insult others at a time later than the time of utterance;
(vii) “Down with F-things” is sincere[s,t] in English iff the speaker
disapproves at the time of utterance of F-things; (therefore)
(viii) “If insulting others is wrong, then I won’t insult others” is
successful[s,t,] in English iff [if insulting others is F then the speaker
does not insult others at a time later than the time of utterance, and
the speaker disapproves at the time of utterance of F-things].
Thus a hybrid expressivist who adopts SCS can hold that ethical sentences
contribute both truth conditions and sincerity conditions to the more complex
sentences containing them and thereby satisfy Contribution.
It also satisfies Logical Preservation. Consider the intuitive inconsistency
between “Insulting others is wrong” and “Insulting others is not wrong,” which
on the hybrid view are rendered (84) and (85):
(84) Insulting others is F; down with F-things!

(85) Insulting others is not F; down with F-things!
The inconsistency is explained as the impossibility of joint truth[s,t,] and hence of

joint success[s,t,]: it is impossible for insulting others to be both F and not-F, so
it is impossible for both (84) and (85) to be true[s,t,], so it is impossible for both
(84) and (85) to be successful[s,t,].29
Now consider Logical Preservation in light of the following intuitively valid
transition:
(86) If insulting others is wrong, I won’t insult others;

(87) Insulting others is wrong; (therefore)
(88) I won’t insult others.
According to the hybrid view that adopts SCS, this argument would be
rendered as:
(89) If insulting others is F, I won’t insult others; down with F-things;

(90) Insulting others is F; down with F-things; (therefore)
(91) I won’t insult others.
Clearly, (91) must be successful[s,t] (i.e., true[s,t]) if (89) and (90) are successful[s,t,]
(i.e., both true[s,t] and sincere[s,t]), and hence such a transition would preserve
success. Consider also a different intuitively valid argument in which the con-
clusion is a moral declarative:
(92) If insulting others causes unnecessary emotional pain, then

insulting others is wrong;
(93) Insulting others causes unnecessary emotional pain; (therefore)
This transition would be rendered as:
(95) If insulting others causes unnecessary emotional pain, then

insulting others is F; down with F-things;
(96) Insulting others causes unnecessary emotional pain; (therefore)
29. Perhaps another way to understand the inconsistency is in terms of the impossibility of
joint sincerity:
(i) Insulting others is F; down with F-things;

(ii) Insulting others is F; okay for F-things!
In this case as well, then, it is impossible for the pair to be jointly successful[s,t].
Again, (97) must be successful[s,t] (i.e., true[s,t] and sincere[s,t]) if (95) is

successful[s,t] (i.e., true[s,t] and sincere[s,t]) and (96) is successful[s,t] (i.e., true[s,t]).
Hence (94) must be successful[s,t] if (92) and (93) are successful[s,t]. Again, intui-
tive validity is ideally respected.
Adopting SCS also allows hybrid expressivism to nicely handle the Dreier-like
putative counterexamples. Recall this putative counterexample:

On the hybrid view under discussion, this transition would be rendered:

And it is certainly possible for (101) to be unsuccessful[s,t,] even if (100) is

successful[s,t,] (i.e., true[s,t,]). For example, it is perfectly possible for it to be false
that insulting others is F—and hence for (101) to be unsuccessful[s,t]—even if
everyone from the speaker’s hometown, and hence the speaker, disapproves of
insulting others.30 Thus this view ideally respects the intuitive invalidity of these
transitions and so ideally respects Logical Preservation.
Example (100)–(101) also clarifies why a hybrid view (of the kind I have
described) that adopts SCS is not by that fact committed to an objectionable
relativism. For that view does not imply that insulting others is wrong sim-
ply because a particular group of agents disapproves of insulting others; to be
wrong, insulting others would need to be F.
Finally, unlike pure expressivism, whose adoption of SCS increases its
burden to explain why moral declaratives and nondeclaratives are unequally
embeddable, hybrid expressivism easily discharges this burden. Recall that
declaratives, including moral declaratives, but neither exclamatives nor imper-
atives, are embeddable as antecedents of some conditionals (“If go home, then
I’m tired”), within scopes of negations (“It is not the case that yes!”), and as
complements of modal operators (“It is possible that yes!”) and attitude verbs
(“Jill believes that yes!”). Hybrid expressivism can plausibly claim that such
contexts require sentences that describe, and all declaratives, including moral
declaratives, describe.
30. I am assuming of course that F-ness is not disapproval of insulting others, that is, I am
assuming that relativism is false.
CONCLUSION
The Frege-Geach problem is the semantic problem of jointly satisfying the

Contribution and (modified) Logical Preservation requirements. It is the cen-
tral problem facing metaethical expressivism, which holds that moral declara-
tives, like nondeclaratives, are at least in part expressive or prescriptive. What
is not widely appreciated is that, since natural languages contain sentences that
are not wholly truth evaluable—most importantly nondeclaratives and more
complex sentences containing them—but yet stand in what are intuitively logi-
cal relations, the Frege-Geach problem is a problem for every semantic theory
that seeks to explain very large portions of natural languages. SCS overcomes
the Frege-Geach problem for nondeclaratives and so offers hope to metaethi-
cal expressivism. Unfortunately, hope runs dry for pure expressivism. For even
if pure expressivism’s adoption of SCS jointly satisfies the Contribution and
Logical Preservation requirements, its adoption of SCS commits it to a very
strong form of moral relativism that most pure expressivists want to reject and
increases an already hefty explanatory burden. Hope is more apt for hybrid
expressivism. For if hybrid expressivism of the kind described here adopts SCS,
it avoids these challenges. That it does is a strong reason to be a hybrid rather
than a pure expressivist, assuming that one wishes to be an expressivist at all.
References
Ayer, A. J. 1936. Language, Truth, and Logic. London: Victor Gollancz.
Barker, Stephen. 2000. “Is Value Content a Component of Conventional Implicature?”
Analysis 60 (3): 268–279.
Blackburn, Simon. 1984. Spreading the Word: Groundings in the Philosophy of Language.
Oxford: Clarendon.
———. 1988. “Attitudes and Contents.” Ethics 98 (3): 501–517.
Boisvert, Daniel R. 1999. “Pragmatics and Semantics of Mixed Sentential Mood
Sentences.” MA thesis, University of Florida.
———. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89 (2): 169–203.
———. Forthcoming. “Expressivism and the Hybrid Option.” In Arguing about Ethics,
ed. Joshua Glasgow. New York, NY: Routledge.
Boisvert, Daniel R., and Kirk Ludwig. 2006. “Semantics for Nondeclaratives.” In The
Oxford Handbook of Philosophy of Language, ed. Ernest Lepore and Barry Smith,
864–892. Oxford: Oxford University Press.
Copp, David. 2001. “Realist Expressivism: A Neglected Option for Moral Realism.”
Social Philosophy and Policy 18 (2): 1–43.
———. 2009. “Realist Expressivism and Conventional Implicature.” In Oxford Studies
in Metaethics, vol. 4, ed. Russ Shafer-Landau, 167–202. Oxford: Oxford University
Press.
Davidson, Donald. 1967. “Truth and Meaning.” Synthese 17 (1): 304–323.
———. 2001. Inquiries into Truth and Interpretation. Oxford: Clarendon.

Frege, Gottlob. 1997. “Negation.” In The Frege Reader, ed. Michael Beaney, 346–361.
Malden, MA: Blackwell.
Geach, P. T. 1958. “Imperative and Deontic Logic.” Analysis 18 (3): 49–56.
———. 1960. “Ascriptivism.” Philosophical Review 69 (2): 221–225.
———. 1965. “Assertion.” Philosophical Review 74 (4): 449–465.
Gibbard, Allan. 1990. Wise Choices, Apt Feelings: A Theory of Normative Judgment.
Cambridge, MA: Harvard University Press.
———. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Hale, Bob. 1993. “Can There Be a Logic of Attitudes?” In Reality, Representation, and
Projection, ed. John Haldane and Crispin Wright, 337–361. New York: Oxford
University Press.
Hare, R. M. 1952. The Language of Morals. Oxford: Clarendon.
———. 1963. Freedom and Reason. Oxford: Clarendon.
———. 1970. “Meaning and Speech Acts.” Philosophical Review 79 (1): 3–24.
Harnish, Robert. 1994. “Mood, Meaning, and Speech Acts.” In Foundations of Speech
Act Theory, ed. Savas L. Tsohatzidis, 407–459. London: Routledge.
Hay, Ryan. 2011. “Hybrid Expressivism and the Analogy between Pejoratives and Moral
Language.” European Journal of Philosophy 21 (3): 450–474. doi: 10.1111/j.1468-037
8.2011.00455.x.
Lepore, Ernest, and Kirk Ludwig. 2007. Donald Davidson’s Truth-Theoretic Semantics.
Oxford: Oxford University Press.
Ludwig, Kirk. 1997. “The Truth about Moods.” Protosociology: Cognitive Semantics I,
Conceptions of Meaning 10: 19–66.
McGinn, Colin. 1977. “Semantics for Nonindicative Sentences.” Philosophical Studies
32 (3): 301–311.
Ridge, Michael. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116
(2): 302–336.
———. 2007. “Ecumenical Expressivism: The Best of Both Worlds.” In Oxford Studies in
Metaethics, vol. 2, ed. Russ Shafer-Landau, 51–76. Oxford: Oxford University Press.
Ross, W. D. 1939. Foundations of Ethics. Oxford: Clarendon.
Schroeder, Mark. 2008a. Being For: Evaluating the Semantic Program of Expressivism.
New York: Oxford University Press.
———. 2008b. “Expression for Expressivists.” Philosophy and Phenomenological Research
76 (1): 86–116.
Segal, Gabriel. 1990–1991. “In the Mood for a Semantic Theory.” Proceedings of the
Aristotelian Society 91: 103–118.
Smart, J. J. C. 1984. Ethics, Persuasion, and Truth. London: Routledge and Kegan Paul.
Unwin, Nicholas. 1999. “Quasi-Realism, Negation, and the Frege-Geach Problem.”
Philosophical Quarterly 49 (196): 337–352.
———. 2001. “Norms and Negation: A Problem for Gibbard’s Logic.” Philosophical
Quarterly 51 (202): 60–75.
van Roojen, Mark, 1996. “Expressivism and Irrationality.” Philosophical Review 105
(3): 311–335.
Vranas, Peter, 2008. “New Foundations for Imperative Logic I: Logical Connectives,
Consistency, and Quantifiers.” Noȗs 42 (4): 529–572.
3
Can a Hybrid Theory Have It

Both Ways? Moral Thought, Open
Questions, and Moral Motivation
DAVID COPP n
Realist-expressivism is a hybrid of moral realism and expressivism.1 It holds that

simple moral assertions standardly express both a basic moral belief and a corre-
sponding conative state.2 For example, a person who asserts “Torture is wrong” in
standard circumstances thereby expresses a belief that torture is wrong as well as a
kind of disapproval. According to the specific version of realist-expressivism that
I propose, such a person thereby expresses commitment to a policy of opposing
and avoiding wrongful acts.3 I maintain that this is so in virtue of conventions gov-
erning the use of moral terms, such as “wrong.” This position combines the chief
doctrines of moral realism with a central positive view of moral expressivism. It
seems to me that the position has clear advantages over standard kinds of moral
realism without having any of the problems faced by noncognitivist expressivism.
In this chapter I respond to two objections. First is the objection that my posi-
tion cannot account for the element of endorsement that distinguishes moral
1. I thank the participants in the Hybrid Theories Conference held at the University of
Edinburgh, July 2012, for helpful comments and suggestions. I am especially grateful to
Kent Bach, Matt Bedke, Dan Boisvert, Steve Findlay, Guy Fletcher, Michael Ridge, Mark
Schroeder, and Jon Tresan.
2. See Copp 2007b. This article was originally published in Social Philosophy and Policy 18
(2001): 1–43. See also Copp 2009. A basic moral belief is a belief, for some moral property
Mness, to the effect that something is M.
3. Copp 2007b. For the idea of a policy, see Bratman 1989.
thought from thought about nonnormative matters. Second is the objection that
my position cannot accommodate the so-called inference-licensing property
of valid arguments and that, like other similar hybrid theories, it is committed
to an implausible assumption about moral belief, the so-called Big Hypothesis.
Both objections, but especially the second, are found in a systematic appraisal of
hybrid theories by Mark Schroeder (2009). Schroeder’s reasoning seems to show
that hybrid theories face difficult challenges in accounting for the logic of moral
arguments. Moreover, they cannot achieve the advantages they seemed to promise
unless they accept the Big Hypothesis. If he is correct, hybrid theories are prob-
lematic. I shall argue that he is not correct.
In the first section I outline the main ideas of realist-expressivism. In the second
section I explain the apparent advantages of the theory, and I propose a response
to the first objection, the objection about moral thought. In the remaining sec-
tions of the chapter I respond to Schroeder’s objections about the logic of moral
arguments. In the third section I show that Schroeder’s argument depends on the
“inference-licensing thesis.” In the fourth section I distinguish two interpretations
of this thesis and argue that, on the interpretation needed for Schroeder’s argu-
ment, it is false. In the fifth section I investigate what is left of his argument given
that its key premise is false. In the sixth section I argue that Schroeder is correct to
think that the Big Hypothesis is false. In the seventh section, however, I argue that
hybrid theories do not need the hypothesis. The final section is a brief conclusion.
T H E P R OP O S A L
I was led to realist-expressivism by reflecting on Allan Gibbard’s suggestion in

Wise Choices, Apt Feelings (1990) that “descriptivistic analyses” of moral belief
cannot account for the element of endorsement found in moral thought and
language. Gibbard (1990, 33) claims that “the special element that makes nor-
mative thought and language normative” is that “it involves a kind of endorse-
ment.” On reflection it seemed to me that Gibbard’s basic idea is plausible. It
seemed plausible that a person making a moral assertion thereby expresses
a relevant conative attitude and not merely a belief. And it seemed plausible
that this is due to the meanings of the moral predicates—in the broad sense of
“meaning” I will be using in this chapter.4
The idea can be brought out with an example (Copp 2009, 167). Anna
says that capital punishment is morally wrong, yet whenever she hears that a
4. There is a view according to which the “meaning” or “semantic content” of an expression

is exhausted by the expression’s contribution to the truth conditions of sentences in which it
is used. On this understanding, my view that the assertion of certain sentences containing
Can a Hybrid Theory Have It Both Ways? 53
criminal is about to be executed, she is indifferent. There is never any indication

that she disapproves of capital punishment other than the fact that she says it
is wrong. She would in fact concede that she has no negative attitude toward
it, nor has she any policy that would lead her to oppose it. Something is amiss.
If Anna were to claim that capital punishment is morally wrong but then add
that she has no negative attitude toward it, we would be puzzled. We might
doubt the sincerity of her claim. Perhaps she does not really believe that capital
punishment is wrong. Perhaps she does not know her own mind. Or perhaps
she does not have a thorough understanding of what she is saying when she
says capital punishment is wrong. Perhaps she thinks she is merely categorizing
it—bringing it under a descriptive category.
The important point, it seemed to me, is that moral realists and other cogni-
tivists have no need to dispute Gibbard’s basic idea. They can agree that a per-
son making a moral assertion thereby expresses a relevant conative attitude and
not merely a belief and that this is due to the meanings of the moral predicates
in a broad sense of “meaning.” This idea is compatible with the guiding idea that
the semantic values of moral predicates are moral properties and that moral
judgments express moral beliefs. After all, there is the familiar example of pejo-
rative terms,5 terms such as “Pommy,” “Rosbif,” and “Inselaffe” that are used in
Australia, France, and Germany, respectively, to refer to the English. Pejoratives
ascribe properties. On a plausible view, to be a Pommy is to be English. But
in Gottlob Frege’s terms, pejorative terms are also “colored.”6 To call Brenda
a “pom” is not merely to describe her as English; it is also to express a kind of
disdain or perhaps even contempt toward her for being English. On this view,
then, pejoratives have a “hybrid” meaning. They take ordinary properties as
their semantic values but are also used to express attitudes such as contempt,
presumably in virtue of conventions governing their use. Given that the lan-
guage includes pejoratives, it seemed to me, there should be no difficulty devel-
oping a theory according to which moral predicates similarly take properties as
their semantic values but are also used to express conative attitudes in virtue of
conventions governing their use. Could there be a problem with such a view?
In light of this reasoning, I took pejoratives as my model for an account of
the meaning of moral terms, and I assumed a hybrid account of the meanings
moral terms expresses a relevant conative attitude is not a view about the meanings of the
terms, because on my view the truth conditions of the sentences are not thereby affected. But
I here understand “meaning” in a more ordinary sense, according to which the meaning of a
term is determined by conventions governing its use such that competence in the use of the
term depends on an awareness of what these conventions permit.
5. I use the term “pejorative” broadly to include slurs and derogatory terms.
6. See Frege 1979, 140–141, 197–198; 1984, 161, 185, 357.
of pejoratives. But no model is perfect of course, and there are other accounts of
the meanings of pejoratives.7 I have no need to insist that the hybrid account is
correct. The point is that I am taking a hybrid account as my model. According
to the model, a pejorative is used both to ascribe a property P and to express
contempt (or some other negative attitude) toward those who are P, where con-
tempt is thereby expressed because the term is associated with a relevant assert-
ability norm.
It is true nevertheless that pejoratives do not all work in the same way and
that moral predicates work differently from some such terms. First, pejoratives
typically are associated with corresponding “neutral” or nonpejorative terms in
the way that “pom” is associated with the term “English.” The moral predicates
do not have similarly neutral counterparts (Finlay 2005, 19). The role of neutral
counterpart has to be filled by a complex predicate, such as “has the property,
whatever it is, that is ascribed by the term ‘wrong.’ ” Similarly, the term “jerk”
lacks a neutral counterpart other than “has the property, whatever it is, that
is ascribed by the term ‘jerk.’ ”8 Second, when typical pejoratives are used in
belief reports, contempt is expressed by the reporter rather than being ascribed
to the person whose belief is being reported (see Potts 2005; Schroeder 2009,
306; 2010, 205). If I report “Anna thinks Brenda is a pom,” I myself express con-
tempt for the English, and I do not seem to ascribe contempt for the English to
Anna. The moral predicates work differently. If I assert that Anna thinks capital
punishment is wrong, I do not myself express an aversion to wrong actions.
Similarly, if I assert that Anna thinks that Brenda is a jerk, I do not myself
express contempt for Brenda. I ascribe the contempt to Anna. Third, different
pejoratives are governed by somewhat different conventions. Some very strong
pejoratives may be governed by a rule that says not to use them at all unless
you have contempt. “Jerk” seems to be governed by a rule that says not to call
anyone a jerk unless you have contempt (Copp 2009, 186–187). “Wrong” may
be governed by a rule that says not to use the term in ascribing wrongness to an
action unless you disapprove of wrongful actions.9 All of this needs to be inves-
tigated systematically, but to do so is beyond what I hope to accomplish here.
Realist-expressivism is, then, the thesis that each moral predicate “M” takes
a moral property Mness as its semantic value and is also used to express a
7. According to “extended descriptions accounts,” for example, “pom” is used to ascribe both
the property of being English and some other related property, such as, perhaps, the property
of being contemptible on account of being English. See Bach 2014; Hom 2008.
8. Ryan Hay (2013) suggests that moral predicates behave in a way that is more similar to
offensive terms like “jerk” than to pejoratives.
9. Copp 2009, 185–189. Note that to report that Anna believes capital punishment is wrong
is not to ascribe wrongness to anything.
characteristic conative attitude C-M in virtue of linguistic conventions or

assertability norms governing its use. A large family of theories could qualify
as versions of realist-expressivism. Realist-expressivism as such is compatible
with a variety of positions regarding the moral properties, regarding the nature
of the conative attitudes expressed by moral assertions, and regarding the idea
of expression.
In brief, the version of realist-expressivism that I favor has the following
characteristics. It is a kind of metaethical naturalism according to which the
moral properties are “natural” properties. It holds that there are interesting,
nontrivial statements of the form “To be M is to be N,” where “M” is a moral
predicate and “N” is used to ascribe a natural property. It denies, however, that
such statements are conceptual or analytic truths (see Copp 1995a, 2007a). It
holds that typical moral assertions express commitment to a relevant moral
policy. For example, a person who asserts that torture is wrong expresses her
commitment to a policy of opposing and avoiding wrongful acts and thereby,
given the belief she also expresses, her commitment to a policy of opposing
torture (Copp 2007b, 181–185). This is due to a convention that governs the use
of “wrong.” Realist-expressivism holds that similar conventions govern the use
of other moral terms (Copp 2007b, 180–181, 185–193; 2009, 186–187). A sim-
pler version of realist-expressivism, a view I call Aversion, holds that a speaker’s
assertion that some kind of action is wrong expresses an aversion to wrongful
acts. According to Aversion, the predicate “wrong” is governed by a convention
according to which it is not to be used in assertions to the effect that something
is wrong unless the speaker has a relevant aversion.
It would be a delicate matter to determine the exact content of these conven-
tions. The key idea, however, is that the expressive role of the moral predicates
is due to the existence of usage conventions. Given the existence of such con-
ventions, it would be infelicitous to call something wrong, for example, if one
lacked the relevant conative attitude. Given the existence of such conventions,
I maintain, a person using a moral predicate in asserting that a kind of action
has a given moral property would conventionally implicate (or simplicate) that
she is in the relevant conative state.10 This proposal is in accord with Schroeder’s
(2009, 284) suggestion that “the mental state expressed by a sentence is the
mental state that it is the semantic assertability condition of the sentence that
the speaker be in.” As Schroeder suggests, “For a sentence, S, to express a men-
tal state, M, S must be semantically associated with a rule of the form: Assert
10. In Kent Bach’s (1999) view, to implicate something is to mean to convey it. Since I do not
think that a person making a moral assertion typically means to convey that she has a specific
conative attitude, I introduced the idea of simplicature, where, roughly, a simplicature would
qualify as an implicature if it were intended (Copp 2009, 182–185).
S only if you are in M!” (see also Schroeder 2008). On my proposal, the moral
predicates are associated with assertability rules of this kind.
T H E P R OMI S E O F R E ALI ST- E X PR E SSI VIS M
Let me turn now to what I take to be the advantages of realist-expressivism.

Here, among other things, I will answer the objection that realist-expressivism
cannot account for the element of endorsement in moral thought. First, I will
contend that realist-expressivism can explain what is correct in Gibbard’s basic
idea in a way that is compatible with moral realism. Second, I will contend that
it can explain the intuitions that lead people to accept moral judgment internal-
ism (or judgment motivational internalism) in a way that is compatible with
externalism. And finally, I will contend that it can defuse the familiar “open
question” challenge to moral naturalism.
Consider again the example of Anna. Anna says that capital punishment is
morally wrong, yet she is indifferent to whether or not criminals are executed.
She has no negative attitude toward capital punishment, nor has she any policy
that would lead her to oppose it. According to realist-expressivism, something
is indeed amiss. Her claim is infelicitous. It violates a usage convention. In a
similar way, given the usage convention governing the French pronoun “tu,” it
would be infelicitous to use “tu” in addressing someone you are not on familiar
terms with. In some contexts, a person who did this would be viewed as insin-
cere and perhaps manipulative. Similarly, if Anna says that capital punishment
is wrong but has no negative attitude toward it, she speaks infelicitously. She
might be viewed as insincere. We might be puzzled about her state of mind.
As yet this says nothing about moral thought as distinguished from moral
discourse. Gibbard was concerned with explaining the aspect of endorsement
that he viewed as “the special element” in both “normative thought and lan-
guage” (1990, 33, my emphasis). Realist-expressivism speaks primarily to the
expressive element in moral discourse. It implies that moral discourse involves a
kind of endorsement, but it does not say anything directly about moral thought.
According to realist-expressivism, a person’s having a moral belief consists in
her having the familiar cognitive attitude to a moral proposition, and a person
can have this very attitude to a moral proposition whether or not she has a suit-
able conative attitude. Nevertheless, attitudes of approval and disapproval are
typically involved in moral thought, and realist-expressivism can help explain
in what way they are involved.11
11. Bach impressed on me the importance of explaining this.

On my view, even if it is a nontrivial truth that a moral property Mness is

identical to a natural property Nness, there is a difference between believing
that something is N and believing it is M. There is similarly a difference between
believing there is water in Donner Lake and believing there is H2O in the lake.
Assuming a “fine-grained” account of the individuation of propositions, we can
explain this on the basis that the beliefs in question take different propositions
as their objects (see King 2007). It is arguable as well that there are differences
in the concepts involved. It is arguable, for example, that normative concepts
differ significantly from nonnormative concepts (Copp forthcoming). Nothing
here is surprising. Realist-expressivism can say more than this, however.
On the hybrid account of the meaning of pejoratives that I have been assum-
ing, it is infelicitous to assert that Brenda is a “pom” unless one has a relevant
attitude toward the English. Arguably it would be infelicitous in a similar way
to think that Brenda is a “pom”—to think of Brenda as a “pom,” to have the
episodic or occurrent thought that she is a “pom”—unless one has a relevant
attitude. This would be to have a thought it would be infelicitous to express
in the straightforward way by uttering “Brenda is a pom.” That is, assuming
that the meaning of “pom” is such that a speaker who calls someone a “pom”
(s)implicates that she has a relevant attitude, there would be something infelici-
tous in having the occurrent thought that Brenda is a “pom” if one lacked the
relevant attitude. This seems to be a difference between thinking that Brenda is
“English” and thinking she is a “pom.” (Similarly, given the convention govern-
ing the French “tu,” it would be infelicitous of someone not on familiar terms
with you to form the thought, say, “Tu es belle.”)12
Let me stipulate that to “accept” a claim is to be in the psychological states
that one would express in asserting that claim. On this stipulation, the differ-
ence in question between thinking that Brenda is “English” and thinking she is a
“pom” corresponds to the difference between accepting that she is English and
accepting that she is a pom. It is due to the difference between the assertability
norms governing the use of expressive terms such as “pom” and the assertabil-
ity norms governing the use of neutral terms.
According to my version of realist-expressivism, moral terms are similarly
governed by assertability norms, so realist-expressivism implies that we should
find a similar difference between thinking something is wrong and any purely
descriptive thought. On my version of realist-expressivism, it would be infelici-
tous for a person to think that torture is wrong—to think of torture as wrong,
to have the episodic occurrent thought that torture is wrong—if she lacked a
policy of avoiding wrongdoing. In this case she would have a thought it would
be infelicitous to express in the straightforward way by asserting “Torture is
12. I thank Adam Sennet. See Copp and Sennet forthcoming.

wrong.” So if there are assertability norms governing moral uses of “wrong”

such that a speaker who calls something wrong (s)implicates that she has a
policy of avoiding wrongdoing, then there would be an infelicity in having the
episodic occurrent thought that, say, torture is wrong if one lacked a policy of
avoiding wrongdoing.
Even if it would be infelicitous to think that torture is wrong if one lacked
the relevant attitude, it does not follow that a person who lacked the relevant
attitude could not think this. Nor does it follow that she would be irrational to
have this thought. Similarly, nothing I have said implies that a person could
not think or would be irrational to think that someone is a pom if she lacked
the relevant attitude toward the English. Perhaps “pom” is the term Anna is
most familiar with for the English, and perhaps she simply finds herself with
the thought that Brenda is a pom even though she lacks the relevant attitude.
It hardly seems that this thought is irrational in any way. Nor similarly does it
seem that it would be irrational to think that torture is wrong if one lacked the
relevant attitude.
Rather, what this line of reasoning about the relation between moral discourse
and moral thought suggests to me is that the thoughts of articulate beings are
governed indirectly by norms that affect the meaning of the sentences and terms
they would use to express these thoughts.13 In this sense the thoughts of articu-
late beings can have a linguistic aspect. A person who has an occurrent thought
might entertain a sentence she could use to express it. This idea seems to be
supported by the fact that when we vocalize a thought, we normally are able
straightaway to assert it without needing to search for the words. It also seems
to be supported by the fact, familiar to people who are bilingual, that there is a
difference between, say, “thinking in French” and “thinking in English.” If our
thoughts can have a linguistic aspect in this way, then a person with an occurrent
moral thought might entertain a sentence she could use to express this thought.
Given this, realist-expressivism can help explain the difference between
thinking something is wrong and any purely descriptive thought. The differ-
ence corresponds to the difference between accepting that something is wrong
(in my stipulated sense of “acceptance”) and accepting any purely descriptive
proposition. Suppose it is a nontrivial truth that a moral property Mness is
identical to a natural property Nness. Given my stipulated sense of “accept,”
there is a conative element that is involved in accepting that something is M
that presumably is not involved in accepting that something is N. By taking this
into account, realist-expressivism can help explain the element of endorsement
in moral thought.
13. See note 4. Michael Ridge pointed out that some conversational norms, such as those
governing the pace of speech, do not plausibly govern thought.
The second potential advantage of realist-expressivism is that it can explain

the intuitions that lead people to accept moral judgment internalism, but it
can do so in a way that is compatible with externalism. I accept a form of
moral judgment externalism according to which it is possible for a person to
have a moral belief without being appropriately motivated. I agree that it is
typical for a person with a moral belief to be motivated appropriately, and
I have just suggested that there is an infelicity in having an occurrent episodic
moral thought without having a relevant conative attitude. But I think it is
possible for a person with a moral belief to lack the relevant conative attitude
and for moral motivation to be entirely lacking. As has often been pointed
out, depression can sap a person’s moral motivation without eliminating her
moral beliefs. Realist-expressivism can explain this, and it can also explain
the intuitive attractiveness of moral judgment internalism. For according to
realist-expressivism, anyone who expresses a moral belief in a standard way
thereby also expresses a corresponding attitude of some relevant kind, such
as approval or disapproval. Hence a conative element is involved in accepting
that something is wrong. According to Aversion, if Anna accepts that torture
is wrong, she has an aversion to wrongful acts. Moreover, as I have just argued,
it would be infelicitous of her to think that torture is wrong but to lack an
aversion to wrongful acts. It does not follow, however, that it is not possible for
her to have the belief without having the motivation. The claims I have been
making are only about what states of mind we express in making moral asser-
tions, about what is involved in accepting a moral claim in my stipulated sense,
and about the felicity of thoughts in an extended, somewhat technical sense.
Realist-expressivism is compatible with the falsity of moral judgment internal-
ism even though it explains why moral judgment internalism might appear to
be true. I see this as an attractive result.
The third advantage of realist-expressivism is that it is potentially able
to defuse the familiar “open question” challenge to moral naturalism. The
challenge is to explain why it seems that, no matter what nonmoral and
nonnormative naturalistic description of a thing we are given, it is always
an open question whether that thing is good or whether it has any other
moral or normative characteristic. Something always seems to be left out.
Realist-expressivism can say that what is left out is the “element of endorse-
ment.” Any normative or moral characterization of the thing would express
some conative attitude toward the thing, such as approval, and no such atti-
tude is expressed by any purely naturalistic description. So any purely natu-
ralistic description will leave out the endorsement (or dis-endorsement) that
is expressed in a normative characterization.
Suppose it is nontrivially true that a moral property Mness is identical to a
natural property Nness. On this view, it is not possible for something to be N
without being M. It may nevertheless seem to be an open question whether

something that is N must also be M. For first, on my view, in the interesting
cases it is not a conceptual truth that to be N is to be M. Second, on my view
to call something M is to express a related conative attitude. There is a conative
element that is involved in accepting that something is M, and in the interest-
ing cases this conative element is not involved in accepting that something is
N. For both of these reasons, a person who knows that something is N might
coherently fail to accept that the thing is M. The absence of the conative element
in accepting that a thing is N, and the lack of a conceptual link, can leave an
open question whether to accept that the thing is M. Moreover, on my account
it would be infelicitous to have the occurrent thought that the thing is M if one
lacked the relevant conative attitude.
I recognize, however, that this explanation might not seem satisfying.
Suppose we decide that the property of being a jerk is the property of being
JKY although we do not take this to be a conceptual truth. We think there is a
conceptual gap between the claim that someone is JKY and the claim that he is
a jerk, and there is also an expressive element in calling someone a jerk that is
missing when we say he is JKY. Yet in this case I doubt that we would have the
nagging sense that something is left out, that it could not be that being JKY is
all there is to being a jerk. In the moral case, however, many philosophers have
this nagging sense of impossibility (see Enoch 2011, 100, 108). To fully explain
this, I think we need to explore the peculiarities of the normative concepts,
which I have attempted to do in another place (Copp forthcoming). So there is
more to say, but let me now turn to Schroeder’s objections regarding the logic
of moral arguments.
S C H R O E D E R ’S AR G U M E N T AN D THE
INF E R E N C E -L I C E NSI N G PR O PE R TY
Schroeder’s arguments do not fully engage with my position. He views hybrid

theories as offering “certain subtle but potentially attractive improvements over
ordinary cognitivist realism in the way that [they handle] both Open Question
phenomena and in making good on the claim that judgment motivational inter-
nalism is both necessary and a conceptual truth” (2009, 263–264). Notice, how-
ever, that the truth of judgment motivational internalism (or moral judgment
internalism) is no part of my view. Moreover, my idea is that moral realism can
gain some of the advantages of expressivism by going hybrid. Schroeder (2010,
191–192) seems mainly to be asking whether an irrealist form of expressiv-
ism can gain some of the advantages of cognitivism by going hybrid. Hence
realist-expressivism is perhaps not squarely in Schroeder’s crosshairs. Despite
this, his argument does challenge my position. For he contends that the “poten-
tially attractive advantages” of hybrid theories “are contingent on a certain
hypothesis about how the expressive meaning of moral words interacts with
intentional attitude verbs” (2009, 263–264). This is the so-called Big Hypothesis,
according to which, for example, if realist-expressivism is true, a person believes
that torture is wrong only if she both accepts the relevant proposition and has
the relevant conative attitude. Schroeder contends that the Big Hypothesis is
false, and I agree. But if so, and if Schroeder’s argument is solid, hybrid theories
face a serious problem. Their apparent advantages rest on a false hypothesis.
The lynchpin of the argument is Schroeder’s thesis that valid arguments
have a property he calls the inference-licensing property. To ensure that valid
arguments with moral content have this property, Schroeder contends, hybrid
theories must have certain features. Theories with these features, however, “are
thereby debarred from having a significant advantage over ordinary cognitivist
views.” He adds, “Hybrid theorists can’t simply have it both ways, getting the
advantages of cognitivism in explaining logic and inference and the advantages
of expressivism in explaining moral motivation [and in explaining open ques-
tion phenomena]” (2009, 259). What, then, is the inference-licensing property?
Valid arguments have three characteristic properties, says Schroeder. Most
important for our purposes, accepting the premises commits one “in some
sense” to accepting the conclusion (2009, 265). This is the inference-licensing
property. Call the thesis that valid arguments have this property the inference-
licensing thesis. Schroeder says the “whole interest” of valid arguments “is that
we can create rational pressure on people to either accept the conclusion of the
argument or else give up on one of its premises (2009, 266).” For example, we
are interested in whether there is a valid argument for the existence of God,
because such an argument would place us under rational pressure to accept
God’s existence unless we reject one of the premises.
The idea is not that a person would necessarily be irrational in some way if
she accepted the premises of a valid argument without accepting the conclu-
sion. As Schroeder says, “She may simply not have put the premises together, or
. . . she may . . . have strong evidence against the conclusion and be as yet unsure
how to proceed” (2009, 266). But “something is going wrong if someone accepts
the premises and has considered the argument but simply declines to accept
its conclusion—even when explicitly confronted with the argument from her
existing beliefs and even in the absence of any countervailing evidence.” Let me
say, for short, that something has gone wrong with a person’s thinking, other
things being equal, if she accepts the premises of a valid argument yet does not
accept the conclusion.
Schroeder formulates the inference-licensing thesis in terms of “acceptance”
rather than belief. The thesis, he says, is that “valid arguments [are such] that
accepting their premises commits someone, in some sense, to accepting their

conclusion” (2009, 265). And he stipulates, as I have done, that one accepts a
claim just in case one is in each state that would be expressed (in the relevant
sense) by a sincere assertion of that claim. He adds that “denying a sentence is
accepting its negation” (2009, 266).
Hybrid theories hold that both a cognitive and a conative state of mind are
expressed by the sincere assertion of a moral claim. Accordingly, hybrid theo-
ries are committed to the view that accepting a moral claim involves both believ-
ing a certain proposition and being in the relevant conative state. For example,
on a hybrid theory a person who accepts “Cursing is wrong” both believes the
proposition, belief in which would be expressed by her sincere assertion of the
sentence, and has the conative attitude, whatever it is, that would thereby be
expressed. The trouble is that on most hybrid views it would be possible to have
the belief that would be expressed by the sincere assertion of a moral claim but
to lack the relevant conative attitude and therefore not to accept the claim.14 But
if so, then, to ensure that valid moral arguments have the inference-licensing
property, such a theory would have to show that something will have gone
wrong, other things being equal, if someone accepts the premises of a valid
argument with a moral conclusion but not the conclusion. In what follows, let
us use Aversion as my example.
The following is a valid argument: (1) It is wrong to eat bacon. (2) If it is
wrong to eat bacon, then it would be wrong to eat this freshly cooked crispy
bacon. Hence it would be wrong to eat this bacon. On Schroeder’s stipulation,
assuming Aversion is true, one does not accept the conclusion unless one has
an aversion to eating bacon. To show that valid arguments have the inference-
licensing property, a defender of Aversion must claim, with respect to this
argument and with respect to every other relevantly similar valid argument,
either that accepting the premises guarantees having a relevantly directed aver-
sion or that accepting the premises guarantees that something has gone wrong,
other things being equal, if one lacks the relevantly directed aversion. Call the
first option the Guaranteed Conative State option and the second option the
Guaranteed Failure option. The trouble is that some valid arguments with a
moral conclusion seem to fit neither option.
Consider, for example, the following argument (compare Schroeder 2009,
291): (1) Everything Anna thinks is true. (2) Anna thinks eating bacon is
wrong. Hence eating bacon is wrong. It seems clear that I might accept the
premises of this argument and conclude that eating bacon is wrong without
14. Ridge pointed out to me that this would not be a problem for hybrid theories according
to which the content of the belief that would be expressed by a moral claim is a function of
the attitude that would be expressed.
having any aversion to eating bacon. My acceptance of premise (1) does not
involve my having such an aversion, nor does my acceptance of premise (2),
nor, it seems, does my acceptance of both. Hence it appears that the defender of
Aversion cannot make good on the Guaranteed Conative State option. And the
Guaranteed Failure option is also problematic. For on this option a defender
of Aversion would need to explain why something has gone wrong, other
things being equal, if a person accepts the premises but not the conclusion, not
because she does not believe it, but because she lacks the required aversion. As
Schroeder says, “What could be wrong with that?” (2009, 269)
Schroeder proposes, on the basis of a very detailed investigation, that to give
a hybrid theory “leverage” in explaining the inference-licensing property, “by
far the most promising option, and possibly the only real option, is to . . . hold
that different sentences containing ‘wrong’ all express the very same desire-like
state” (2009, 275).15 Call this the Same State Proposal. It proposes a negative
answer to the first question he poses to hybrid theories: “Do different sentences
containing the word ‘wrong’ express different desire-like states?” (2009, 262).
Schroeder notes that, on this proposal, “someone who makes a moral inference
does not come to have a new desire-like attitude but only comes to acquire a
new belief,” for the attitude involved in accepting the conclusion is already pres-
ent due to the person’s accepting the premises (2009, 275). Because of this, the
idea is, hybrid theories that take up the Same State Proposal should be able to
explain the inference-licensing property.
Unfortunately, the Same State Proposal is implausible. According to the
proposal, a hybrid theory should hold that every sentence containing “wrong”
expresses the very same desire-like state. This would mean that my assertion of
“Anna thinks eating bacon is wrong” would express the “very same desire-like
state” as would be expressed by my assertion of “Eating bacon is wrong.” A moral
nihilist’s assertion of “Nothing is wrong” would also express this desire-like
state. These are implausible implications of the proposal.
A worse problem is that hybrid theories cannot explain the inference-licensing
thesis even if they accept the Same State Proposal. For here is a valid argument
(Prior 1976, 90–93; see also Pidgen 1989, 133): Undertakers are church officers.
Therefore if church officers are wrong to be irreverent, undertakers are wrong
to be irreverent. No hybrid theory would say that accepting that undertakers
are church officers includes having whatever conative attitude is expressed by
sentences containing “wrong.” The sentence “Undertakers are church officers”
does not contain the word “wrong,” so the Same State Proposal does not imply
15. That is, as he says, hybrid theories should answer “no” to his first question, “Q1”, “Do
different sentences containing the word ‘wrong’ express different desire-like states?” (2009,
262).
that accepting the sentence guarantees having the attitude expressed by sen-
tences containing “wrong.” Furthermore, it is completely unclear why accept-
ing the premise of the argument commits one in any relevant sense to having
the conative attitude that is expressed by sentences containing “wrong.” Hence
even if a theory accepts the Same State Proposal, it cannot thereby put itself
in a position to show that valid arguments with moral conclusions have the
inference-licensing property.
For these reasons and others, I think we should abandon the Same State
Proposal. If we do this, however, then if Schroeder is correct, hybrid theo-
ries still need somehow to show that valid arguments with moral conclu-
sions have the inference-licensing property. I want to argue, however, that this
challenge rests on a mistake. For valid arguments do not in general have the
inference-licensing property that Schroeder defines. If I am correct, hybrid
theories are not faced with the difficult theoretical choices pressed on them by
Schroeder’s arguments.
P R O B L E MS W I TH THE I NFE R E N CE - LI CENS ING THES IS
Recall that Schroeder formulates the inference-licensing thesis in terms of

acceptance rather than belief. The thesis, he says, is that “valid arguments [are
such] that accepting their premises commits someone, in some sense, to accept-
ing their conclusion” (2009, 265). Once we bear firmly in mind Schroeder’s
stipulated meaning of “accept,” however, it is easy to find apparent counterex-
amples involving pejoratives.
It is important in this context that realist-expressivism takes a hybrid
account of the meaning of pejoratives as a model for its account of the meaning
of moral predicates. For if valid arguments involving pejoratives do not have
the inference-licensing property, as I will argue—assuming that they have a
hybrid meaning—then it would be question begging and ad hoc to suppose
without argument that valid arguments involving moral predicates do have the
inference-licensing property. This said, however, in what follows I do not need
to rely on a hybrid account of the meaning of pejoratives. For I can stipulate that
for some nonpejorative “neutral” predicate P there is a corresponding predicate
“P*” that has a hybrid meaning of the kind in question. So let me stipulate that
the predicate “English*” is used to predicate being English as well as to express
contempt for the English.16
Consider the following argument: Everything Anna thinks is true; hence if
Anna thinks Brenda is English*, Brenda is English*. This argument is valid.
16. I am grateful to Ridge and Bach for helpful discussion (see Bach 2014).
Any possible world in which the premise is true is a world in which the conclu-
sion is true. Moreover, on my stipulation an assertion of the sentence “If Anna
thinks Brenda is English*, she is English*” would express contempt for the
English. Hence accepting the conclusion of the argument would involve having
contempt for the English. It nevertheless appears clearly possible to accept the
premise of the argument but not to accept the conclusion—because one lacks
contempt for the English—without one’s thinking being faulty in any relevant
way, not even if all else is equal. Here is another example: Brenda is English;
therefore if the English are English*, Brenda is English*. On my stipulation an
assertion of the conclusion would express contempt for the English. Yet it is cer-
tainly possible to accept the premise that Brenda is English but not to accept the
conclusion—because one lacks contempt for the English—without one’s think-
ing being faulty in any relevant way, not even if all else is equal.
The following argument is more controversial: Brenda is English; therefore
Brenda is English*. The argument is valid on my stipulation.17 Yet accepting
that Brenda is English obviously does not commit one to having contempt for
the English. Hence accepting the premise does not commit one to accepting
the conclusion. Anna might know that Brenda is English, know the meaning
of “English*”, and yet decline to “accept” that Brenda is English* “even when
explicitly confronted with the argument from her existing beliefs and even
in the absence of any countervailing evidence” (Schroeder 2009, 266). Other
things are equal, yet nothing need have gone wrong in any relevant way in
Anna’s thinking provided that she believes the conclusion. But she does believe
this, for the proposition expressed by the conclusion is the proposition that
Brenda is English, which Anna believes, since she accepts the premise of the
argument. This therefore also appears to be a valid argument that lacks the
inference-licensing property.
To better understand the issues here, we need to distinguish between two
“inference-licensing properties” and two “inference-licensing theses.”18 The
acceptance-inference-licensing thesis is the thesis Schroeder formulates, according
to which, other things being equal, something has gone wrong in one’s thinking
if one accepts the premises of a valid argument but fails to accept the conclu-
sion. According to this thesis, valid arguments have the acceptance-inference-
licensing property. There is also the belief-inference-licensing thesis, according to
which, other things being equal, something has gone wrong in one’s thinking if
one believes the premises of a valid argument but fails to believe the conclusion.
17. As Ridge pointed out, it is not formally or logically valid even though the truth of the
premise guarantees the truth of the conclusion on my stipulation.
18. Schroeder (2010, 197) notices this distinction.
According to this thesis, valid arguments have the belief-inference-licensing

property. I claim that the belief-inference-licensing thesis is true, at least when
restricted to arguments the validity of which is knowable a priori, but the
acceptance-inference-licensing thesis is false.19
The rationale Schroeder gives for “the” inference-licensing thesis is actu-
ally a rationale for the belief-inference-licensing thesis rather than the
acceptance-inference-licensing thesis. Valid arguments are important, he says,
because “we can create rational pressure on people to either accept the con-
clusion of the argument or else give up on one of its premises” (2009, 266).
But the point is that we invoke the validity of an argument in order to cre-
ate rational pressure on people to either believe the conclusion of the argu-
ment or else deny one of its premises. We do not think that the validity of the
argument from Brenda is English to the conclusion that Brenda is English*
would create rational pressure on anyone to have contempt for the English.
So the acceptance-inference-licensing thesis is not supported by Schroeder’s
rationale. Only the belief-inference-licensing thesis is so supported. The
belief-inference-licensing thesis is the one that is intuitively plausible.
Moreover, intuitively the reason a person who believes the premises of a valid
argument is committed to believing the conclusion is that, given her beliefs, the
conclusion must be true. The conclusion is true in every world that is possible
given what she already believes. It is true in every epistemically possible world.
The same obviously cannot be said for conative states that might be expressed
by asserting the premises or conclusion of a valid argument in a case in which
these claims have expressive content, for conative states are not apt to have
truth values.
The important point is that, on any plausible hybrid theory, valid moral argu-
ments do have the belief-inference-licensing property given the restriction to
arguments the validity of which is knowable a priori. Realist-expressivism has
no difficulty explaining how it is that something has gone wrong in a person’s
thinking, other things being equal, if she believes the premises of a valid argu-
ment with a moral conclusion yet fails to believe the conclusion. The explana-
tion is the same as would be given in the case of valid nonmoral arguments
and the same as would be given in the case of valid moral arguments by any
standard cognitivist theory. Given that the validity of an argument is know-
able a priori, something has gone wrong in a person’s thinking if she believes
the premises of the argument and if she has considered the argument but
19. The argument from Donner Lake contains water to Donner Lake contains H2O is valid,
since it is not metaphysically possible that the premise be true but the conclusion not be true,
but the argument’s validity is not knowable a priori. Nothing need have gone wrong in a per-
son’s thinking if she believes the premise but not the conclusion, not even if all else is equal.
simply does not believe its conclusion “even when explicitly confronted with
the argument from her existing beliefs and even in the absence of any counter-
vailing evidence” (Schroeder 2009, 266). On any plausible theory, valid moral
arguments have the belief-inference-licensing property even if they lack the
acceptance-inference-licensing property.
One might object that, in offering an argument for a moral conclusion, we
often want to achieve more than simply to “create rational pressure” on people
to believe the conclusion (Schroeder 2009, 266). In cases where the conclusion
of the argument is that the people we are addressing ought to do something,
we normally also want to motivate these people to act accordingly. And we
may want to motivate them by creating rational pressure on them. The “practi-
cality requirement” can explain how we can do this, for it says that necessarily
anyone who believes she ought morally to do something is either motivated
to act accordingly or is not fully rational (Smith 1994, 61). I believe that the
practicality requirement is false (Copp 1995b; 2007a, chap. 8). But even if we
assume it is true, it does not follow that all valid arguments with moral con-
clusions have the acceptance-inference-licensing property. For first, there is a
gap between showing that a person would be irrational to lack motivation to
do something and showing that she would be irrational to lack an aversion to
wrongdoing or to lack a policy of avoiding wrongdoing. Second, the practicality
requirement is insufficiently general. It implies only that if a person accepts
the premises of a valid argument the conclusion of which concerns how she
ought to act, then she would be irrational not to be motivated accordingly.
The acceptance-inference-licensing thesis concerns all valid arguments with
moral conclusions.
One might think that my views about moral thought support the thesis that
arguments with moral conclusions have the acceptance-inference-licencing
property. On my view there is an infelicity in a person’s thinking if she has the
occurrent thought that something is wrong but lacks the relevant conative atti-
tude. Suppose Brenda believes that everything Anna thinks is true and believes
as well that Anna thinks eating bacon is wrong. Suppose Brenda accordingly
comes to have the occurrent thought that eating bacon is wrong. On my
account, if she lacks the relevant conative attitude, her thinking is infelicitous.
But even if this is correct, it does not follow that arguments with moral conclu-
sions have the acceptance-inference-licencing property. For, first, on my view
the infelicity in Brenda’s thinking has nothing to do with the fact that she came
to have the problematic thought in virtue of valid reasoning from premises she
accepts. Second, there would not be a similar infelicity in Brenda believing that
eating bacon is wrong without having the relevant conative attitude. My claim
concerns an infelicity in having certain occurrent or episodic thoughts without
having relevant conative attitudes. Third, infelicity in thought is not a failure of
the relevant kind. It is not a failure of the kind that would be exhibited, other
things being equal, by a person who did not believe the conclusion of a valid
argument the premises of which she accepts. The infelicity in Brenda’s think-
ing in the example simply amounts to her having a thought that she could not
felicitously express in the standard way, by asserting “Eating bacon is wrong.”
As I argued in the second section, it does not follow that she is irrational to have
this thought, nor does it follow that she failing to be in a psychological state she
is committed to being in.
As we have seen, there seem to be valid arguments that do not have the
acceptance-inference-licensing property. Moreover, the reason for thinking that
valid arguments have the belief-inference-licensing property does not support
the idea that they have the acceptance-inference-licensing property. It is there-
fore problematic to insist that a plausible hybrid theory must show that valid
arguments with moral conclusions have the acceptance-inference-licensing
property. The belief-inference-licensing property is not similarly problematic,
however, and this thesis raises no problems for plausible hybrid theories.
WH E R E A R E W E ?
As we saw in the third section, Schroeder contends that any valid inference
has the acceptance-inference-licensing property. He then points out that hybrid
theories have trouble explaining this, and he proposes that to solve this prob-
lem, a hybrid theory should adopt the Same State Proposal. On this proposal
a defender of Aversion would say that every sentence containing “wrong”
expresses aversion to wrongful acts. We then saw that the Same State Proposal
is implausible and worse, even if it were true, a hybrid theory could not use it
to show that valid arguments have the acceptance-inference-licensing property.
As I argued in the fourth section, however, some valid arguments do not have
the acceptance-inference-licensing property. The property valid arguments
must have is the belief-inference-licensing property, and plausible hybrid theo-
ries have no difficulty explaining this.
The upshot is to overturn much of Schroeder’s argument. In summarizing his
argument, Schroeder says that he has given reasons why hybrid theorists should
adopt the Same State Proposal (2009, 262). And Schroeder says, if a hybrid
theory rejects the Same State Proposal, then “the inference-licensing property
is no more explicable than it would be on a traditional expressivist account. So
even though such views are possible, they don’t offer any clear progress (2009,
297).” I have argued to the contrary. Realist-expressivism can explain the belief-
inference-licensing property in the same way that it would be explained on any
standard cognitivist account. And it is not the case that all valid inferences have
the acceptance-inference-licensing property.
This is not the end of the matter, however. For, Schroeder contends, hybrid
theories still need to explain how it is that “new moral conclusions [have] the
ability to independently motivate” (2009, 297). He adds, “It is precisely this
idea”—that new moral conclusions have the ability to motivate—“that is at
the heart of the theoretical grounds for the sort of internalism that motivates
hybrid theories in the first place” (2009, 297). As I have explained, however, my
motivation for proposing realist-expressivism was not to secure the truth of
moral judgment internalism. It was, inter alia, to secure an explanation for the
intuition that moral judgment internalism is true without committing myself to
its actually being true.
This brings us to Schroeder’s Big Hypothesis. For as it turns out, if we assume
a suitable hybrid theory, moral judgment internalism follows from the Big
Hypothesis. Indeed Schroeder introduces the Big Hypothesis because he thinks
hybrid theories need the Big Hypothesis in order to make good on their claim
that moral judgment internalism is true (2009, 263–264). Since I reject moral
judgment internalism, I need to reject the Big Hypothesis.
T H E B I G H Y P O TH E SI S
The Big Hypothesis is an hypothesis about the meaning of “believes” (and per-
haps other propositional attitude verbs). Schroeder formulates it as follows, fol-
lowing Daniel Boisvert (2008):
Big Hypothesis: If ‘P’ is a sentence expressing mental states M1 . . . Mn,

then the descriptive content of ‘S believes that P’ is that S is in each of men-
tal states M1 . . . Mn (2009, 301).
According to the Big Hypothesis, if the meaning of “wrong” is such that a

speaker expresses aversion to wrongful actions in saying X “is wrong,” then
sentences to the effect that someone believes X to be “wrong” are true only if the
believer has an aversion to wrongful actions. As Schroeder puts it, “What the
Big Hypothesis says is that attitude ascriptions (at least, ‘believes that’ ascrip-
tions) attribute not only belief in the descriptive content of the complement
clause but the desire-like attitudes that it expresses as well” (2009, 301).
Given a suitable hybrid theory, the Big Hypothesis entails moral judgment
internalism. According to Aversion, for example, if the Big Hypothesis is true,
Anna does not believe that torture is wrong unless she has an aversion to
wrongful actions. Such an aversion obviously would motivate someone not to
engage in torture if she believed that torture is wrong. According to Aversion,

then, if the Big Hypothesis is true, Anna does not believe that torture is wrong
unless she is motivated to some degree not to torture. Similarly, a defender of
Aversion might hold that an assertion of “I ought to A” expresses an aversion
to failing to A. On such an account, again, if we assume the Big Hypothesis, a
person cannot believe she ought to act in a certain way unless she is motivated
appropriately, at least to some degree. The Big Hypothesis therefore seems to
entail moral judgment internalism if we assume a suitable hybrid theory.
In any event, Schroeder argues, and I agree with him, that the Big Hypothesis
is implausible. Schroeder offers two arguments to show that pejoratives do not
obey the hypothesis. Rather than explore and evaluate his arguments, I will
present three counterexamples to the hypothesis (2009, 304–306).
First, suppose Bigot says of Nice Guy, “He believes that Brenda is a pom.”
Considering a similar example, Schroeder suggests in effect that we would take
Bigot’s claim to express his own attitude to the English, not Nice Guy’s, and we
would think that Bigot’s claim might well be true even if Nice Guy lacks any
relevant negative attitude (2009, 304–305). Bigot would not contradict himself
if he added that Nice Guy, “fool that he is, has only positive attitudes toward
the English.” Second, suppose Carol says “My arm hurts.” Her assertion would
express both her pain and her thought that her arm hurts. According to the Big
Hypothesis, then, if we say “Carol believes her arm hurts,” what we say is true
only if Carol is in pain. But perhaps Carol has been having a bad dream and
wakes up thinking her arm hurts although she is not actually in pain. In such a
situation it would still be true that she believes her arm hurts although the Big
Hypothesis implies otherwise. Third, an assertion of “It is sad that the boy died”
would express sadness as well as the speaker’s thought that the boy’s dying “was
sad.” (Perhaps this could be understood as the thought that the boy’s death was
a sufficient reason for feeling sad.) According to the Big Hypothesis, it would
follow that if we say “Donna believes it was sad that the boy died,” what we say
is true only if Donna is actually sad. But we can imagine cases in which Donna
believes that the boy’s dying was sad but does not feel sad. Perhaps Donna is
completely immune to feelings of sadness although she understands that the
boy’s death was a sufficient reason for feeling sad.
These examples are of different kinds. One involves a pejorative, one involves
a sentence used to express pain, and one involves a sentence used to express sad-
ness. Partly because the examples are of such different kinds, it seems reason-
able to take them to provide grounds for skepticism about the Big Hypothesis.
For these reasons, I do not find the Big Hypothesis to be at all plausible.
Would a version restricted to moral belief be more plausible? The restricted
hypothesis would say that, if some version of realist-expressivism is true,
then the descriptive content of “S believes that P,” for a sentence “P” that
contains moral terms, would be that S has both the relevant propositional
attitude and the corresponding conative attitude. Restricted in this way, the
Big Hypothesis would be a hypothesis about the meaning of “believes” when
it takes sentences that contain moral terms as its complement. As such, it
would be ad hoc. Moreover, it would run afoul of certain examples. Brenda
believes it would be wrong to eat bacon, but it seems possible that she lacks
any relevant conative state.
Schroeder argues that hybrid theories need to defend the Big Hypothesis
in order to have the advantages that made them attractive in the first place.
But I will now argue that the falsity of the Big Hypothesis is not a problem for
realist-expressivism.
T H E P R OMI SE O F R E ALI ST- E X PR E S S IVIS M REVIS ITED
What was the promise of realist-expressivism? First, the theory accommodates

the intuitive idea that a person making a moral assertion expresses not only
a belief but also a relevant conative attitude. This is Gibbard’s basic idea, and
it is brought out by the example of Anna. The theory also accommodates the
idea that moral thought involves an element of endorsement. It can say that
Anna’s thought that capital punishment is wrong is infelicitous if Anna lacks
the relevant conative attitude. Second, realist-expressivism helps explain why
it may seem that moral belief guarantees moral motivation, as is implied by
moral judgment internalism. But the explanation it provides is compatible with
the view that moral belief does not in fact guarantee moral motivation. Finally,
realist-expressivism helps explain the “open” feel of inferences from nonmoral
naturalistic premises to moral conclusions, and it explains this in a way that is
compatible with the naturalist’s thesis that moral facts are natural facts.
Schroeder contends, as we saw, that hybrid theories cannot account for the
logic of moral arguments and in particular for the fact that moral arguments
have the inference-licensing property unless they accept the Same State Proposal.
But, he contends, if they accept the Same State Proposal, they cannot achieve the
advantages they promised unless they also assume the Big Hypothesis. I think
we can now see, however, without looking in detail at the complexities that
Schroeder explores, that hybrid theories do not face these difficulties.
Most important for my purposes, realist-expressivism can account for the
logic of moral arguments and for the fact that valid moral arguments have
the belief-inference-licensing property, even though it rejects the Same State
Proposal. Moreover, valid arguments need not have the acceptance-inference-
licensing property. Furthermore, the Same State Proposal is highly implausi-
ble. But if realist-expressivism can account for the belief-inference-licensing
property of valid moral arguments without accepting the Same State Proposal
and if valid arguments need not have the acceptance-inference-licensing prop-
erty, then there is no need for the Big Hypothesis. It appears, then, that at least
some hybrid theories have no need for either the Same State Proposal or the
Big Hypothesis.
Hybrid theories can have it “both ways.” That is, at least realist-expressivism
can rely on its cognitivist aspect to explain the logic of moral arguments. It can
rely on its expressivist aspect to help explain the open feel of certain arguments
and to explain the intuition we might have that moral motivation is guaranteed
by moral belief. Yet realist-expressivism is compatible with the kind of external-
ist moral naturalism that I favor.
CO N C L U S I ON
In order for our actions to be grounded in a moral perspective, moral belief

needs to be accompanied by motivating states of mind, such as, I suggest, com-
mitment to suitable moral policies. Such commitment is partly constituted by
relevant intentions, and because of this, it brings moral policies to bear on our
decisions. As I use the term, a moral conviction is a moral belief that is accom-
panied by the relevant moral policy. I would say, then, that the point of moral
discourse is in part to guide people to form the appropriate convictions and not
mere moral beliefs unaccompanied by commitment to corresponding policies.
The goal of guiding people to have appropriate moral convictions is served by
the existence of conventions governing the use of our moral terms such that
when we use moral terms in expressing our basic moral beliefs we also thereby
express commitment to relevant policies. If this is correct, moral discourse does
not merely express our beliefs. It also expresses our commitment to moral poli-
cies that figure in our decisions about how to act.
This, then, is the basic thesis underlying realist-expressivism in the version
that I favor. The view is intended primarily of course to account for the expres-
sive aspect of moral discourse. As I explained, however, the theory can also
account indirectly for the element of endorsement in moral thought. According
to realist-expressivism, a person’s having a moral belief consists in her taking
the familiar propositional attitude to a moral proposition, and this does not
require that she have any relevant conative attitude. Nevertheless, we can say a
person does not accept a moral proposition and does not have a moral convic-
tion unless she has the relevant conative attitude, such as a commitment to a
moral policy. And as I have explained, it would be infelicitous for a person with
an occurrent basic moral thought to lack the relevant conative attitude, since in
this case she would have an occurrent thought that it would be infelicitous for
her to express. In my view, for example, if a person has the occurrent thought
that torture is wrong but lacks a policy of avoiding wrongdoing, she has a
thought that it would be infelicitous for her to express by a straightforward
assertion of the sentence “Torture is wrong.” Our thoughts are in this way gov-
erned indirectly by the assertability norms that affect the meaning of the sen-
tences we would use to express these thoughts.
Given all of this, realist-expressivism can help explain the difference between
thinking something is wrong and any purely descriptive thought. It can say
that moral discourse and thought are governed (whether directly or indi-
rectly) by assertability norms that do not govern purely descriptive thought
and discourse. Moreover, taking this into account, realist-expressivism can help
explain the element of endorsement in moral thought and discourse. It can do
this in a way that is compatible with externalist and naturalistic moral realism.
References
Bach, Kent. 1999. “The Myth of Conventional Implicature.” Linguistics and Philosophy
22.4: 327–366.
———. 2014. “Loaded Words: On the Semantics and Pragmatics of Slurs.” Presented to
the 2014 meetings of the Pacific Division of the American Philosophical Association.
http://online.sfsu.edu/kbach/Bach.LoadedWords.APA.pdf
Boisvert, Daniel. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly
89.2: 169–203.
Bratman, Michael. 1989. “Intention and Personal Policies.” Philosophical Perspectives
3:443–469.
Copp, David. 1995a. Morality, Normativity, and Society. New York: Oxford University
Press.
———. 1995b. “Moral Obligation and Moral Motivation.” In On the Relevance of
Metaethics: New Essays on Metaethics, ed. Jocyelyne Couture and Kai Nielsen, 187–
219. Canadian Journal of Philosophy supp. vol. 21. Calgary: University of Calgary
Press.
———. 2007a. Morality in a Natural World. Cambridge: Cambridge University Press.
———. 2007b. “Realist-Expressivism: A Neglected Option for Moral Realism.” In
Morality in a Natural World, 153–202. Cambridge: Cambridge University Press.
Originally published in Social Philosophy and Policy 18 (2001): 1–43.
———. 2009. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies
in Metaethics, vol. 4, ed. Russ Shafer-Landau, 167–202. Oxford: Oxford University
Press.
———. Forthcoming. “Normative Naturalism and Normative Nihilism:Parfit's Dilemma
for Naturalism.” In Reading Parfit: On What Matters, ed. Simon Kirchin. London:
Routledge.
Enoch, David. 2011. Taking Morality Seriously: A Defense of Robust Realism.

Finlay, Stephen. 2005. “Value and Implicature.” Philosophers’ Imprint 5.4: 1–20. http://
quod.lib.umich.edu/p/phimp/3521354.0005.004/1/--value-and-implicature?rgn=pu
blications;view=image;q1=dlps.
Frege, Gottob. 1979. Posthumous Writings. Edited by Hans Hermes, Friedrich Kambartel,
and Friedrich Kaulbach. Chicago: University of Chicago Press.
———. 1984. Collected Papers on Mathematics, Logic, and Philosophy. Edited by Brian
McGuinness. Oxford: Basil Blackwell.
Gibbard, Allan. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University
Press.
Language.” European Journal of Philosophy 21.3: 450–474.
Hom, Christopher. 2008. “The Semantics of Racial Epithets.” Journal of Philosophy
105.8: 416–440.
King, Jeffrey C. 2007. The Nature and Structure of Content. Oxford: Oxford University
Press.
Pidgen, Charles R. 1989. “Logic and the Autonomy of Ethics.” Australasian Journal of
Philosophy 67.2: 127–151.
Potts, Christopher. 2005. The Logic of Conventional Implicatures. New York: Oxford
University Press.
Prior, A. N. 1976. “The Autonomy of Ethics.” In Papers in Logic and Ethics, 88–96. Edited
by P. T. Geach and A. J. P. Kenny. London: Duckworth.
Schroeder, Mark. 2008. “Expression for Expressivists.” Philosophy and Phenomenological
Research 76.1: 86–116.
———. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119.2: 257–309.
———. 2010. Noncognitivism in Ethics. London: Routledge.
Sennet, Adam, and David Copp. “What Kind of Mistake is it to Use a Slur?” Forthcoming
in Philosophical Studies.
Smith, Michael. 1994. The Moral Problem. Oxford: Blackwell.
4
Attitudinal Requirements for Moral

Thought and Language
Noncognitive Type-Generality
RYA N H AY n
I N T R OD U C TI O N
While taking stock of the variety of recent “hybrid” metaethical proposals,

Mark Schroeder (2009, 259) maintains, “Hybrid theorists can’t simply have it
both ways, getting the advantages of cognitivism in explaining logic and infer-
ence and the advantages of expressivism in explaining moral motivation.” In
this chapter I will proceed as a naive hybrid expressivist, keeping in mind two
very broad goals. I will initially consider a minimal hybrid expressivist posi-
tion and the theoretical benefits of making certain further commitments. The
view should be able to straightforwardly address issues about logical embed-
ding and account for the connection between moral judgment and motivation.
I will maintain that a hybrid expressivist view that makes a certain commit-
ment about the structure and content of the noncognitive attitude associated
with moral thought and language can get these advantages.
This chapter is divided into three sections, and over the course of it I will nar-
row my focus onto a particular kind of hybrid expressivist view that I think may
be best suited to get what is needed. In the first section, following Schroeder’s
work in assessing the merits of current hybrid views as well as proposals made
by Daniel Boisvert, Michael Ridge, and David Copp, I briefly review why the
hybrid expressivist may be optimistic about “having it both ways.” In the second
section I will argue that interpreting the general principles that allow hybrid
expressivists to be optimistic turns out to be problematic. The issues arise out of
certain (initially advantageous) assumptions made about the general structure of

the content of the mental state associated with moral thought and language. In
the third section I propose a refinement that may allow hybrid expressivists to
avoid these issues. The basic idea is that agents can possess generalized noncogni-
tive attitudes toward types rather than propositions involving universals. I will end
by maintaining that this position may lead to a further advantage in explaining
morality and motivation.
HY B R I D E X P R E SSI V I SM , M O TI VATI O N ,
A N D E MB E D D I NG
One way of generally characterizing “hybrid” metaethical views is to describe

what such views may hold about basic kinds of moral judgments. They may hold
that moral judgments involve a state composed of both cognitive and noncog-
nitive elements. Another way hybridity is typically characterized is as a thesis
about the meaning of moral language. I will proceed here by characterizing a
view along both of these dimensions and the options the hybrid expressivist can
take. I will call any view compatible with the following two very general claims
of the hybrid expressivist. First, the view is expressivist if it is compatible with
the claim that a particular utterance expresses a certain state of mind just in
case it is correct, as a matter of a semantic or conventional rule associated with
meaning, to assert that sentence only if you are in that state of mind.1 Second, a
hybrid expressivist view holds that just as descriptive assertions express repre-
sentational cognitive states of mind, sincere moral assertions express a complex
mental state in some way composed of both cognitive and noncognitive ele-
ments.2 I will represent the general structure of these hybrid expressivist com-
mitments as follows:
1. This minimal expression relation follows Schroeder’s (2008, 108–109) suggestion for
expressivists that he refers to as minimal assertability expressivism. I have this kind of mini-
mal expression relation in mind here for the hybrid expressivist, and I intend the relation to
be even weaker in the sense that it is neutral with respect to a semantic/pragmatic distinc-
tion. The important feature is that the expression relation tells us something about meaning
in general.
2. Here are some terminological clarifications: (1) I use “moral judgment” here to refer to the
internal states individuals are in when they think, for instance, that a particular action has a
particular moral quality. A simple judgment has a simple subject-predicate form. Making a
moral judgment need not involve a linguistic act. (2) The mental components of the hybrid
analysis (both cognitive and noncognitive) are about something, and I will generally use
“content” to refer to whatever it is that they are about. (3) Finally, I assume that hybrid
expressivists can at least decompose, for the purposes of the proposed analysis, the elements
of the complex mental state expressed.
Attitudinal Requirements for Moral Thought and Language 77
Hybrid analysis for an agent MS[ χ] ≡ (В [α] & D [β])

Expression for a speaker SU ⌊χ⌋ ⊃ MS[χ]
Let MS[ . . . ] be a mental state about moral content . . . ; let SU ⌊ . . . ⌋ be an

assertion of . . . in natural language; В is a cognitive element, D is a noncog-
nitive element.
This characterization of hybrid expressivism is meant to be very general. There
are a number of relatively recent specific hybrid-style views that take different
approaches to filling in the details.3 These views differ with respect to the general
form and content of the cognitive and noncognitive elements associated with moral
language and thought. They differ with respect to how moral language might be
said to express such elements. That being said, what we have here, though general,
is at least a starting point from which I will begin to make certain choices, keeping
in mind the benefits that arise from selecting among various theoretical options.
One of the reasons to “have it both ways” with a hybrid view is that it can
apparently address why moral judgments appear to be motivating in ways that
descriptive judgments are not. As David Hume (2010 [1740]) writes near the
outset of book 3 of his Treatise of Human Nature, “Morals excite passions, and
produce or prevent actions.” This apparent practicality requirement of moral
judgment weighs heavily on contemporary metaethical views. In The Moral
Problem (1994) Michael Smith maintains that one of the central problems in
metaethics is to reconcile the ideas that (1) there is an internal connection
between moral judgment and factual belief, (2) there is an internal connec-
tion between moral judgment and motivation, and (3) desires and means-end
beliefs are necessary components of motivation. The hope of course is that
hybrid expressivists can account for (1) by citing the cognitive element associ-
ated with moral thought. Hybrid expressivists can account for (2) and (3) by
explaining how the noncognitive and cognitive states result in the appropriate
motivating state. How that explanation proceeds depends on what the hybrid
expressivist says about the nature of the cognitive and noncognitive elements
associated with moral judgments. So while keeping that goal in mind, I will
consider the options before the hybrid expressivist with respect to analyzing the
associated cognitive and noncognitive elements.
3. Even at the outset many positions that have been discussed in the context of hybridity are
not expressivist in this sense. These include views that I understand to be fundamentally
cognitivist about the meaning of moral language, such as Stephen Finlay’s end-relational
view or Dorit Bar-On and Matthew Chrisman’s neo-expressivism. Many hybrid expressivists
differ in their further commitments about the expression relation. I will remain neutral on
this issue here.
In the case where an agent judges a particular action to be wrong and sin-
cerely makes statements to that effect, the hybrid expressivist may maintain
that the associated noncognitive state is a negative attitude like disapproval or
disgust. Not much turns here on what we call this attitude, so I will just follow
the standard convention of labeling it “desire” to distinguish it from the associ-
ated cognitive element (which I will often just generically refer to as “belief ”).
The more important issue, as we will see, pertains to determining the content of
this attitude. One intuitive option is just to hold that the desire associated with
moral judgment has as its content a relation between the agent and an action.
This option would provide us with the following general recipe, based on what
I will call a “Humean intuition”:4
For any action, x, any moral predicate, φ, and any associated action rela-
tion ψ, an agent judges that φx only if the agent desires that ψx.5 (∀x)
(MS[φx] ⊃ D [ψx])
It is the subject action that is in the content of the agent’s desire. This Humean
intuition may serve as the basis for simple subjectivist and purely noncognitivst
expressivist positions. In turn the hybrid expressivist might adopt it as well.
After all, this would allow for a straightforward answer to the question of why
moral judgments are motivating. Presumably, desiring to do something (along
with further beliefs about how to go about doing it) will motivate you to act.
And insofar as you express this desire by making statements that correspond to
your judgment, this can account for what James Dreier (1990) calls the “inter-
nalist character” of moral claims. This could explain why moral thought and
language is often seen as necessarily motivating. Perhaps this connection is too
strong, but I will return to that issue in the third section. For now, having the
explanatory resources to address the practicality requirement is a desirable fea-
ture of the view.
The problem with relying too heavily on the Humean intuition can be seen
as soon as it is considered in light of another main theoretical advantage sought
4. As Hume (2010 [1740], bk. 3, sec. 1) writes: “Take any action allowed to be vicious: Willful
murder, for instance. Examine it in all lights, and see if you can find that matter of fact, or
real existence, which you call vice. . . . The vice entirely escapes you, as long as you consider
the object. You never can find it, till you turn your reflection into your own breast, and find a
sentiment of disapprobation, which arises in you, towards this action” (my emphasis).
5. Where ψ can be an n-place relational predicate. For example: the judger does or avoids
doing x, or does x and gets others to do so, etc. Not much turns on this here; the primary
issues I will raise pertain to the contents of attitudes more generally, holding fixed the instan-
tiated ψ for any corresponding φ. I will initially treat desire as a propositional attitude.
by the hybrid expressivist. As the basis for a purely noncognitivist expressiv-

ist position, the adoption of the Humean intuition leads to the classic logical
embedding problem.6 Expressivists and their critics have realized that what-
ever proposed noncognitive attitude is necessary for simple claims involving
moral predicates, it must allow us to account for their consideration and use in
reasoning. This leads to the question: How do simple claims involving moral
predicates embed in the standard variety of first-order logical constructions?
Expressivist accounts of moral claims make further commitments about the
structure and content of attitudes to account for logic and general semantic
compositionality. The hope is that hybrid expressivists may be entitled to a sim-
pler account in part because of the cognitivist resources available to them.
Part of the hope for the hybrid expressivist certainly lies in an account of
the cognitive element. For any particular moral judgment made by an agent,
there is a cognitive state that can be thought of as a relation between the
agent and some neutral counterpart proposition. Whenever an agent judges
an act x to have a certain moral quality φ, the agent is in a state in part com-
posed of the belief that x is φ. This neutral counterpart can be considered
even if there is no readily synonymous neutral counterpart predicate in
the language that has the same extension and intension as the moral term
in question.7 Furthermore, if an agent accepts any further first-order logi-
cal constructions involving moral predicates, there are corresponding neutral
counterpart propositions that the agent likewise believes. This propositional
account involving neutral counterparts can be relied on to play just the same
role in cognitive reasoning as any beliefs with ordinary descriptive contents.
Following Ridge’s (2006) observations, inconsistency can be explained “on
the cheap” by appeal to the cognitivist components supposed necessary by
the hybrid view. Accepting both that stealing is wrong and that stealing is
not wrong is inconsistent, since holding the requisite beliefs in their neutral
counterpart propositions is inconsistent.
6. Over time this problem has been variously attributed to Gottlob Frege, P. T. Geach, and
John Searle. Geach (1965) demonstrates the issue with an argument like the following: If
tormenting the cat is wrong, getting your little brother to do it is wrong. Tormenting the cat
is wrong. So, getting your little brother to torment the cat is wrong. The second premise and
conclusion require, respectively, disapproval of cat tormenting and getting your little brother
to do so. But the Humean intuition does not provide us with a recipe for determining the
noncognitive attitude associated with the conditional premise.
7. I make a longer argument for this in Hay 2013, 454–458. Ridge (2006, 313) also notes in
his suggestion for the hybrid expressivist that “the speaker may or may not have a very clear
idea of what the relevant property is.” Similarly, Copp (2009, 188) writes, “Of course, even
though there is no familiar term that is a neutral equivalent for ‘wrong’, there are neutral
equivalent phrases, such as ‘has the property called ‘wrongness.”
While this explanation for inconsistency is straightforward, it turns out that

hybrid expressivists do not entirely avoid the embedding problem by simply
invoking the cognitivist element in their analyses. Hybrid expressivists must
not only explain why, for instance, it is inconsistent to accept the premises of
a valid argument but reject its conclusion (Ridge 2006, 313; Schroeder 2009,
265) but also how, should one consider and accept the premises, one should
accept the conclusion (Schroeder 2009, 266). This still puts pressure on the
hybrid expressivist to explain the relation between the premises and the con-
clusions of valid arguments in terms of the relations between the associated
noncognitive attitudes required for their acceptance.
However, if the content of the noncognitive attitude is structured in a certain
way, hybrid expressivists may be entitled to a plausible and relatively simple
explanation for embedding. In “Ecumenical Expressivism: Finessing Frege”
Ridge (2006, 313) suggests that the hybrid expressivist might account for the
noncognitive element as an “attitude in favor of actions in general insofar as
they have a certain property.”8 Similarly, in “Expressive-Assertivism” Boisvert
(2008, 171) maintains that the speaker’s noncognitive attitude “is not directed
(at least only—and not by way of literal meaning) toward the subject of the
sentence, but toward things of a more general kind, namely, things that have
the property picked out by the ethical predicate.” Boisvert touts this as one of
the key components of his own hybrid view. Copp (2009, 187) also considers
that idea that it may be “semantically inappropriate to use the term in making
a moral judgment unless one disapproves of actions that have the property of
being wrong.” And Schroeder (2009), in evaluating the range of current hybrid
offerings, characterizes these kinds of positions as holding something like the
desire not to do whatever has the property picked out by “wrong.” The basic
idea is that, for any subject-predicate construction “φx” involving a moral term
“φ” there is an associated noncognitive attitude about whatever actions gener-
ally are φ. This is the structure of what I take to be the principle of generality.9
Boisvert maintains that an agent has a general attitude whether the predicate is
used in either making sincere simple moral claims (which for the hybrid expres-
sivist express simple moral judgments) or in embedded constructions involving
first-order logical operators.10 Along with this further assumption, generality pro-
vides a recipe for the attitude an agent has when moral predicates are used in com-
plex logical constructions: they always express the same one. As far as whether or
8. My emphasis on “actions in general.”

9. “Generality principle” is also the name Boisvert (2008, 171) provides for this idea as part
of his own hybrid view. I have reformulated it here as a broader principle that I take it is
compatible with Boisvert’s more specific formulation.
10. Boisvert (2008) refers to this as the “extensionality principle.”
not such a position is intuitively plausible, hybrid theorists have strongly pushed
on the analogy between moral language and other kinds of language that appear to
be dual-use, like slurs and other pejoratives.11 For instance, slurs seem to have sim-
ilar embedding features (they appear to express attitudes when embedded in any
logical construction by the speaker). This has been discussed elsewhere, so I will
not rehash the arguments here. But it seems plausible that other parts of language
may be dual-use, express attitudes with such general contents, and furthermore
express those same general attitudes in a wide variety of logically complex con-
structions. The associated generalized attitude (about whatever is φ) is the same
across simple judgments, conditionals, negations, disjunctions, and conjunctions
as well as constructions involving existential and universal quantification.
Now inferential relations can be explained rather inexpensively by appeal to
both the cognitive and the noncognitive elements. The acceptance of the claims
involved in a modus ponens argument with embedded moral predicates all involve
the same attitudes toward the same contents (whatever is φ).12 So explaining how
an agent should be committed to the consequent, given her or his acceptance of
a conditional and corresponding antecedent, is straightforward. The agent would
have to be in just the right states: she or he would have the neutral counterpart
belief (the content of which is the neutral counterpart proposition associated with
the consequent) and the general desire (which is invariable between the premises
and the conclusion). Insofar as this explanation generalizes, the hybrid expressiv-
ist can provide a straightforward explanation for embedding, and this was one of
the primary theoretical motivations for adopting a hybrid view.
In cataloging the benefits of recent specific hybrid positions, Schroeder
(2009) argues that this is precisely the attitudinal relation the hybrid expressiv-
ist should want to address embedding issues. All sentences containing moral
predicates should be associated with the same attitude (or at least the same one
for the same moral predicate) toward the same generalized content. By postu-
lating that the agent has this desire along with the corresponding belief, hybrid
expressivism can easily meet the theoretical constraint the embedding chal-
lenge places on expressivist views. On Schroeder’s analysis, this is an advantage
for a hybrid view that arises specifically out of its hybrid features.
In addition, the original Humean intuition can also be explained by appeal to
both the cognitive and the noncognitive elements. For instance, when an agent
makes the judgment that a particular action is morally obligatory, it follows from
this analysis that the agent has a resultant desire to do that particular action. For
11. Boisvert 2008; Copp 2001, 2009; Hay 2013.

12. Furthermore, arguments involving multiple moral predicates would express all of the
corresponding attitude types to the corresponding generalized contents.
instance, suppose that Timothy makes the judgment that donating 10% of his
2013 income to charity is morally obligatory. Timothy’s general desire that he
do whatever he is morally obligated to do along with his belief that he is mor-
ally obligated to donate to charity commits him to having the desire to donate
to charity. And once again, insofar as this desire to do that action motivates
Timothy to do it, then the hybrid expressivist has a straightforward account of
how moral judgments motivate. So hybrid expressivists who accept generality
can explain the Humean intuition and furthermore provide an explanation for
the practicality requirement. At this point it appears that hybrid expressivists
can be optimistic about “having it both ways” in the sense that their hybrid
view can be used to explain both how constructions involving moral predicates
embed in more complex constructions and how the beliefs and desires associ-
ated with moral judgments result in motivation.
To get these advantages, the hybrid expressivist must be committed to cer-
tain kinds of accounts about the contents of the moral thought and the way the
contents of the associated beliefs and desires are related. We can summarize the
features of such an account in the following way:
Requirement 1: Generality of Desire. The content of the associated desire

must be general, so that any simple or complex construction involving
a moral predicate must always express the same desire toward the same
content.
Requirement 2: Propositional Belief. The belief must have propositional
content that does the same sort of inferential work that ordinary descrip-
tive beliefs do.
Requirement 3: Content Relation. The content of these states must be
related in such a way that simple moral judgments result in the agent hav-
ing a desire to do something with respect to the action she or he has made
the judgment about.
Requirements 1 and 2 must be met by a hybrid view to provide an explana-

tion for embedding, and requirement 3 must be met to explain the Humean
intuition and in turn explain how it is that ordinary moral judgments are moti-
vating. In the next section I will examine the potential structure and content of
the general desire in more detail.
GE N E R A L I T Y A ND DE SI R E
In the last section the principle of generality was characterized by the claim that
the noncognitive states associated with moral thought and language (associated
desires) are directed at whatever is φ. I will turn now to discuss a variety of ways
this claim might be interpreted. I will consider four different candidate inter-
pretations for generality, and I will present what I think is the basic logical form
of each. In this section I will argue that different interpretations involving the
scope of the agent’s desire and the universal quantifier are problematic. In the
next section I will propose a different interpretation that is intended to avoid
these issues and address some of its further advantages.
I will begin with an example. Suppose my wife Flannery accepts, much to our
cat Lucy’s relief, that feeding Lucy at least twice a day is morally obligatory. On the
hybrid analysis Flannery must have the corresponding neutral counterpart belief
and a desire that she do whatever she is morally obligated to do. In reference to
the requirements laid out at the end of the previous section, we can say that:
(1) Flannery desires that she do whatever is morally obligatory. (Generality

Requirement)
(2) Flannery believes that feeding Lucy at least twice per day is morally
obligatory. (Belief Requirement)
(3) Flannery desires that she feed Lucy at least twice per day. (Resultant
Requirement)
I will represent the general form of claim (2) as: В [φf] and claim (3) as D
[ψf]. The question I will examine is: What does the generality claim made in
(1) amount to? What is it for Flannery to have such a general desire?
In explaining the generality principle, Boisvert (2008, 178) writes, “A speaker
has an attitude of the appropriate [general kind] just in case she is disposed,
upon being presented with things of the kind that he or she believes have the
properties rightness or wrongness, to approve or disapprove respectively of
those things.” This kind of dispositional account, based on the beliefs of the
agent, appears to have the following kind of structure. Flannery has the attitude
(1) if and only if for any action, if Flannery believes that it is morally obligatory,
then she desires that she do that action.
(1a) (∀x)(В [φx] ⊃ D [ψx])
If (1a) is equivalent to (1), then (3) is a logical consequence of (1) and (2).

The content relation requirement is met, but perhaps it is met too strongly;
the sense in which (1) and (2) necessitate (3) may be more psychological than
logical. Regardless, (1a) appears to more appropriately be thought of as a con-
sequence of Flannery’s general desire (1) rather than equivalent to or consti-
tutive of it. On the hybrid expressivist view under consideration, being in a
noncognitive mental state is a necessary feature of moral judgment. However,

(1a) is not itself a mental state but rather a general relation between them that
is supposed to hold for an agent.
One way of demonstrating this is to consider the case of a moral nihilist.
Upon matriculating to a university and signing up for a course on ethics, Nina
the Nihilist is asked about the wrongness of cheating on exams. Not only does
Nina see nothing wrong with cheating, Nina grandly proclaims, “No actions are
morally obligatory!” Nina’s utterance should express a state composed of her gen-
eral desire to do whatever is morally obligatory and her belief that no actions are
morally obligatory. The problem is not that Nina cannot have a general attitude
toward something that she believes does not exist. The problem is that for Nina,
while (1a) is vacuously true, it seems very strange to think she is in any desire
state at all as a result. The relation expressed by (1a) holds even though there is no
desire-like state Nina is in.
Perhaps one could try to rule out such a case by eliminating the agent’s beliefs
from the relation in (1a) and interpreting (1) as follows:
(1b) (∀x)(φx ⊃ D [ψx])
But this is clearly not promising. The generality principle should not be under-
stood as holding that for anything that is morally obligatory, the agent desires
that she or he do it. The wide-scope reading of the universal quantifier as in
(1b) is false insofar as Flannery is not morally omniscient. After all, there are
presumably a large number of actions that Flannery has not made moral judg-
ments about, and some of those may be morally obligatory, and Flannery may
make mistakes about which actions are morally obligatory.13
So perhaps the natural way of understanding (1) would be to hold that the
universal quantifier indicated by “whatever” should be seen as taking narrow
scope relative to Flannery’s desire.
(1c) D [(∀x)(φx ⊃ ψx)]
On such a reading, the content of Flannery’s desire could be understood as

unspecific or referentially opaque. Similarly, we can provide a referentially
opaque reading for the claim:
(4) I desire to pet all dogs in the park on Wednesday.
13. I mean to be neutral here about realist/antirealist positions about descriptive content of
moral predicates; on all but the most radical speaker-relative accounts, there is some space
between what the agent believes to be wrong and what is wrong.
I can have that desire even though I do not have a desire to pet any particular
dog. I just have a general desire to pet all the dogs in the park but no spe-
cific ideas about which ones will be there. In the parallel case for (1), Flannery
desires to do what is morally obligatory, but that does not entail that for any
action that is morally obligatory she desires to do it. The problem, though, is
that if this “but no specific one” reading is plausible, then this would allow for
situations in which the agent had the general desire (1), the specific belief (2),
but not the specific desire (3).
Flannery is making a mistake if she desires that she do whatever she is morally
obligated to do, believes that feeding Lucy at least twice per day is morally obliga-
tory, but does not desire to feed Lucy at least twice per day. Similarly, suppose
I explicitly assert 4 and 5:
(5) I believe Dakota will be in the park on Wednesday.
and explicitly deny 6:
(6) I desire to pet Dakota.
If that were the case, I would be making a kind of mistake. I could not get you
to excuse my apparent folly by just saying “But my desire to pet every dog in the
park is general and unspecific; there is no particular dog I desire to pet.” I can-
not simultaneously have a completely nonspecific general desire and also have
specific beliefs about which things fall into the category my desire is about. If
I have the desire as in (4) and the belief as in (5), then I should conclude (6).
The contents of my beliefs serve to specify the content of my general desire.
Returning to the case of moral judgments, whatever the relation between the
contents of (1) and (2) is, it must result, in the case of simple judgments, in the
agent having a desire to do something about the specific thing she or he has
made the judgment about.
Taking this into account, what if the consequent within the scope of the
desire in (1c) is made conditional on the agent’s belief?
(1d) D [(∀x)(В [φx] ⊃ ψx)]
Then the content of the agent’s desire is specified by her or his beliefs. Flannery
desires that for any action (or type of action) that she believes to be mor-
ally obligatory that she do it. This option fails to meet the original generality
requirement needed to address the embedding problem. Once again, suppose
Flannery accepts that feeding Lucy at least twice a day is morally obligatory.
Flannery then comes to accept the conditional premise “If feeding Lucy at least
twice a day is morally obligatory, then getting Ryan to do it is morally obliga-

tory.” Now, as soon as she accepts this conditional premise, she is in a state that
corresponds with the neutral counterpart proposition. Should she accept the
conditional premise, she will come to have the belief that the action of getting
me to feed Lucy is morally obligatory. But now the content of her desire has
changed. But recall that the primary theoretical advantage of accepting the gen-
erality principle was the idea that the agent’s desire and corresponding content
is identical between the premises and conclusions of a valid argument.
I will conclude with a brief overview of my points in this section. I presented
four ways we might interpret the generality principle as appealed to in:
Flannery desires that she do whatever she is morally obligated to do.
The question posed in this section was: What is it for Flannery to have such
a general desire? Whatever it is, it must meet the requirements presented in the
last section if it is to be appealed to by the hybrid expressivist. I presented four
candidates:
(1a) (∀x)(В [φx] ⊃ D [ψx])

(1b) (∀x)(φx ⊃ D [ψx])
(1c) D [(∀x)(φx ⊃ ψx)]
(1d) D [(∀x)(В [φx]⊃ ψx)]
(1a) may satisfy the content relation requirement but does not adequately explain
what it is to have the general desire. (1b) is an implausible wide-scope reading of
the principle. It is false, since the agent may not have desires about everything that,
for instance, is morally obligatory. (1c) is an implausible narrow-scope reading of
the principle; there are specific actions that her desire is about, namely the ones
determined by her beliefs. (1d) has a variable content dependent on the beliefs of
the speaker, and this changes when her beliefs change. In the next section I will
consider a completely different interpretation for the generality requirement. I will
then turn to argue that there are still advantages for the hybrid expressivist view if
the generality requirement is given the proposed interpretation.
T Y P E -G E N E R A LI TY AND M O TI VATI O N
The generality principle must be formulated in such a way that ensures there
is a “wall of separation” between the contents of the cognitive and noncogni-
tive states associated with moral thought and language. I propose that one way
to do this is to suppose that an agent, rather than desiring to not do anything

that has the property of wrongness, just disapproves of the type of actions that
are wrong. In this section I will explain how this is different and maintain that
interpreting the generality principle in this way avoids the issues raised in the
last section. In addition, I will show how it deals with nihilists, amoralists, and
the apparent connection between moral judgment and motivation.
The first maneuver here is to maintain that the general content of the non-
cognitive state associated with moral thought is not propositional. Rather, it is
desire-like states just like those that by themselves may be directed at objects.
Approval and disapproval appear to work this way. An agent may disapprove of
things as varied as Abraham Lincoln, the assassination of Lincoln, politicians,
or assassinations. The original generality principle can be formulated in terms
of these states. It would hold that the disapproval state associated with moral
thoughts involving “wrong” is just disapproval of anything that is wrong. I will
represent disapproval with the “disapproving look” symbol “ ಠ_ಠ ” rendering
generality as ಠ_ಠ [(∀x)(φx)]. But given that formulation involving a universal
quantifier, all of the same problems involving relative scope discussed in the last
section arise once again.
My proposal here in a sense modifies generality to make it even more gen-
eral. The content of the disapproving state is not whatever has the property of
being wrong (regardless of whether or not “whatever” is given an unspecific
reading). Instead, the content is just the type—by which I mean the abstract
quality shared by wrong things.14 I will call this proposal type-generality and
represent the type of action associated with a moral predicate φ as Tφ.15
Type-Generality: (∀x)(MS[φx] ⊃ ಠ_ಠ[Tφ])
But what does it mean to say that the content of the state can just be a type?
14. The idea is that types can play the kind of single object role and can be distinguished
from the universal property sense. Types, like single objects, can bear relations to individu-
als in the way that properties or sets cannot. One can disapprove of the types of things that
are wrong in a way that one cannot seemingly disapprove of the property of wrongness or
perhaps even less plausibly the set (itself) of wrong actions. If that works, it would fall in line
with a type/universal distinction addressed in Wollheim 1968, 76–77, where predicates and
relations that hold for tokens also hold for types.
15. The view can allow for higher-order “type-of-types” disapproval as well. The disapproval
here may be directed at the type of types of actions that are wrong, from which disapproval of
types of actions psychologically follows, from which disapproval of token instances of such
types psychologically follows.
Given that the noncognitive states are approving or disapproving states,

I will demonstrate this point with a nonmoral example. Consider a person who
despises a particular color, for instance, an interior designer who disapproves of
beige. In general such an individual, when presented with specific beige things
(or at least things that appeared before her senses to be beige), would disapprove
of those things. And perhaps on that basis it would appear correct to character-
ize her negative attitude as being about anything that is beige generally. But it
does not follow from her disapproval of beige that she disapproves of any beige
thing. It still may very well be the case that she actually approves of some par-
ticular things that are beige. Not because she has mistaken beliefs or is uncertain
about what is beige but because she can make exceptions for particular objects,
even if they are of a type she disapproves of. She might like a particular beige
Ikea sofa because that is the only sofa of a certain variety in stock (I did not say
she was a good interior designer). She disapproves of its beigeness, but this is
compatible with her approval of the sofa. Of course her disapproval of beige is
cited when explaining why she does not like many other beige things. The point
is that an agent may disapprove of a type without being committed to disapprov-
ing of anything that has the associated property.
One advantage of this kind of hybrid structure is that it avoids the scope
interpretation issues discussed in the last section. The content of the desire ele-
ment is related to the belief element in that they both involve content identified
by the use of a moral predicate. But since the belief has a propositional content
and the desire does not, there is no problematic entangling of these contents.
Making a new moral judgment entails coming to be in a state with a different
cognitive element, but the noncognitive element remains the same, though it
is still of course required. What changes, as far as the noncognitive element is
concerned, is the resultant approving or disapproving state the agent is in aris-
ing from her new belief and her very general type-directed attitudinal state.
Another advantage is that this view can make sense of the thoughts and
claims of moral nihilists. In the last section, on a dispositional attitude inter-
pretation of the generality principle (as in [1a]), Nina’s belief that nothing is
right seemed to imply she possessed no general attitude at all. Suppose that
Nina also claims “No actions are wrong!” On a type-generality-based hybrid
expressivist view, Nina’s utterance, if sincere, expresses a state composed of
(1) Nina’s neutral counterpart belief that there are no wrong actions and
(2) Nina’s disapproval of that type of action. In this case there is no conflict in
the compositional content of Nina’s state, though it may be a sort of strange
position to be in. She can have such a belief (it may very well be factually
incorrect), and she can disapprove of the type of things that are wrong. That
might seem strange, but we can also consider other cognitive-noncognitive
state pairs that are allowable in this way.
One can positively believe that no trolls exist but still fear trolls. Such an
individual does not fear any troll, as she believes none exist; the object of her
fear is the type. Given her belief, the unspecific universal reading (that she fears
any troll but no particular one) strikes me as simply incorrect. Again, we might
think that such a person is strange for having such a fear, but it does not rise
to the level of being the kind of failure discussed in the last section. Types can
be treated as objects that can be feared, loved, approved of, or disapproved of
regardless of whether, from the intentional perspective of the agent, tokens or
occurrences of the type exist.
This idea can also be appealed to in order to explain how certain agents can
fail to be motivated even when they sincerely think that something is wrong.
The adoption of type-generality by the hybrid expressivist makes room for the
idea that Nina may still have the requisite noncognitive state but not be moti-
vated to act in any way. If Nina disapproves of the type of actions that are wrong
but does not believe that anything has the property of wrongness, she likewise
will not be motivated to act. Furthermore, the idea can explain why in more
typical cases agents may fail to be motivated in every case where they think
that actions are wrong. Generally, if one believes that an action has a certain
property and one disapproves of the type of action that has it, then one would
disapprove of that action. Insofar as disapproving of an action motivates one
to avoid doing it (or to get others to avoid doing so in addition), the resultant
disapproval state is then motivating.
So if I believe that eating meat has the property of wrongness and I disap-
prove of the type of actions that are wrong, then it seems as though I should
disapprove of eating meat, but this is not necessarily so and certainly not as a
matter of logical consequence. Since eating meat has other qualities that I do
not disapprove of, I am not committed to disapproving of that action. Of course
there must be some variance in the extent to which I can consider actions
wrong that also bears relation to the extent to which I think certain actions
have that property and the intensity of my resultant attitude. Presumably, I can
make exceptions for meat eating but not for, say, genocide. Of course one can
be mistaken about the extent to which things are wrong, and this corresponds
directly to the intensity of my resulting motivation.
If that is right, then this may be a way the hybrid expressivist could appeal to
type-generality to avoid some of the issues with strong motivational internal-
ism. Of course being able to explain the motivating features of morality is, as
I stated at the outset, one of the theoretical advantages of a hybrid view. Views
such as Ridge’s (2006) version of “ecumenical expressivism,” Boisvert’s (2008)
“expressive-assertivism,” and Stephen Barker’s (2000) conventional implicature
view all would fall into the category of holding that moral judgments and cor-
responding sincere utterances involve being in states that will result in the right
cognitive and noncognitive elements to result in being motivated. One worry,

though, is that since hybrid expressivists end up committed to the idea that
agents are in just the right belief/desire states to be motivated when they make
moral judgments and sincere moral claims, they cannot account for amoralists
and other standard cases where it looks as though moral judgments are made
but motivation is lacking.
Stephen Finlay (2004, 2005) is highly skeptical about internalism for this
reason. Finlay (2004), who endorses a cognitivist view but holds that attitudes
are communicated conversationally, maintains that “being in certain motiva-
tional or attitudinal states is not a necessary condition for making genuine
moral judgments.”Dorit Bar-On and Matthew Chrisman do not go this far
but also think that amoralist-style cases should call into question the build-
ing of motivational internalism into the semantics of moral language. Instead,
they distinguish between different senses of expression and see motivation as
a necessary condition for making genuine moral judgments but not a nec-
essary condition for understanding the semantic rules governing the use of
moral language.16 And Copp’s “realist-expressivism” may leave open the pos-
sibility of just the opposite; that a speaker is in a corresponding motivational
state may follow from what she asserts and conversationally (s)implicates, but
she can think that actions have certain moral properties with necessarily pos-
sessing corresponding attitudes. Needless to say, intuitions and corresponding
theoretical options vary widely across those who have endorsed hybrid and
hybrid-like views.
Of course the hope is that the kind of type-generality hybrid expressivism
I am considering in this chapter may explain the general connection between
moral judgment, corresponding utterances, and motivation in a way that does
not commit itself to the strongest form of motivational internalism. It allows for
cases where agents make judgments about the moral quality of an action but
make an exception for it.
16. On their view making a moral judgment involves both (thought) a-expression of a moti-
vational state and (thought) s-expression of a descriptive content, but since a-expression
is external to the meaning of moral language, there is no internal connection between the
meaning of moral language and motivational states. Bar-On and Chrisman’s position is that
motivational internalism about meaning is too strong; the failure to be in an appropriate
motivational state while labeling a particular action as wrong is on their view neither a con-
ceptual impossibility nor a failure in the mastery of the term “wrong.” Instead, using “wrong”
without being in the appropriate motivational state violates what they call the “propriety
conditions” for the use of the term. They conclude that “denying that the connection to moti-
vation is forged in the semantic content of ethical claims makes room for the following con-
ceptual possibility: a person could issue a claim such as ‘torturing cats is wrong,’ . . . sincerely
and while understanding what it says, yet without being motivated to act in accordance with
it” (Bar-On and Chrisman 2009, 144).
That being said, it is not clear that the type-generality view can avoid one
typical kind of example that may push us toward externalism. That kind of
example involves an agent who is capable of picking out moral properties and
uses moral terms to do so but simply lacks the (very general) noncognitive
attitude. In contrast to the moral nihilist, consider the following amoralist case:
Sally the Sociopath: Every year, Sally travels to a large city, sneaks into
a high-rise building, and defenestrates an unsuspecting maintenance
worker. She feels absolutely no remorse. She does not disapprove at all
of defenestrating maintenance workers. Finally Sally is caught and under
intense questioning she admits, “What I did was wrong!”
On the type-generality view Sally’s claim, if sincere, would express a state

composed of her belief that her action has the property of wrongness and a neg-
ative attitude toward the type of actions that are wrong. Now, Sally’s admission
might seem suspect, but suppose that she has the neutral counterpart belief.
Perhaps she has it because she has witnessed others use the term to describe
certain actions, and she has come to understand not only what falls under the
extension of the term but also how to use it and apply it as well as nonsociopaths
in identifying actions. Sally can perfectly well have this belief without any dis-
approving attitude toward that property. But type-generality will be committed
to the view that Sally cannot really think defenestrating high-rise maintenance
workers is wrong. Finlay finds this kind of analysis wanting. He writes, “The
amoralist appears to be asserting the same thing as the rest of us when ascribing
moral goodness, and the nondescriptivist’s only reason for denying this seems
to be that it is incompatible with nondescriptivism” (Finlay 2004, 209).
But in this case there does seem to be something quite off about Sally as
far as her moral understanding goes. She may be able to master the extension
and intension of “wrong,” but she just does not think about it or use it in the
same way that ordinary moral agents do. The failure is still present even if
Sally proclaims that “no actions are wrong” while also failing to disapprove
of anything. Even in that case Sally is missing a key part of the understanding
about what it is to employ moral concepts and terms in thought and language.
She does not really understand what it would be for anything to be wrong. In
much the same way, if we imagine a nonbigot in the land of the bigoted, she
may use a slur for a class of people as merely a way of identifying them. She
can think and speak about the class of people by using the term, but she does
not have the required attitudes to employ it with its full bigoted meaning. In
an analogous sense Sally can think and speak about actions by using the term
“wrong,” but she lacks the required very general attitude to employ it with its
full moral meaning.
Since Sally lacks the general attitude, she is just not motivated to avoid
defenestrating maintenance workers. If Sally only has the belief that defenes-
trating maintenance workers has the property of wrongness, her belief alone
is not motivating. Along a similar line, Dreier asks us to imagine a group of
people who recognize certain natural facts about what is right or wrong (or
good or bad) but simply do not care about such things. He tells us that in this
case, “if they have a term that applies to all and only things with those natural
properties, it might have the same extension, intension, the same content as our
word ‘good,’ but there is a sense in which it does not mean what ‘good’ means”
(Dreier 1990, 8). After considering examples of amoralists, Dreier maintains
that they appear most plausible in virtue of the fact that they stand in contrast
to the normal moralist who has normal motivations. But this still allows room
for individuals in, say, a subcommunity of amoralists to sensibly use a term like
“wrong” without its normal motivational force but still pick out the property of
wrongness (Dreier 1990, 11–14). In this way what Dreier labels the “internalist
character” is still part of the meaning of the moral language, but every indi-
vidual who uses the term is not necessarily motivated.
This response mirrors how I think the hybrid expressivist can respond to amor-
alist cases like Sally the Sociopath. That being said, the motivational connection
supposed here by type-generality is even weaker than that which Dreier supposes.
On the view under consideration, there may be cases where agents really do have
the requisite attitude but are also not motivated. Approving of the type of actions
that are right does not necessarily motivate one to do every particular action that
the agent believes to be right and that she or he is capable of carrying out.
If anything the type-generality view I consider in this final section is an
attempt to accommodate multiple theoretical concerns that arise in trying to
have things both ways. In trying to get some of the main theoretical advantages
of both cognitivism (embedding) and noncognitivism (motivation), we are left
with a view that ends up requiring a very general attitude; a general attitude that
is present in the vast majority of people who do have moral sentiments. How
did we get here? Well, any view that is in any way expressivist (in the sense that
noncognitive attitudes are tied to moral claims) faces the embedding challenge.
Without proposing a novel logic of attitudes, the challenge can be met easily by
a hybrid view that meets a kind of generality requirement. But the content of
the proposed generality requirement must not be such that it either shapes or
is shaped by the content of the cognitive element. I have argued here that mod-
ifying (or at least further articulating) the generality principle is required to
ensure that there is no entangling of the content of the associated cognitive and
noncognitive states. Furthermore, the view that accepts this type-generality has
the advantage of being able to explain why agents are typically motivated by
moral judgments, but such motivation is not necessary in all cases.
To get this type-generality, the hybrid expressivist must make certain fur-
ther commitments about the associated noncognitive state and its correspond-
ing content. As always, having it both ways has involved making such further
commitments to retain desired explanatory advantages. The hope is that the
further commitments do not render the position simply intuitively implau-
sible. I think there are two ways to take the suggestions made here. The first,
more optimistic position is that a view with this content, structure, and men-
tal and metaphysical commitments is plausible, and the analogy with other
seemingly dual-use states supports it. On the other hand, for the opponents
of hybridism, perhaps the structure endorsed here is further evidence that the
theoretical commitment of hybrid expressivism must become so exotic that it
is intuitively untenable. 17
References
Barker, Stephen. “Is Value Content a Component of Conventional Implicature?”
Analysis 60, no. 3 (2000): 268–279.
Bar-On, Dorit, and Matthew Chrisman. “Ethical Neo-Expressivism.” In Oxford Studies
in Metaethics, vol. 4, edited by Russ Schafer-Landau, 133–165. New York: Oxford
University Press, 2009.
Boisvert, Daniel. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89, no. 2
(2008): 169–203.
Copp, David. “Realist-Expressivism: A Neglected Option for Moral Realism.” Social
Philosophy and Policy 18 (2001): 1–43.
Copp, David. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies
in Metaethics, vol. 4, edited by Russ Schafer-Landau, 167–202. New York: Oxford
University Press, 2009.
Dreier, James. “Internalism and Speaker Relativism.” Ethics 101, no. 1 (1990): 6–26.
Finlay, Stephen. “The Conversational Practicality of Value Judgment.” Journal of Ethics
8, no. 3 (2004): 205–223.
Finlay, Stephen. “Value and Implicature.” Philosopher’s Imprint 5, no. 4 (2005): 1–20.
Geach, P. T. “Assertion.” Philosophical Review 74, no. 4 (1965): 449–465.
Hay, Ryan J. “Hybrid Expressivism and the Analogy between Pejoratives and Moral
Language.” European Journal of Philosophy 21, no. 3 (2013): 450–474.
Hume, David. A Treatise of Human Nature. Project Gutenberg, 2010 [1740]. http://
www.gutenberg.org/files/4705/4705-h/4705-h.htm.
Ridge, Michael. “Ecumenical Expressivism: Finessing Frege.” Ethics 116, no. 2
(2006): 302–336.
17. I would like to thank all of the participants of the Hybrid Theories in Metaethics
Conference that took place in Edinburgh in the summer of 2012, at which an early version of
this chapter was presented. I am particularly grateful to Michael Ridge and Guy Fletcher for
hosting and organizing the conference and for the insightful comments made and questions
posed by the participants.
Schroeder, Mark. “Expression for Expressivists.” Philosophy and Phenomenlogical

Research 76, no. 1 (2008): 86–116.
Schroeder, Mark. “Hybrid Expressivism: Virtues and Vices.” Ethics 119, no. 2
(2009): 257–309.
Smith, Michael. The Moral Problem. Oxford: Blackwell, 1994.
Wollheim, Richard. Art and Its Objects. New York: Harper and Row, 1968.
5
Diachronic Hybrid Moral Realism

JON TRESAN n
This chapter introduces a novel hybrid metaethical view and describes its unique
advantages. A classic metaethical challenge is capturing the ways morality is
objective and the ways it is practical, without extravagance: the best balance of
objectivity, practicality, and frugality. Hybrid metaethical views are attractive
because they can help with this project. The advantage of the view presented
here is that it allows for otherwise impossible combinations of extremes of
objectivity, practicality, and frugality.
One such extreme is that it incorporates an objectivistic version of moral
realism. So the first section introduces the view by comparing it to existing ver-
sions of hybrid realism. For reasons explained there, I call the view “diachronic
hybrid moral realism” (DHR).
The second section explains the key to its ability to reconcile objective real-
ism with extremes of practicality and frugality. This key is that it breaks the
entailment, by objective realism, of the content-individuation of moral judg-
ments. It is natural to suppose that if objective realism is true, then for a judg-
ment to be moral is for it to be about moral reality and for a judgment to be of
a certain moral type is for it to be about the relevant part of moral reality. For
instance, suppose there is an objectively real property of moral wrongness. It
would seem to follow that judgments about what is morally wrong are precisely
those about what has that property. But if DHR is correct, though objective
realism is true, moral judgments are not content-individuated.
The third and fourth sections show the extremes of practicality and frugality
(respectively), which can be reconciled with objective realism once it loses its
commitment to content-individuation. The third section shows how objective
realism without content-individuation can be combined with a robust func-
tionalism about the moral realm. Functionalist views identify a characteristic
cluster of pro/con-attitudes and practices and say that being moral (vs. non-
moral) is being among or relevantly related to those. For instance, we use moral
judgments to display, guide, and mobilize co-targeted non-ulterior attitudes
like indignation and guilt, and on one functionalist view their being so used is
what their being moral judgments consists in. Combining objective realism and
functionalism is difficult—if objective realism entails content-individuation.
The difficulty goes away if it does not.
Section four describes an extremely frugal metaethical position and argues that
it is easier to combine with objective realism if content-individuation is dropped.
It is well known that metaethical views can be more or less metaphysically frugal.
For instance, though everyone posits natural properties, some go beyond this
and posit nonnatural ones. Metaethical views are also more or less anthropologi-
cally frugal. For instance, every metaethical view acknowledges a human (i.e.,
species-typical) tendency to think, talk, and care in characteristically moral ways
about objectively real behavioral properties. But they differ regarding whether
humans as such tend to converge on the same properties. Antirealists say that—as
with such things as food, music, and sexiness—behavioral cathexes converge at
the human level only on domains, with specific profiles determined by personal
and cultural idiosyncrasies. Objective realists, on the other hand, tend to say that
humans think, talk, and care about the very same properties, for example, wrong-
ness. Even naturalistic objective realists, though they deny that moral properties
are metaphysically special, suppose that they are in this sense humanly special.
Such a commitment carries a heavy burden, as appearances suggest no such con-
vergence. Nevertheless, objective realists seem forced to it—if they are committed
to the content-individuation of moral judgments. If they are not, they can avoid
the burdensome posit of humanly special moral properties.
A major motivation for functionalism, and a thorn in objective realists’ side,
are certain interpretative tendencies. I mean our tendencies regarding the inter-
pretation of judgments as moral judgments or not and as moral judgments of
a specific type. These tendencies seem role sensitive in the sense that the judg-
ments’ relations to attitudes and practices directly influence our confidence that
they are moral. Metaethicists have explored these tendencies, reflecting on two
kinds of cases: amoralists and moral twins. With amoralists we explore how
the absence of the relevant role can make us less confident that the judgments
are moral. For instance, if judgments are never accompanied by co-targeted
con-attitudes, are we less apt to interpret them as wrong-judgments (judgments
about what is morally wrong). With moral twins we consider whether judg-
ments playing a certain role makes us more confident that they are moral. For
instance, we take twin-wrong-judgments to be wrong-judgments, despite sig-
nificant and fundamental differences between twin-wrong-judgments and our
wrong-judgments, for example, in which behaviors they target (are made about).
Diachronic Hybrid Moral Realism 97
A natural suggestion is that competent folk interpret these ways because moral
judgments and wrong-judgments are in fact those that play the relevant role.
I suggest that these interpretative tendencies have split philosophers into
two fundamentally different camps, “demythologizing antirealists” and
“mythologizing realists.” The shared assumption that objective realism entails
content-individuation forces this split. For it is hard to explain why we interpret
judgments as wrong-judgments when and only when they play the relevant role
(roughly), consistently with content-individuation, without taking wrongness
to move or interest us in some special way. We are forced either to excessive
demythologization (objectively real moral properties are a myth) or indulgence
of myth (objectively real moral properties are special). If so, then by letting us
have objective realism without content-individuation, DHR alone makes a fully
demythologized objective realism possible.
S Y N C H R O N I C AN D DI AC HR O N I C HY BRID
MO R A L R E A LI SM
DHR is in deep respects similar to existing versions of hybrid realism, such as

those of David Copp and Daniel Boisvert.1 As noted, hybrid views in general
are motivated by their contributions to giving us the best balance of objectivity,
practicality, and frugality. Hybrid realist views help with this project by exploit-
ing six truths:
HR1. Some claims about descriptive moral semantics suffice for moral real-
ism. By “descriptive moral semantics” I mean the aspects of the mean-
ing of moral terms relevant to their truth-conditional contribution when
they are used for moralizing. Arguably any view according to which moral
terms have a descriptive moral meaning and that adds that the truth con-
ditions in question are sometimes satisfied is realist. But whether that is
so need not worry us, as DHR posits an especially objective descriptive
moral semantics. It holds that there is an objectively real (instantiated,
nonrelative) property W that exhausts the descriptive moral meaning of
“wrong.” That is, “wrong,” when used for moralizing, contributes W to the
truth conditions of sentences, and that is the only truth-conditionally rel-
evant aspect of its meaning. It has no descriptive context-sensitivity, no
Kaplanian character, no primary versus secondary intension, and so forth.
“Wrong” is in this way about W as “spherical” is about sphericality. (And
1. Boisvert 2008; Copp 2001.

mutatis mutandis for other moral terms.) That surely suffices for realism,
and henceforth I use “realism” to mean this objective realist view. (So as
used here the “antirealist” label covers even contextualist and relativist
realists.)
HR2. The descriptive moral semantic claims sufficient for realism are com-
patible with moral terms having further semantic features. Boisvert and
Copp suggest that moral terms also have an expressive moral meaning.
For example, Boisvert suggests that “x is wrong” means something like “x
is F; and down with F things!” Since the descriptive element suffices for
moral realism, the additional expressive meaning does not conflict with
realism. It just means that we talk about moral properties in a biased way,
a bias that has shaped even the meanings of the terms we use to talk about
them. If humans fell in love with mass and the term “mass” took on a posi-
tive expressive element, this would not undercut realism about mass.
HR3. Some such further semantic features would, if had by moral terms,
underwrite a more robust account of morality’s practicality. For instance,
if calling an action “wrong” does not just ascribe a property to it but
expresses a con-attitude, we can better understand why there is such a
strong correlation between such judgments and attitudes. Indeed we can
better understand doubts about whether one could judge that x is wrong
without having the relevant con-attitudes and thus can better understand
attraction to the “motivational internalist” idea that such a correlation is
necessary. (We return to this in the third section.)
HR4. The more robust account of morality’s practicality they underwrite is
itself a reason for believing those further semantic claims, insofar as we have
reason to believe that morality is practical in the relevant way, that realism
is true, and that there is no better way of accommodating that practicality
consistently with realism. For instance, if realism otherwise renders inex-
plicable our doubts about whether one could judge that x is wrong without
con-attitudes, then the existence of those doubts is itself reason to posit
the further meaning.
HR5. Because they are posited in addition to—and not instead of or as an
aspect or implication of—the realism-sufficient descriptive moral seman-
tic features, the further semantic features carry none of the standard threats
to moral objectivity. Insofar as those threats involve semantic claims, they
are about descriptive moral semantics only. For instance, many worry
that moral semantic contextualism undermines our ability to accommo-
date important aspects of morality’s objectivity. And contextualism is an
account of descriptive moral semantics (truth-conditional contributions
are context sensitive). Similarly, insofar as relativist accounts of moral
properties have semantic implications, they all pertain to the descriptive

moral content of moral terms (e.g., “wrong” is about a relative property).
HR6. Because this combination of realism and practicality is achieved via
semantic posits, no questionable properties are required. Combining real-
ism and motivational internalism has been thought to require such things
as necessarily motivating properties. But no such things are required by
the semantic posits in question. That “wrong” has some meaning in addi-
tion to its being about a property W tells us nothing about the nature of W.
Where DHR and existing versions differ is in the nature of the “further seman-
tic features.” There are two ways of having two meanings: a term can exhibit both
meanings on the same occasions of use or switch back and forth between them. So
in the case at hand the second meaning could be a further, nondescriptive moral
meaning (exhibited when used for moralizing). Or it could be a meaning exhibited
when moral terms are used for something other than moralizing. Existing hybrid
realist views are of the former, synchronic sort. What is distinctive about DHR is
that it says moral terms exhibit different meanings on different occasions, that they
are polysemous (ambiguous).2 Hence diachronic hybrid realism.
Note that synchronic and diachronic hybrid realism are not incompatible.
Moral terms could have two meanings when used for moralizing and a further
nonmoralizing meaning. Indeed I find synchronic hybrid realism plausible in
addition to DHR. Thus as used here “synchronic hybrid realism” means mere
synchronic hybrid realism (i.e., without DHR).
DHR posits a pervasive, entrenched, intelligible, unnoticed polysemy in
moral terms. When we use them for moralizing, they mean one thing; when we
use them for metamoralizing, they mean another. “Metamoralizing” is think-
ing and talking about moral attitudes and practices. For instance, when we say
“they think dancing is morally wrong,” we use a moral term (“morally wrong”).
But we are not moralizing. Error theorists metamoralize without contradiction.
In a nutshell, what DHR says is that when we use moral terms for moraliz-
ing, they have the realist meaning indicated above. But when we use them for
metamoralizing they have an attitudinal or functional meaning along the lines
posited by emotivists and antirealist expressivists. For instance, “they think
dancing is morally wrong” ascribes to “them” a certain sort of negative orienta-
tion to dancing: they are against it in a characteristic way.3
2. Polysemy is noncoincidental ambiguity; e.g., “bank” is not polysemous, because the two mean-
ings are not related, but “run” is multiply polysemous, because the various meanings are con-
nected (“run a mile,” “run for office,” “run in a stocking,” “run the motor,” “run your mouth,” etc.).
3. If this is correct, then metamoral uses of “morally wrong” have a semantic impact on
the phrases in which they are embedded: if the only shift were in the meaning of “morally
Although DHR is novel, two of its major elements are quite familiar: function-
alism about metamoral semantics and the posit of a moral/metamoral polysemy
in moral terms. Antirealists tend to be functionalists, for example, prescriptiv-
ists, expressivists, projectivists, and contextualists. And everyone recognizes that
“moral” is moral/metamoral polysemous. When moralizing, the contrast is with
“immoral,” when metamoralizing with “nonmoral” (“moral judgment”).
DHR extends these familiar ideas in two ways. First, it combines them with
realism about moral semantics. Second, it generalizes. Not just “moral” is moral/
metamoral polysemous with a realist moral meaning and a functionalist metamoral
meaning. Moral terms in general are like this.
DHR faces serious challenges. This is presumably the main reason why it has
been ignored by metaethicists: whereas only a glance is needed to see the chal-
lenges, much more thought is needed to see that they might be met. I cannot argue
wrong,” then “they think x is morally wrong” would mean they think x is something (though
not W), not that they are against x. The effect is similar to that which occurs in idioms.
For instance, “bucket” in “he kicked the bucket” does not just mean something different
than it does in “grab me that bucket,” it changes the meaning of the entire phrase: “kicked
the bucket” does not just mean kicked something (only not a bucket). However, I should
make explicit that not all metamoralizing uses of moral terms involves embedding in famil-
iar metamoral phrases. R. M. Hare seems right that we sometimes use moral terms as if in
inverted commas, even when the linguistic form of our utterances suggests moralizing. The
Simpsons episode “Itchy and Scratchy and Marge” provides a nice example. Marge has led a
successful public protest against cartoon violence, creating a problem for the makers of The
Itchy and Scratchy Show ("Itchy & Scratchy & Marge" The Simpsons. Fox. 20 Dec. 1990. Web
Transcript), who call her for help:
Roger Meyers Jr. [owner of the studio]: Listen you’re so smart, how do we end
this picture?
Marge: Hmm . . . well, what’s the problem you’re having?
Roger Meyers Jr.: Ok, here it is: Itchy [a mouse] just stole Scratchy’s pie and Scratchy
[a cat] is understandably upset.
Marge: Uh huh.
Roger Meyers Jr.: So we figured he could, you know, just grab Itchy and toss him in
a bucket of acid.
Marge: Oh, dear!
Roger Meyers Jr.: But then we remembered that this might be interpreted as violence,
which is morally wrong now thanks to you. So, what’s your big idea? How do we end
this?
Marge: Hmm . . . let’s see. Umm . . . oh! Couldn’t Itchy share his pie with Scratchy?
Then they would both have pie! (Italics added)
Evidently, Meyers is not to be understood as making a moral judgment to the effect that
cartoon violence has become wrong, thanks to Marge. What he means is that she has turned
people against it.
for this here, but I will sketch my reply to what I take to be the main challenge.4
Much can be said briefly and makes DHR more intelligible and plausible.
DHR posits something surprising, nonidentity in moralizing and metamoral-
izing uses of “wrong” and other moral terms. This is surprising, because the rela-
tion between the two uses seems so like the ordinary relation between first-order
and metalevel uses of a term. We describe things as “red” and ascribe beliefs about
what is “red,” and “red” has the same meaning in both uses. Of course “red” is
itself ambiguous (color, commie). But for each first-order meaning there is a cor-
responding metalevel use in which it has that same meaning. But DHR suggests
that that is not true of moral terms. When we use “morally wrong” for moralizing,
it means W, but “judgment about what is morally wrong” does not mean judg-
ment about what is W. This raises two questions. First, why do we not talk about
W judgments—or if we do, what terms do we use and why those? Second, even
assuming it makes sense for us to acquire some terms with the functional meaning
DHR ascribes to moral terms in metamoralizing uses, why those terms, given their
preexisting nonfunctional moral meanings? (Why not terms explicitly about roles
or if necessary new terms?)
My answer to the first question is that we do talk about W judgments and we
do so by using moral terms. “Wrong” in metamoral uses does not always have
functional meaning; sometimes it does mean W. When metamoralizing in dis-
course contexts in which moral attitudes and practices are of interest, we use moral
terms functionally; in contexts in which moral properties are of interest, we use
them for those properties. For instance, suppose we are genuinely interested in
whether some action is morally wrong (has W, ex hypothesi). I seek out Jones’s
advice, because we take her to be a reliable moral judge, and report her as “judg-
ing it wrong.” In such a context we are interested in her judgment as relevant to
W, so we characterize it as such: my report means that she judges it to have W.5 It
is because our interest in moral attitudes and practices is so much more pervasive
than our interest in moral properties that I say “the” metamoral meaning is func-
tional. Really the suggestion is that its standard metamoral meaning is functional.
As is our wont, we naturally and unreflectively disambiguate.6
4. There are familiar challenges to realism and metamoral functionalism, and of course DHR
faces these (except the challenges to each that stem from the other). So far as I am aware, the
only challenge that is distinctive to DHR is that mentioned in the text.
5. This suggestion of contextual disambiguation can be empirically tested: we can see how,
if at all, role-sensitive interpretative tendencies are influenced by discourse contexts. The
hypothesis predicts that they will be weaker in contexts in which moral facts are of interest
than those in which attitudes and practices are.
6. One of the worries about DHR is that it entails the truth of some sentences of the form
“they truly judge x-ing morally wrong, though x-ing is not morally wrong.” These sentences
My answer to the second question is that when the need for functional
metamoral language arose, moral terms were well positioned to be recruited to
serve that need by a common process of semantic change. The process involves
two stages: a certain way of using language figuratively and the conventional-
ization of figurative language so that it becomes literal, now meaning what it
used to pragmatically convey.
The figurative use of language involves a certain type of metonymy. In meton-
ymy we use terms for things associated with X as terms for X. This is very common.
For instance, “the White House announced yesterday . . . ,” “do you take plastic?”
“I’m reading Shakespeare,” “the ham sandwich is getting impatient . . . ,” “lend
me a hand,” and so forth. In each case we explicitly mention something saliently
associated with what we want to convey information about: the White House is
associated with the administration, plastic with credit cards, William Shakespeare
with his works, the ham sandwich with the person ordering it, lending a hand with
helping out with one’s body. Terms for the former are metonymic vehicles used to
convey information about an implicitly referenced metonymic target.
The type of metonymy at work with moral terms exploits the salient asso-
ciations between certain tokens of a type and the type itself. The metonymic
vehicle is thus a term for a token, the metonymic target the type of which it is
a salient token. This is a common type of metonymy, an obvious instance of
which is brand-name genericization, as with “Kleenex,” “Band-Aid,” “Hoover,”
and “Rollerblade.” In these cases we use a term for a prominent token (brand)
of a type as a term for the type (tissue, adhesive bandages, vacuum cleaners,
in-line skates). Another example involves the use of historical events or persons
as names of types of events or persons. For instance, “9/11” is used as a met-
onymic vehicle for devastating surprise attacks, as in “no more 9/11s”; “Benedict
Arnold” and “quisling” are vehicles for traitors; “Hitler” for bloodthirsty mega-
lomaniacs (“another Hitler”); and so on. Token-for-type metonymy is pervasive
and intelligibly so given the salient link between certain tokens and types. 7
sound bizarre. But the point made in this paragraph gives us an explanation. Since the
first, metamoral “morally wrong” must be read functionally (otherwise the sentence is
self-contradictory), the discourse context must be one in which our interest in attitudes and
practices is greater than our interest in moral facts. But assessments of the truth value of
moral judgments tend to occur in contexts in which moral facts are of primary interest. The
sentence thus suggests incompatible discourse contexts. And if the context is set up to invite
simultaneous functional metamoralizing and truth value assessment, it is not obvious that
the sentences will seem so odd (Sarkissian et al. 2011). Note that confusion caused by poly-
semy is a well-known phenomenon; see, e.g., note 7.
7. For useful discussions, see Falkum 2011; Glucksberg 2001; Koskela 2011. Sam Glucksberg
(2001, 40) reports a conversation that illustrates the kind of confusion to which type-for-token
metonymy can lead:
If realism is true, then when we moralize with “wrong” we mean W. But

our judgments have features other than their meaning. In the case of
wrong-judgments, one salient feature is their role. For instance, they are
used to display, guide, and mobilize co-targeted, non-ulterior con-attitudes.
Let us dub “WR judgments” those judgments that play the relevant role
(the “wrong role”). W judgments are token WR judgments: the tokens most
salient to English speakers. Prior to the semantic change posited, we had
terms at hand for W judgments; our ordinary metalevel phrases with moral
terms embedded (like “they think x is wrong”). We were not delivered terms
for the roles on a platter like that. Insofar as we were moved to acquire a term
for WR judgments as such, we might well have used those available meta-
level phrases as metonymic vehicles, implicitly targeting the role-playing
type. If so, calling something a “judgment about what is morally wrong”
would convey that it is a WR judgment, and conventionalization would have
led to it meaning that.8 All the while, moralizing uses of “wrong” could con-
tinue to mean W, just as one can use “Benedict Arnold” to refer to the unique
historical individual.
This account gains credibility from the fact that metamoralizing is an especially
important manifestation of our social intelligence, which long predates moral-
izing (moralizing exploits it). Plausibly, we are a self-domesticated species: given
the fitness impact of moralizing, selection for capacities and tendencies to
A striking example of a new category that received its name from one of its prototypi-
cal members was reported in a newspaper article about the war crimes trial of John
Demjanjuk. Demjanjuk had been accused of being “Ivan the Terrible,” a cruel and sadis-
tic prison guard at the Treblinka death camp [and] extradited to stand trial in Israel.
A conversation between a native Israeli and an American reporter reveals a typical
instance of dual reference (emphases added):
Israeli: “If he is a Demjanjuk, then he should be condemned to death.”
Reporter: “But he is Demjanjuk, his name is John Demjanjuk.”
Israeli: “I know his name is Demjanjuk, but I don’t know if he is a Demjanjuk.”
As the newspaper article pointed out, the term Demjanjuk was used in this conversation
in two ways: to refer to the person John Demjanjuk and also to refer to the category of
people that he exemplified, a Demjanjuk. “The name Demjanjuk has become a noun in
Israel, a word to identify an ordinary person capable of committing unspeakable acts.” . . .
As it turned out, John Demjanjuk was found not guilty and has since applied for permis-
sion to return to the United States. It would be quite apropos to assert, in this context,
that John Demjanjuk was not a Demjanjuk after all!
8. One could agree with everything about DHR except deny that conventionalization ever
took place: what DHR says is literal is actually conveyed pragmatically. Indeed this fits
nicely with debunking accounts of role-sensitive interpretations (Levy 2011; Merli 2002;
Sonderholm 2013; Strandberg 2011, 2012; Viggiano 2008; and see especially Copp 2007,
212–216, 241–243).
represent and navigate it are predictable, or at least intelligible, and we do seem

to have them.9 Indeed moral attitudes like guilt, shame, and a sense of moral duty
may be, at least in part, adaptations to moralizing. Adaptations to moralizing
may also include sociolinguistic tendencies that promote efficient and reliable
tracking of moral norms, and these might include or cause the exploitation of
available linguistic resources for metamoral uses, such as terms for salient tokens
of WR judgments. Moreover, since moralizing is episodic, metamoralizing often
occurs while concern for W is dormant. Nonmoralizing metamoralizers may
have little incentive to sustain a prior devotion of linguistic resources to W.
This is speculative of course. But it means that if we have evidence for real-
ism and metamoral functionalism, we may not have to decide between them.
That decision is forced only if moral terms are moral/metamoral monosemous.
If metamoral functionalism is correct, an interesting question is what roles
like WR really amount to. I cannot say much here, but I will report two assump-
tions relevant later. First, WR judgments will at least be widely correlated with
co-targeted, co-valenced non-ulterior attitudes of the characteristic moral sort
(disapproval, guilt, shame, aversion, outrage, contempt, disgust, etc.). And
these include de dicto attitudes, which, if realism is true, are explicitly about
W. Second, I assume that, at least among humans, if judgments have the feature
just mentioned (widely correlated with co-targeted, co-valenced non-ulterior
attitudes of the characteristic sort), then they are WR judgments.
RE A L I S M, C O NTE NT- I NDI V I DUATI O N ,

A N D H Y B R I D S E M AN TI CS
To say that moral judgments are content-individuated is to say that, for each
moral judgment type, there is an objectively real property such that being a
judgment of that type is being a judgment about that property and nothing more
than that. This seems a natural corollary of realism. As noted earlier, realism
holds that “wrong” relates semantically to W as “spherical” relates to spherical-
ity. And to judge that something is spherical is to judge that it has sphericality.
Of current relevance is not the content-individuation claim itself but two
modal implications of it, “content sufficiency” (CS) and “content necessity” (CN).
CS Necessarily, W judgments are wrong-judgments.

CN Necessarily, wrong-judgments are W judgments.
9. Bering 2011; Chudek and Henrich 2011; Chudek et al. 2013; Kitcher 2006; O’Gormana
et al. 2008.
It might seem puzzling why realism should entail CS and CN. For realism
as understood here is a contingent, empirical claim about the meaning of an
English word in certain uses, and CS and CN are modal claims about all pos-
sible judgments of certain types. How can the former entail the latter?
There are three ways facts about a word meaning can have implications
regarding a type of judgment: the judgments can be about the word, they can be
expressed with the word, or they can be characterized with the word. Obviously,
neither of the first two can underwrite the entailment of CS or CN by realism.
Few wrong-judgments are about the word “wrong,” and many are not expressed
with it either (e.g., those made in different languages). However, the third can
and does underwrite the entailment—at least in the absence of any hybrid view.
Let us dub “semantic monism” the claim that moral terms have only one
meaning where this rules out both synchronic and diachronic hybrid seman-
tics. Synchronically, monism means semantic “purity”: when used for moral-
izing, moral terms have only one kind of meaning. Diachronically, monism
means moral/metamoral monosemy: when used for metamoralizing, moral
terms have the same meaning they do when used for moralizing.
If semantic monism is true, then realism entails that “wrong” means W, and
nothing more, when used for moralizing or metamoralizing. And if that is true,
then both CS and CN are true. For substituting synonyms for synonyms (“W”
for “morally wrong”), both become equivalent to “necessarily, W judgments are
W judgments.”
We can now see how DHR breaks the entailment of content-individuation
by realism. For that entailment obtains only if the word defined by realism
(“wrong”) has the same meaning in CS and CN as it does when used in the
way to which the definition applies. And that is precisely what DHR denies. For
the definition applies to moralizing uses of “wrong,” and CS and CN involve
metamoral uses. In both the term “morally wrong” is used—“wrong-judgments”
is shorthand for “judgments about what is morally wrong”—but neither entails
a first-order claim about what is morally wrong. Rather, “morally wrong” is
used to characterize judgments as being of a certain type.
Turning to synchronic hybrid realism, its implications depend on the contri-
bution to metamoral meaning of a further, nondescriptive semantic feature. For
instance, suppose (as Boisvert suggests) “wrong” has both a descriptive mean-
ing (W) and an expressive meaning (it expresses a con-attitude to W things). If
so, then what does “judgment that x is wrong” mean?
The answer is under dispute.10 There are two main possibilities. According to
one, the expressive element remains expressive and adds an extra condition to
appropriate (sincere and literal) uses of the phrase. On this view the phrase means
10. Schroeder 2009; see also Boisvert, Copp and Schroeder in this volume.
something like “judgment that x is W; and down with W things.” (Compare the
effect of “damn” in “Jones thinks that the damn Republicans are just wonder-
ful.”) On the second possibility, the expressive element itself becomes ascribed,
adding a truth condition to the ascription. If so, then it means something like
“judgment that x is W that expresses a con-attitude to W things.”
If the first proposal is correct, then synchronic hybrid realism avoids neither
CS nor CN. For both would then mean “necessarily, W judgments are W judg-
ments; and down with W things.” And that is true, even if properly expressed
only by a disliker of W things. If the second proposal is correct, then CS is not
true. For it would then mean “necessarily, W judgments are W judgments that
express a con-attitude to W things,” and that is not so (unless W is necessarily
motivating). However, CN is still true even if the second proposal is correct. For
it would then mean “necessarily, W judgments that express a con-attitude to W
things are W judgments.”
Note that the failure to break the entailment of CN applies to any version of
synchronic hybrid realism, be it couched in terms of expressive or imperati-
val meaning, conventional implicature, presupposition, or whatnot. It does not
matter what we add to our account of the meaning of “wrong” so long as we
do not remove W: realism plus the assumption of moral/metamoral monosemy
entails CN. Thus the only way to affirm realism without thereby being committed
to CN is to adopt DHR.
In sum, DHR lets us combine realism with the rejection of both CS and CN.
Arguably, synchronic hybrid realism lets us combine realism with the rejection
of CS. But there is no way for realists to avoid CN without DHR, not even with
synchronic hybrid realism. So henceforth I will focus on what can be combined
with realism only if CN is dropped, for that is where DHR’s unquestionably
unique advantages lie.
RE A L I S M, F U N C TI O NALI SM , AND M O RA L TW INS

WI T H OU T S P E CI AL PR O PE R TI E S
I pointed out in the introduction that (intuitive) interpretations of amoralists

and moral twins underwrite functionalism. Competent folk take judgments as
wrong-judgments when and only when they are WR judgments (more or less),
and a straightforward vindicating explanation of that tendency is that they are
thereby manifesting their understanding of what it is to be a wrong-judgment.
Similarly, there could hardly be better evidence that money is a functional type
than that we do not take something to be money unless it plays the relevant
role, and if something plays the relevant role we take it to be money (even if
otherwise unlike our money, e.g., in its material constitution).
I should make explicit that I am thinking of the relevant roles as played by judg-
ment types, not token judgments. Like Michael Stocker, David Brink, and oth-
ers, I find it easy to suppose that individual amoralists can make moral judgment
despite lacking the typical attitudes. However, James Dreier, James Lenman, and
others seem right that our intuitions quite differ when it comes to entire isolated
communities of amoralists. If judgments of some type are never accompanied by
any co-targeted con-attitudes, then they do not seem to be wrong-judgments,
whatever else is true of them. This supports a kind of communal—or, better,
holistic—version of motivational internalism. (Money is like this too, as witness
bills used as bookmarks, to light cigars, or framed and hung on a wall.)11
As with content-individuation, what is crucial about functionalist
role-individuation are two modal corollaries, “role necessity” (RN) and “role
sufficiency” (RS).
RN Necessarily, wrong-judgments are WR judgments.

RS Necessarily, WR judgments are wrong-judgments.
RN and RS give us a straightforward vindicating explanation of negative and

positive role-sensitive interpretations, respectively. (Negative role-sensitive
interpretations involve role absences decreasing confidence in interpretations
as moral, positive role-sensitive interpretations involve role presences increas-
ing confidence.)
If semantic monism is presupposed, it is hard to see how realists can accom-
modate RN or RS without taking moral properties to be special in some way
(indulging myths, as I see it). Indeed realism, semantic monism, and RN entail
that W is necessarily motivating. Realism and semantic monism entail CS,
that necessarily W judgments are wrong-judgments. RN says that necessarily
wrong-judgments are WR judgments. Combining, we get that necessarily W
judgments are WR judgments: there cannot be judgments about W that do not
play the distinctive role WR. Thus one way to force a choice between antireal-
ism and myth—assuming there are no necessarily motivating properties—is to
assume semantic monism and then try to give a vindicating explanation of our
negative role-sensitive interpretations. So one way DHR lets us be nonmytholo-
gizing realists is by undercutting the entailment of CS by realism. But as we saw,
this virtue may not be unique to DHR; synchronic hybrid realism may share it.
So let us focus on positive role-sensitive interpretations.
A common worry about claims like RS is that they are too permissive. Many
would follow Philippa Foot, for instance, in denying that WR judgments that
11. For further discussion and references, see Tresan 2009.

target all and only finger snapping as such are wrong-judgments.12 Perhaps
their targets have to include harmful behaviors or certain paradigms like bru-
tality, betrayal, neglect, hogging, and shirking.
However, even if this worry is justified, positive role-sensitive interpreta-
tions still pose a challenge to realists. The most obvious form of the challenge
is that—even if there are nonrole conditions (like having certain targets) on
wrong-judgments as well as role conditions—the conditions sufficient for
wrong-judgments might not be enough to guarantee that the judgments that
satisfy them are all about the same objectively real property. If there are condi-
tions sufficient for judgments being wrong-judgments but insufficient to guar-
antee they are all about the same objectively real property, then CN is false.
DHR would then be realists’ only hope.
Plausibly our interpretations, however constrained by nonrole conditions,
do not support a guarantee that wrong-judgments are all about the same objec-
tively real property. Or they do so only if the property in question is special
in some way. If so, positive role-sensitive interpretations force us to choose
between antirealism and mythologizing (or DHR). But the matter is tricky for
three reasons. First, distinguishing role and nonrole conditions is not always
easy. The relevant roles themselves presumably impose some constraints that
we would not naturally think of as role constraints. For instance, at a minimum,
for judgments to play the role WR, they must target behaviors. They cannot be
about all and only red things, or chairs, or numbers between 8.3 and 9.7. And
what counts as a role anyway? Suppose we describe the role as opposition to W.
If that counts as a role, then RS may be true but compatible with CN. Second,
what makes it the case that two judgments are about the same property is a
vexed issue, since the metasemantic question of what makes judgments about
one property rather than another is vexed. Third, the relevant interpretative
tendencies are mixed. The classic example is R. M. Hare versus Foot. For Hare,
their being used to commend even such things as brutality and scalping can
make it apt to interpret judgments as good judgments. For Foot, as we have
seen, targets cannot be so bizarre.
So it is worth pointing out another, less problematic way positive role-sensitive
interpretations force a choice between antirealism and special properties. This
way does not require us to mention roles at all and does not depend on any
metasemantic assumptions. It depends on only two assumptions: that we are
justified in taking twin-wrong-judgments to be wrong-judgments and a truistic
epistemic principle.
The epistemic principle regards the sorts of evidence that can justify us in
taking two causally independent judgment types J1 and J2 to attribute the very
12. Foot 1958.

same property. Broadly speaking, there are two sorts: the basic standards13 guid-
ing J1 and J2 judgments and the nature of the candidate properties compat-
ible with the first sort of evidence. The most favorable evidence occurs when
basic standards are identical—by stipulation in the case of thought experiments,
manifestly (empirically) in real-life cases. At the other extreme, basic standards
might diverge so much that there is no chance of interpreting the judgments as
attributing the same property. For instance, if J1 judgments are made about all
and only red things (even if not round) and J2 judgments are made about all and
only round things (even if not red), it is hard to see what further evidence might
suggest that J1 and J2 judgments actually attribute the same property.
When the evidence regarding basic standards neither mandates nor excludes
the judgments attributing the same property, the second sort of evidence is rel-
evant. What matters is whether among the remaining candidates there is one that
is conspicuous: likely to be talked about. I say more about conspicuousness in the
next section. But to see the point, suppose you are interpreting a tribe. The mem-
bers have a term “squircle,” and they make squircle judgments about things that
are roughly circular; that is all you have been able to observe. You seem justified in
taking squircle judgments to be about circularity, even though our evidence from
targets does not mandate that (they could be about some roughly circular shape).
That is because circularity is so much more likely to be talked about as such than
any of the other candidates. Now suppose you visit another tribe, and the members
13. By the “basic standards guiding S’s wrong-judgments” I mean the set of properties F
such that
(a) F is an objectively real natural property,

(b) S is disposed such that if she or he believes that x is F then she or he judges that x is
wrong, and
(c) It is not the case that (b) is true only because there is a distinct property F* such that
(i) correspondents of (a) and (b) are true of F*, and
(ii) S is disposed such that if she or he believes that x is F then she believes that
x is F*.
For instance, the basic standards of a consistent utilitarian consist in the property of fail-
ing to maximize happiness, since (a) it is an objectively real natural property, (b) she or he
is disposed such that if she or he believes x fails to maximize happiness, she or he judges x
wrong, and (c) she or he does not judge failures to maximize happiness wrong, because she
or he judges them to have some other objectively real natural property, which makes her or
him judge actions wrong. For most of us our basic standards are much more messy than
this, consisting in multiple properties and changing over time. (Moreover, I have omitted the
parameters of confidence in beliefs and judgments and stringency of wrong-judgment. For
Jones a belief that x is F might elicit a weak judgment that x is slightly wrong, whereas for
Smith it might elicit certainty that x is very wrong. Ultimately these should be distinguished,
though it is not necessary here.) That our wrong-judgments are guided by basic standards
make umpersand judgments about figures roughly similar to ampersands (&).

The only difference is that the ends connect to form a third loop, and the figure is
rotated clockwise ninety degrees. Are you justified in taking them to be thereby
making ampersand judgments? No, because ampersands are no more likely to be
talked about than any of countless other similar shapes compatible with what you
have observed. They could attribute being an ampersand, but by the same token
they could attribute a property even less like the ampersand than it seems.
It follows that if the basic standards guiding them differ significantly and system-
atically and there is no conspicuous candidate, we are not justified in taking J1 and
J2 judgments to attribute the same property. And this brings out another way that,
barring DHR, positive role-sensitive interpretations pressure us to choose between
antirealism and special properties. In this case the special properties are special in
being conspicuous. The pressure is exerted by the fact that, on any version of the
moral twin earth thought experiment, it is stipulated there are significant, sys-
tematic differences between the basic standards guiding twin-wrong-judgments
and our wrong-judgments. (“Our wrong-judgments” = wrong-judgments made
with the English term “morally wrong”.) It follows that if we are justified in taking
wrong-judgments and twin-wrong-judgments to attribute the same property, that
property must be conspicuous. Thus realists who affirm CN must say that if we are
justified in taking twin-wrong-judgments to be wrong-judgments, W is conspicu-
ous. Since the only way to be realist without affirming CN is DHR, it gives us the
only way to avoid antirealism or special (conspicuous) properties.
Here is a fuller presentation of the argument:
1. If CN is true, then both our wrong-judgments and twin-wrong-

judgments are wrong-judgments only if they attribute the same
property (viz. W). (def. CN)
2. We are justified in taking causally independent judgments to attribute
the same property only if they manifestly converge14 or there is a
conspicuous candidate. (epistemic principle discussed above.)15
3. Our wrong-judgments and twin-wrong-judgments are causally
independent and do not manifestly converge. (stipulated)
in this sense is uncontroversial, even nonnaturalistic realists and antirealists grant it. For
ease, in the text I speak of “wrong-judgments about all and only F things” as shorthand for
wrong-judgments guided by basic standards consisting in F.
14. By “manifestly converge” I mean that identity of basic standards is either stipulated (in
thought experiments) or supported by empirical evidence (in real-life cases).
15. The premise does need qualification but not one that matters. Strictly speaking, one
might be justified in taking judgments to attribute the same property in virtue of being justi-
fied in either taking them to manifestly converge or positing a conspicuous candidate, even
so
4. If CN is true, we are justified in taking both our wrong-judgments

and twin-wrong-judgments to be wrong-judgments only if W is
conspicuous. (1–3)16
5. We are justified in taking both our wrong-judgments and
twin-wrong-judgments to be wrong-judgments. (part of
straightforward vindicating explanation of interpretative tendencies
of competent folk)
so,
6. If CN is true, W is conspicuous. (4, 5)

7. If realism is true, then CN is true (or DHR is true). (See the second
section.)
so,
8. Either antirealism is true, or W is conspicuous (or DHR is true). (6, 7)
Three final points. First, note that this argument relies on no metasemantic
assumptions about what makes judgments attribute one property rather than
another. It relies merely on a weak principle about the sorts of evidence that
might justify taking independent judgments to attribute the same property
if they do not and there is not. But it is not easy to see what could justify these false beliefs.
And avoiding the argument this way would provide only a hollow “vindication” of positive
role-sensitive interpretations. Any theorist who adopted it would have to take herself or him-
self to be in possession of evidence that defeats premise 5, at least for her or him.
16. Strictly speaking, two further assumptions are needed for 4 to follow from 1–3. The first
is that if being about a certain property is a necessary condition on being a P judgment, then
being justified in taking two judgments to be P judgments depends on being justified in tak-
ing them to be about the same property. That seems true at least where the latter clause is
read de re: where taking them to be about the same property might consist in taking one to
be about F and the other about G where in fact F = G. And premise 2 seems true even for
this de re sort of taking to attribute the same property: note that we are not justified in taking
umpersand and ampersand judgments to attribute the same property, even de re. (There is an
exception to this epistemic dependence assumption in the case of testimonial justification: if
a source known to be reliable tells you that J1 judgments and J2 judgments are both P judg-
ments, you might be justified in believing this independently of your justification for believ-
ing they are about the same property. But our justification for taking twin-wrong-judgments
to be wrong-judgments is not testimonial.) The second assumption is that if we are justified
(premise 2). This principle is compatible with any plausible metasemantic

theory.
Second, I noted earlier that our positive role-sensitive interpretative ten-
dencies are mixed (e.g., Foot vs. Hare). This too needs explanation, and DHR
gives us a powerful resource: it lets us posit different metamoral dialects (and
idiolects). We can thus give a straightforward vindicating explanation of the
nominally conflicting interpretations: in Foot’s mouth, “the anti–finger snap-
pers do not think finger snapping is morally wrong” is true, whereas in Hare’s
mouth “they think it is morally wrong” is true. Such variations fit nicely with
the account of the development of the polysemy sketched in the first section.
For semantic changes are not uniform in a population. Brand-name genericiza-
tion, for instance, occurs unevenly, with some parts of the linguistic commu-
nity still using as a brand name what others use as a type term or both (e.g., in
some parts of the United States “coke” means simply Coca-Cola, and in other
parts it is polysemous, meaning either that or soda pop). And of course the
variations need not simply involve different stages of the same process but the
acquisition of different polysemous meanings. DHR, then, can give us a vin-
dicating explanation not just of role-sensitive interpretations but of nominally
conflicting role-sensitive interpretations.17
Third, a worry that will have occurred to some readers is that DHR does
not really vindicate our interpretations of moral twins insofar as we interpret
them as disagreeing with us. For the vindication works by letting judgments
with different contents be wrong-judgments. And such judgments need not be
in disagreement.
This is a serious worry about DHR that unfortunately I cannot do justice to
here.18 I will make just four brief points. First, this worry is not specific to DHR
but is a well-known challenge to any view that rejects conspicuous moral proper-
ties. For without taking wrongness to be conspicuous, there is no way to defend
the idea that both our wrong-judgments and twin-wrong-judgments are about
the same property. Without that, the intuition that the judgments are in disagree-
ment stands in need of explanation. And the explanations cited by other views
in taking our wrong-judgments and twin-wrong-judgments to be about the same property

because there is a conspicuous candidate, then that conspicuous candidate is W (the prop-
erty our wrong-judgments attribute). But this just follows from the fact that a conspicuous
candidate justifies taking judgments to attribute the same property by justifying our taking
the judgments to attribute it.
17. See Francén Olinder 2010, 2012 for a much more thorough explanation.
18. I say more about it in “Naturalistic Moral Realism and Motivational Internalism: From
Negative to Positive” (forthcoming).
are available to proponents of DHR as well. For instance, as many have noted, our
wrong-judgments and twin-wrong-judgments disagree in attitude, which might
account for at least some of the intuition of disagreement. Proponents of DHR
can say this as well. Second, synchronic hybrid realism fares no better when it
comes to vindicating the intuition of disagreement. For synchronic hybrid real-
ism does not help us see how twin-wrong-judgments and our wrong-judgments
might attribute the same property. Third, recent empirical work suggests that phi-
losophers may overestimate the intuition that there is disagreement in moral twin
cases.19 Fourth, DHR does have a resource for explaining the intuition of disagree-
ment that other views lack: the moral/metamoral polysemy insofar as we do not
recognize it. In confidently (and correctly) interpreting twin-wrong-judgments
as wrong-judgments, we may take ourselves to have justifiably interpreted them
as making W judgments. If so, we will suppose that they disagree.
R E A L I S M W I TH O UT HUM ANLY SPE CIA L P ROP ER TIES
What are realists unavoidably committed to? Realism entails that when English
speakers moralize, they think and talk about objectively real moral properties
like W. That is already a robust commitment that one might well doubt, not
least because of variations in the basic standards guiding wrong-judgments
among English speakers. I have nothing to contribute about that problem here,
and DHR is not helpful to that project, so far as I can see.20
Basically, I think realists ought to stop there. At least they should not posit
a certain further type of thing, “humanly special properties.” The best way to
explain what I mean by this is to explain why realists who reject DHR are com-
mitted to them. To be precise, they are committed given a plausible background
assumption, one that often goes without saying in metaethics. This assump-
tion—which I shall call “moralizing is human”—has two components.
The first is that wrong-judging is human: species typical, not just an idio-
syncrasy of some particular human cultural tradition. In this sense sports are
human, but baseball is cultural: cultures independently come to play sports but
not baseball. This first part of the assumption, then, is that wrong-judging is not
distinctive of some cultural tradition—not something just English speakers do
or even just those influenced by Western civilization—but a human tendency.
(And likewise for other thin moral judgment types.)
19. I have in mind especially Sarkissian et al. 2011.

20. For a helpful discussion, see the contribution to this volume by Laura Schroeter and
François Schroeter and Schroeter and Schroeter 2009.
The assumption that wrong-judging is human already forces realists to a

significant commitment if moral terms are moral/metamoral monosemous.
As we saw in the second section, realism and the monosemy assumption entail
CN, that wrong-judgments are always W judgments. If so and if humans are
wrong-judgers, then humans are W judgers: there is a species-typical ten-
dency to think and talk about the very property we talk about when we mor-
alize with “wrong.”
That is already a way for W to be “humanly special,” but there is more.
For the second component of the “moralizing is human” assumption is that
human wrong-judgments always play the role WR. If RN is true, this is a nec-
essary truth, and the “human” bit is superfluous (as with “human bachelors
are unmarried”). But even if we suspend judgment about that, it is empirically
obvious that wrong-judgments among humans are WR judgments.21 Although
philosophical thought experiments have featured entire communities of amor-
alists who make judgments that might be wrong-judgments, no actual human
community has ever come close to that. Insofar as anyone actually makes
wrong-judgments, these are bound up with the relevant attitudes and practices.
Monosemous realism plus the assumption that moralizing is human entails
not just that humans think and talk about W but that they care about it too.
And indeed that they care about it in a striking way: widely, negatively, not
derivatively of ulterior concerns, and to some extent as such. Monosemous
realism plus “moralizing is human,” then, entail that W is in this strong sense
humanly special. Let us use the phrase “take interest in” to abbreviate “think,
talk, and care about, including widely, negatively, nonderivatively, and to some
extent as such,” so that we can put the claim that W is humanly special this
way: humans take interest in W.
I think we ought to reject humanly special moral properties: deny that
humans take interest in W or any other moral properties. If we reject them,
then—as with the rejection of metaphysically special properties—we have two
choices: reject realism or take real moral properties to be not humanly special.
In the rest of this section I say why I think humanly special moral properties are
problematic and what realism without humanly special properties looks like.
One way of evaluating whether humans take interest in W would be to iden-
tify which property W is and then investigate whether humans take interest in
it. But that method would be both difficult and unsatisfying, because it can only
be applied if realism is presupposed. A better method is to identify a broader
set of properties to which W must belong if it exists (i.e., if realism is true) and
21. I am here relying on the second assumption about WR mentioned at the end of the first
section.
then see whether humans take interest in any properties in that set. If not, we
can deny that W is humanly special without any commitment to realism.
Doing this requires identifying a set to which we know W must belong,
and we can achieve this by relying on uncontroversial first-order moral intu-
itions. For instance, we know redness is not W, because an absurdity would
follow: necessarily actions are wrong insofar as they are red. We can radically
narrow the candidates for being W in this way. We know that if W exists it
is a behavioral property and not a shape, color, size, location, sound, physical
object, and so forth. And we know it is not just any behavioral property; for
example, it is not finger snapping, dancing the samba, or playing baseball. W
is either identical to or necessarily coinstantiated with a property cited as the
wrong maker by some normative theory that does not seem utterly absurd to us
(or would not if brought to our attention). Let us dub “wrongish” any property
F such that it is even remotely plausible—not utterly absurd—that necessarily
x is wrong insofar as x is F.22 Unless we suffer pathological doubts about our
reliability as moral judges, we can be confident that W is wrongish if it exists.
This means that W is humanly special only if there is some humanly special
wrongish property (HSW).
HSW There is a wrongish property humans take interest in.
HSW can be investigated empirically, without any presupposition of realism.

The set of wrongish properties is huge, probably infinitely large. But that does
not rule out empirical discoveries about our relations to all of them. Here is an
analogy. I am thinking of a number between 83.1708 and 83.1743; it could be
any number of digits, including infinitely many. Since there is an infinite num-
ber of possibilities, I will give you a hint: the one I am thinking of is the one we
talk about far more than the others, the one that is distinctively phenomeno-
logically salient to us, and the one we denote with “pi.” Those hints are useless of
course, because you know we do not bear those relations to any numbers in that
set. And you know that even though there are infinitely many such numbers.
Similarly, we can—at least in principle—learn about our relations to all wrong-
ish properties without knowing which one W is or if it even exists.
Another way of seeing that HSW is empirical is to note that antirealists are
free to accept it. An expressivist, for instance, could hold that humans take inter-
est in failing to maximize happiness: they think, talk, and care about it (widely,
negatively, nonderivatively, and as such). Since failing to maximize happiness
is wrongish, this would be a way of affirming HSW. But such an expressivist
would still deny that the failure to maximize happiness is wrongness. She or
22. F has to be instantiated, since we are looking for a candidate for the realist property W.
he will say that it is merely the universal target of the con-attitudes constitu-
tive of wrong-judging. (Error-theoretic projectivists and contextualists can say
similar things.)
Two plausible claims together suggest that HSW is unjustified. The first is
descriptive relativism; the second an inconspicuousness claim. Descriptive rela-
tivism has it that, empirically, independent human cultures tend to be moral
twins with respect to each other: their WR judgments are guided by signifi-
cantly, systematically different basic standards.23 The inconspicuousness claim is
that no wrongish property is likely to be talked about by human communities.
A scope distinction is important here. It may be that any given human commu-
nity is likely to talk about some or other wrongish property. What the inconspic-
uousness thesis denies is that there is some wrongish property such that humans
are likely to talk about it. I will say more about inconspicuousness, but first a
word about why descriptive relativism and inconspicuousness undercut HSW.
They do so, because, as we saw in the last section, there are two ways of
supporting the claim that independent judgments are about the same prop-
erty: they can have the same basic standards or there can be a conspicuous
property among the candidates compatible with the first sort of evidence.
Descriptive relativism and inconspicuousness rule these out, respectively.
Together they deprive us of any reason we might have for positing an HSW
property. We do not seem to take interest in the same wrongish property, and
there is no particular reason to expect that we would, thus we have no good
reason to believe that we do.
Descriptive relativism is a familiar, plainly empirical issue; I need not say
anything more about it here.24 But I will say a bit about “inconspicuousness.”
As I am understanding it, conspicuousness is always relative to subjects: some-
thing can be conspicuous to some and not to others. Something is conspicuous
to subjects insofar as they are likely to talk about it as such. In the case at hand
the relevant subjects are human communities. The question is whether there is
a wrongish property such that it is likely that human communities would inde-
pendently come to talk about it. Inconspicuousness is the claim that there is not.
If realism is true, then there is a wrongish property English speakers and those
influenced by them talk about: W. But if inconspicuousness is true, it would be
very surprising to find other cultures independently coming to talk about W. (It
would not be impossible, just a huge coincidence, so inconspicuousness does
not entail descriptive relativism, nor does it rule out HSW on its own.)
23. And there is no known case of independent cultures arriving at WR judgments with
identical basic standards.
24. For a powerful defense, see Prinz 2007.
Inconspicuousness is a useful claim to focus on for two reasons. First, as just

noted, if descriptive relativism is true, then whether HSW is justified depends
entirely on it. Second, conspicuousness is never a brute property. It cannot be that
something is likely to be talked about as such but that likelihood has no explana-
tion in its other relations to those likely to talk about it. Conspicuousness depends
on other, empirically rich relations, such as being perceptually salient or causally
potent. The case for inconspicuousness is thus negative: we find no evidence of
any such relations. Rather, for each wrongish property there seem to be countless
similar properties—similar metaphysically, extensionally, phenomenologically,
causally, and functionally. Since a conspicuous wrongish property would have to
stand out in at least one of these ways, there is no conspicuous wrongish property.
Three significant assumptions make this claim especially plausible. The first
is an affirmation of normative-theoretic open-mindedness: countless normative
theories, varying along multiple continua, cannot be ruled out as absurd. For
instance, all the well-known variations of consequentialist and deontologi-
cal theories.25 The second is a permissive position on properties, the relevant
upshot of which is that for every such normative theory there is at least one dis-
tinct wrongish property—namely, the natural “wrong-making” property that, if
that theory is correct, W supervenes on or is.26 The third is reductive naturalism
and in particular the assumption that those “wrong makers” are all the wrong-
ish properties there are. There is not, necessarily coinstantiated with failing to
maximize happiness or any other natural wrongish property, a metaphysically
sui generis property—that would itself be wrongish, as necessarily coinstanti-
ated properties are necessarily identically wrongish.27
Together these suggest that for every wrongish property there are count-
less properties that are similar metaphysically (viz., natural), extensionally
25. For instance, take utilitarianism: x is wrong insofar as x fails to maximize happiness.
Consider all the possible variations: “subjective” aspects can be added (maximize expected
happiness? reasonably expected? [with variations on what counts as reasonable]); versions
of happiness can vary (pleasure? preference satisfaction? objective list? [and what is on the
list?], etc.); there can be alternatives to maximizing (Millian variations in the weight assigned
to different types of happiness, options to fail to maximize happiness to avoid great personal
loss [and so possible supererogatoriness], satisficing, etc.); minor nonutilitarian elements
can be added (e.g., equality, justice, or rights tip the balance in case of a tie for maximizing
happiness); rule (vs. act) elements can be added (with variations in how rules are evaluated);
there are more; and any or all of these possibilities can be added at once, with differing
weights assigned.
26. Error-theoretical normative theories—like that x is wrong insofar as God hates x—are
excepted, even if we are happy being open-minded even about them (see note 22).
27. This third claim has it that wrongish properties are like “samba-ish” and “cricket-ish”
properties: those that epistemically (at least for me) might be necessarily coinstantiated with
dancing the samba or playing cricket, respectively. I know roughly what they involve but not
(differing slightly on continuous dimensions),28 phenomenologically, causally,

and functionally. To be precise, they are phenomenologically and causally simi-
lar except in psyche- or culture-bound ways. For instance, W has causal powers
no other wrongish property has in virtue of English speakers and those influ-
enced by them taking interest in it. But of course culture-bound distinctiveness
cannot underwrite conspicuousness to humans. Likewise, W may be phenom-
enologically salient to us; for example, if we project our attitudes onto it. To oth-
ers a different wrongish property might be salient. Again, such culture-bound
salience cannot explain a human tendency to take interest in W.
For a wrongish property to be conspicuous functionally is for it to be likely
talked about, because talking about it is or was especially useful or adaptive.
Realists often tell functional stories the upshot of which is that it makes sense
for there to be humanly special properties.29 Because of the popularity of this
move, I will offer four warnings about relying on any such line of reasoning to
counter the case against HSW.
First, functional stories are always weaker than empirical data. A compelling
evolutionary case can be given that there is no homosexuality, but it is too weak
to make us doubt the manifest existence of homosexuality. If some feature is
observable but conflicts with our best functional stories, we should look for bet-
ter stories, not doubt our eyes. Thus there are severe limits to a function-based
account of conspicuousness. A functional story is better suited to supplement
another account than as a stand-alone case for a conspicuous wrongish property.
Second, if there are no wrongish properties that are conspicuous in any other
way, there are unlikely to be any that are functionally conspicuous. F might be
more functional to care about than G, because F is more easily noticeable than
G or because the behavioral upshot and impact of caring about F is quite differ-
ent than that of caring about G. But if F and G are similar phenomenologically
and causally, they will be similarly noticeable. And if they are similar extension-
ally and causally, then the behavioral upshot and impact of caring about them
will be similar. Thus functional differences seem to require other differences.
A case for functional conspicuousness alone is not likely to do the job.
Third, it can be easy to ignore the difference between showing that domains are
functional to relate in a certain way and showing that specific properties are. To show
the former is to show that, for some domain of properties, it is functional to relate to
the details, not even where the indeterminacies lie. Nevertheless, there are no necessarily
coinstantiated but metaphysically distinct samba- or cricket-ish properties, of that I am sure.
28. Including from zero to a little bit, as when we add to utilitarianism that equality of distri-
bution breaks ties in total amounts of happiness.
29. E.g., Boyd 1988; Brink 1989; Casebeer 2003; Railton 1986.
some or other member of D in a certain way. It is easier to give a domain-functional

account than to show that there is a uniquely functional property. But a defense of a
functionally conspicuous wrongish property would have to do that.
Fourth, flexibility in which properties we take interest in may be useful or
adaptive (in the latter case it yields phenotypic plasticity). Humans do not all
face precisely the same opportunities and challenges, and this sort of flexibility
allows us to develop relevantly useful or adaptive attitudes and not waste psy-
chological resources caring about what is irrelevant. Consider the needs of a
tribe living amid great abundance but also near many competing, warlike tribes
as against those of a tribe mostly free of competition but in a harsh environ-
ment. The kinds of “wrongish” behaviors especially maladaptive in such tribes
might differ, for example, insufficient martial courage or in-group loyalty in
the first, insufficient industry and compassion in the second. Insofar as moral
con-attitudes are designed to suppress such maladaptive behaviors, the adap-
tive triggers in these communities would vary. Flexibility would allow for this,
whereas a fixed human tendency to acquire the same con-attitudinal profile
would not. Thus a challenge to any functional defense of conspicuousness is
not just that we might not know the relevant function or that the functional
pressures might be too weak to canalize a unique human profile but that an
intelligible function might actually speak against conspicuousness.
I have argued that DHR lets realists accommodate the inconspicuousness of
wrongish properties. Their inconspicuousness also helps DHR in another way.
As noted in the first section, proponents of DHR must explain why—even grant-
ing the need for terms for WR judgments—we use moral terms polysemously
rather than other terms. My answer was that moral terms are well placed to be
used metonymously, as salient tokens of a salient type. Inconspicuousness helps
makes this plausible, because it removes a source of resistance to metamoral-
izers exploiting moral terms for content-liberated functional uses. Why forego
linguistic resources for metamoralizing to ensure we can track a property that is
just one among countless similar properties? The desire for terms for WR judg-
ments might override hesitation about losing terms for W judgments, espe-
cially since we can have both by disambiguating.30
Descriptive relativism and inconspicuousness vindicate the common sen-
timentalist idea that morality can be modeled on such things as tastiness and
sexiness. In these cases we find variable interest in otherwise undifferentiated
30. Descriptive relativism can also help DHR with this reply. For the more variety among the
contents of WR judgments we encounter, the more valuable are terms for WR judgments
and the less valuable terms for W judgments. Similarly, the more variety of tissues around,
the greater the need for a term for “tissue” and the less need for a term for a particular brand
like Kleenex.
domains. The specific line between what any individual or culture likes and
does not like in these ways does not reflect any significant metaphysical, exten-
sional, phenomenological, causal, or functional difference in the things liked
or not. Those who like certain flavors will find them significant because they
are likable, and the rest of us will find them significant in that they are liked.
But other than that they are just like other flavors. The only other potential dif-
ference of note is that they may be within a domain such that there is a human
tendency to like some or other flavors in that domain. If there are no humanly
special moral properties, then that is how it is with moral interest in behaviors.
This may suggest antirealism, because antirealism is true of these other dis-
courses. Our standard terms of evaluation in these other fields (“tasty,” “sexy”)
do not seem to have a descriptive content exhausted by an objectively real prop-
erty. But that may just mean the cases are not alike in every respect. So far as
realism goes, a better model for moral terms are terms for units of length like
“inch,” “foot,” “mile,” and so forth. Here too we find variable interest in an oth-
erwise undifferentiated domain. There is no significant metaphysical, exten-
sional, phenomenological, causal, or functional difference between, say, an inch
and 1.0001 inches or any two precise lengths. The only significant difference
between them is that we talk about one rather than the other. But realism is true
here: these terms have a descriptive content exhausted by an objectively real
property. (Of course there are differences between moral and length terms.)31
CO N C L U S I ON
DHR is the only way to have realism without CN. So it is the only way to com-
bine realism with robust vindicating explanations of our positive role-sensitive
interpretations that cite RS. Indeed it is the only way to combine realism with any
account of those interpretative tendencies that posits conditions that (i) suffice
for wrong-judgments but (ii) do not guarantee that all possible judgments that
satisfy them are about the same real property. It is also the only way to vindicate
31. Here are seven. First, moral terms are about behavioral properties, length terms about
lengths. Second, length terms pick out properties that vary along a single dimension, whereas
moral terms pick out properties that vary along multiple dimensions (in this way more like
terms for dances, sports, or games than lengths). Third, length terms get their meanings in a
manner much closer to sheer stipulation than moral terms. Fourth, moral terms presumably
have a significantly greater degree of indeterminacy. Fifth, our concern about the properties
picked out by length terms is entirely ulterior. Sixth, our concern about moral properties
tends to be uni-valenced—we dislike wrongness—whereas our concern about lengths var-
ies: sometimes being a foot is positive, sometimes negative, sometimes neutral. Seventh, the
practical needs served by the terms are very different.
the minimal assumption that we are justified in taking twin-wrong-judgments

to be wrong-judgments without positing conspicuous properties. Finally, it is
the only way to combine realism with the rejection of humanly special prop-
erties, at least if we assume that moralizing—including wrong-judging—is a
human tendency and not a cultural idiosyncrasy.
My aim has just been to show that DHR lets us have all of these claims, not
that we should want any of them. But note that all the claims DHR uniquely
reconciles with realism are popular. They are popular among antirealists, but
that is to be expected given that DHR has not been taken seriously—even dis-
cussed—and it is the only way to reconcile them with realism. And note that
antirealists do not accept these claims grudgingly, as an undesirable commit-
ment given their antirealism. Rather, antirealism is attractive partly because it
makes it possible to believe them.
If DHR is defensible, it can liberate metaethicists from defending unneces-
sary and problematic positions. It can liberate demythologizers from defend-
ing antirealism, freeing them up to adopt a more objective account of moral
semantics. It can liberate realists from defending conspicuous, humanly spe-
cial properties. Both positions would thus be transformed, but realism much
more so. For antirealists, adopting DHR would amount to adopting a some-
what more objective account of moral semantics than they had thought avail-
able. For realists, rejecting conspicuous, humanly special properties would
amount to a significant thinning out of their worldview. I suspect and hope
that the demythologization of conspicuous, humanly special properties is just
a matter of time. If so and if wrong-judging really is human, the fate of realism
rests with DHR.32
References
Bering, J. 2011. The Belief Instinct: The Psychology of Souls, Destiny, and the Meaning of
Life. New York: Norton.
Boisvert, D. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89
(2): 169–203.
Boyd, R. 1988. “How to Be a Moral Realist.” In Essays on Moral Realism, ed. G.
Sayre-McCord, 181–228. Ithaca, NY: Cornell University Press.
Brink, D. 1989. Moral Realism and the Foundations of Ethics. Cambridge: Cambridge
University Press.
32. For helpful feedback and discussion, thanks to Dean Pettit, Daniel Boisvert, Geoffrey
Sayre-McCord, David Copp, Gene Witmer, Iskra Fileva, Kirk Ludwig, William Lycan,
Catherine Marcinkiewcz, James Evans, Gunnar Björnsson, Laura Schroeter, François
Schroeter, other participants at the 2012 Edinburgh conference on Hybrid Theories in
Metaethics, and especially Greg Littmann, Guy Fletcher, and Michael Ridge.
Casebeer, W. 2003. Natural Ethical Facts: Evolution, Connectionism, and Moral Cognition.

Cambridge, MA: MIT Press.
Chudek, M., and J. Henrich. 2011. “Culture-Gene Coevolution, Norm-Psychology, and
the Emergence of Human Prosociality.” Trends in Cognitive Science 15 (5): 218–226.
Chudek, M., W. Zhao, and J. Henrich. 2013. “Culture-Gene Coevolution, Large-scale
Cooperation, and the Shaping of Human Social Psychology.” In Cooperation and Its
Evolution, ed. K. Sterelny, R. Joyce, B. Calcott, and B. Fraser, 425–458. Cambridge,
MA: MIT Press.
Copp, D. 2001. “Realist-Expressivism: A Neglected Option for Moral Realism.” Social
———. 2007. Morality in a Natural World: Selected Essays in Metaethics.
Cambridge: Cambridge University Press.
Falkum, I. L. 2011. “The Semantics and Pragmatics of Polysemy: A Relevance-Theoretic
Account.” PhD diss., University College of London. http://www.google.com/url?sa=t
&rct=j&q=&esrc=s&source=web&cd=4&cad=rja&ved=0CEgQFjAD&url=http%3A
%2F%2Fdiscovery.ucl.ac.uk%2F1139079%2F1%2F1139079.pdf&ei=4udhUvneKIrI
yAHG0IGoCw&usg=AFQjCNENBPzISwGiPsTBy8oFx0aFNQ_IBw&sig2=7nIKkN
wJOpy4895rYccMZw&bvm=bv.54934254,d.aWc.
Foot, P. 1958. “Moral Beliefs.” Proceedings of the Aristotelian Society (1958–59) 59:83–104.
Francén Olinder, R. 2010. “Moral Motivation Pluralism.” Journal of Ethics 14
(2): 117–148.
———. 2012. “Moral and Metaethical Pluralism: Unity in Variation.” Southern Journal of
Philosophy 50 (4): 583–601.
Glucksberg, S. 2001. Understanding Figurative Language: From Metaphor to Idioms.
Kitcher, P. 2006. “Between Fragile Altruism and Morality: Evolution and the Emergence
of Normative Guidance.” In Evolutionary Ethics and Contemporary Biology, ed. G.
Boniolo and G. De Anna, 159–177. Cambridge: Cambridge University Press.
Koskela, A. 2011. “Metonymy, Category Broadening and Narrowing, and
Vertical Polysemy.” In Defining Metonymy in Cognitive Linguistics: Towards a
Consensus View, ed. R. Benczes, A. Barcelona, and F. Mendoze Ibáñez, 125–146.
Amsterdam: Benjamins.
Levy, N. 2011. “Moore on Twin Earth.” Erkenntnis 75 (1): 137–146.
Merli, D. 2002. “Return to Moral Twin Earth.” Canadian Journal of Philosophy 32
(2): 207–240.
O’Gorman, R., D. Wilson, and R. Miller. 2008. “An Evolved Cognitive Bias for Social
Norms.” Evolution and Human Behavior 29 (2):71–78.
Prinz, J. 2007. The Emotional Construction of Morals. Oxford: Oxford University Press.
Railton, P. 1986. “Moral Realism.” Philosophical Review 95 (2):163–207.
Sarkissian, H., J. Park, D. Tien, J. Wright, and J. Knobe. 2011. “Folk Moral Relativism.”
Mind and Language 26 (4): 482–505.
Schroeder, M. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119 (2): 257–309.
Schroeter, L., and F. Schroeter. 2009. “A Third Way in Metaethics.” Noûs 43 (1): 1–30.
Sonderholm, J. 2013. “Unreliable Intuitions: A New Reply to the Moral Twin-Earth
Argument.” Theoria 79 (1): 76–88.
Strandberg, C. 2011. “The Pragmatics of Moral Motivation.” Journal of Ethics 15

(4): 341–369.
———. 2012. “A Dual Aspect Account of Moral Language.” Philosophy and
Phenomenological Research 84 (1): 87–122.
Tresan, J. 2009. “The Challenge of Communal Internalism.” Journal of Value Inquiry 43
(2): 179–199.
———. Forthcoming. “Naturalistic Moral Realism and Motivational Internalism: From
Negative to Positive.” In Motivational Internalism, ed. G. Björnsson, C. Strandberg, R.
Francén Olinder, J. Eriksson, and F. Björklund. New York: Oxford University Press.
Viggiano, A. 2008. “Ethical Naturalism and Moral Twin Earth.” Ethical Theory and
Moral Practice 11 (2): 213–224.
6
The Pragmatics of Normative

Disagreement
S T E P H E N F I N L AY n
On their surface, moral and normative sentences appear descriptive. But what
kinds of facts could they describe? Many have been drawn to the idea that nor-
mative words like ‘ought’ and ‘good’ are used to refer to relational properties,
consisting in a relation to something like a standard, rule, desire, or end, which
can vary between contexts. On my preferred view, to say that s ought to φ is
roughly to assert the proposition that some implicit end, e, is more likely to
obtain if s φs than if s does anything else instead.
But relativistic metaethical theories face a familiar, seemingly fatal problem
in accounting for the intuitive extent of normative disagreement (and agree-
ment). They imply that superficially incompatible moral or normative claims
made by speakers who are concerned with different ends (etc.) express logically
consistent propositions. Given the natural idea that disagreement involves logi-
cal inconsistency, these theories seem to imply that utterances of sentences like
‘s ought to φ’ and ‘s ought not to φ’ do not really express disagreements even
when they intuitively do.
In particular, moral disagreements evidently occur between people who are
not united by preference for any common end; call such disagreements funda-
mental. It is widely thought that accommodating these disagreements requires
instead some kind of absolutist semantics: either a form of nonrelational
descriptivism identifying a common property of concern to disagreeing parties
or a form of expressivism identifying some kind of disagreement in attitudes
other than belief, like preferences.
I will first explore a different source of disagreement problems. There
are compelling reasons for thinking that normative propositions are
The Pragmatics of Normative Disagreement 125
information relative, which generates structurally identical problems about

disagreement and shows that normative disagreement is everybody’s prob-
lem. I will argue that any adequate account must appeal to pragmatics and
that the end-relational theory (and so any relevantly similar view) provides a
“quasi-expressivist” solution in terms of the conflicting preferences that are
pragmatically expressed by asserting logically consistent normative proposi-
tions—a solution that extends straightforwardly from information relativity
to end relativity.
So far this recapitulates my moves in a previous paper cowritten with Gunnar
Björnsson (Björnsson and Finlay 2010). But whereas we there offered a prag-
matic solution only for a specific kind of case involving a deliberating agent and
a better-informed adviser, here I will advance a general theory of normative
disagreement. I will also argue that despite being a purely descriptivist seman-
tics, the end-relational theory actually provides a better account of disagree-
ment in attitude than expressivism, making it redundant to posit expressive
conventions in the meanings of normative words. Contrary to the received
wisdom, intuitions of normative disagreement support a relational semantics
over either expressivism or nonrelational descriptivism. For reasons of space,
however, I will not be able to discuss related issues about how normative claims
are reported and evaluated as “true” or “false.”1
T H E P R OB L E M O F DI SAG R E E M E NT FROM INFORMATION

R E L AT I V I T Y
An initial, naive formulation of the natural idea that disagreement requires

inconsistent propositions is
Inconsistent Belief: B’s assertion/belief that p disagrees (agrees) with A’s

assertion/belief that q iff p and q are logically inconsistent (equivalent).
If one of B’s assertions or beliefs disagrees with one of A’s, then by extension B
himself can be said to disagree with A herself in that respect. This can be said in
a familiar sense even where there is not any interaction between them, though
talk about “disagreement” may sometimes suggest it.
Consider the following scenario:2
1. A longer version of this paper appears as chap. 8 of Finlay 2014.

2. This is a variant of the miner case in Kolodny and MacFarlane 2010; Regan 1980. I intro-
duce Two Gun Roulette to address the problem relative to a single end, which avoids compli-
cations for the end-relational theory addressed in Finlay 2014, chap. 6.
Two Gun Roulette. Angie has been kidnapped by Sadie, who tells her that
Angie will be tortured to death unless she plays a special game of Russian
roulette. She must choose between two six-chambered revolvers, L to her
left, and R to her right. L has been loaded randomly with one round, R
with three rounds. Angie must aim her chosen revolver at her head and
pull the trigger once. If she survives she will be released unharmed. She
has no other realistic options.
Evidently, Angie could appropriately conclude her deliberations by saying,
Angie: ‘I ought to use L, not R.’ (Deliberation)
The end-relational semantics (ERT) accommodates this as assertion of the

true proposition that she is more likely relative to her information to survive
unharmed if she uses L than if she uses R. Now let us supplement the scenario:
Advice: Bertie is a fellow captive who has all of Angie’s information plus
the following: since the revolvers were loaded, five captives have played
Sadie’s game. (In this version, the cylinders are spun once upon loading
but not between each use.) All five chose L and survived. Bertie has one
opportunity to whisper briefly in her ear but not enough time to share this
information.
The implication of Bertie’s information is that the next use of L will be lethal.
Evidently he would speak appropriately by saying,
Bertie: ‘You ought to use R, not L.’ (Advice)
ERT accommodates this as assertion of the true proposition that Angie is more
likely relative to Bertie’s information to survive unharmed if she uses R than if
she uses L.
To accommodate these claims, ERT apparently must interpret them as rela-
tivized to different information. Consequently, Angie and Bertie assert logically
consistent propositions. It further seems they need not have any inconsistent
beliefs. Bertie accepts the proposition assigned as the content of Angie’s claim,
and Angie does not believe anything inconsistent with the content assigned to
Bertie’s claim. By Inconsistent Belief they therefore do not disagree, yet intui-
tively Bertie does disagree with Angie’s statement. (It is not so intuitive that
Angie disagrees with Bertie; we will return to this asymmetry below.) He seems
able to say appropriately, ‘No, you ought to use R, not L,’ and while ERT suggests
that Angie asserted a proposition that Bertie may know to be true, for him to
say ‘Yes, that’s right. You ought to use L, not R’ would clearly be perverse and
deceptive. Here we confront a puzzle: Angie and Bertie seem to assert propo-
sitions whose truth conditions are determined wholly by their own differing
information and are therefore logically consistent, but Bertie seems thereby to
disagree with Angie.
This puzzle arises for any theory that relativizes normative propositions to
different information. One response simply denies that these normative claims
really are information relative or interprets both as relativized to something
like all the facts.3 On this view, Angie asserts a proposition concerning what
she objectively ought to do, which is simply false though she may be justified in
believing it, and Bertie is therefore correct to reject it. But this analysis cannot
be correct, because Angie may know it is quite likely she objectively ought to
use R. She can appropriately say ‘I ought to use L, although it’s quite likely that
given all the facts I ought to use R,’ but she cannot say ‘Given all the facts I ought
to use L, although it’s quite likely that given all the facts I ought to use R,’ since
accepting that p is likely false is incompatible with warranted assertion of it.
Angie thereby cannot be using ‘ought’ in an objective sense.4 Similar reasoning
shows that Bertie also cannot be using ‘ought’ objectively: he knows that using
R is quite likely also to be fatal, which would also make it false that Angie objec-
tively ought to use R. Making sense of these claims requires recognizing their
relativity to the speakers’ incomplete information.
A second response is that Angie’s and Bertie’s statements are relative to the
same incomplete information. Deliberating agents presumably do not aim to
make decisions on the basis of merely the information in their possession when
they begin deliberating but on roughly the fullest information they can utilize at
the time of decision, which will include information others are in a position to
make available to them. Whereas Bertie is unable to share his information with
Angie, information can be “available” for use without being possessed if others
communicate the probability of p relative to it. Angie’s use of ‘ought’ can there-
fore be expected to be relativized to information selected this way rather than
merely to her own information: call this news sensitivity.5 Her statement would
then be false because sensitive to Bertie’s additional information, although
perhaps still justified since she lacks reason to suspect this information exists.
3. Judith Jarvis Thomson (2008, 195) holds that all ‘ought’ claims are sensitive to (“objective”)
probability relative to the information available to humans at the time. Two Gun Roulette is
inspired by her scenario.
4. This is clearer where the speaker knows one of the other options is objectively best, like the
miner case; see Kolodny and MacFarlane 2010.
5. Janice Dowell (2013) argues that this is a sufficient solution.
Bertie’s statement would directly contradict hers, accounting for disagreement

by Inconsistent Belief. But while news sensitivity seems plausible for cases of
advice like Bertie’s, it is not a sufficiently general solution. If Bertie were, rather,
a long-distance eavesdropper, intuitively he could still disagree with Angie,
though his information would not be available to her.6 To simplify discussion
I will proceed under the assumption that news sensitivity does not account for
disagreement in the original case either.
A third response rejects these ordinary intuitions as mistaken: Bertie does
not genuinely disagree with Angie, because their claims are consistent. The
intuitions result from failing to appreciate that the claims are implicitly rela-
tivized to different information, as Bertie does himself. If intuitions of nor-
mative disagreement are systematically unreliable, then the entire problem
could be dismissed as based on faulty data. But this proposal fails to account
for the subtlety of ordinary intuitions of normative disagreement. Not just
any two speakers asserting superficially incompatible normative sentences
are taken to be disagreeing—such claims are often recognized as talking past
each other—and we will see that these intuitions exhibit a high level of sensi-
tivity to information relativity. (I take no stand on what should be classified
as “genuine” disagreement; the task is, rather, to explain the intuitive sense
of disagreement.)
These cases of disagreement from diverging backgrounds thus pose a gen-
eral puzzle for metaethics—or rather for philosophy in general, since analogous
problems are observed for many words that encourage relativistic treatments,
such as language about possibility (‘might’), knowledge (‘knows that’), taste
(‘disgusting’), and gradable adjectives (‘tall’). For example:
Angie: ‘I might survive if I use L.’
Bertie: ‘No, you cannot survive if you use L. But you might survive if you use R.’
Sadie: ‘No, you’re both wrong. Angie cannot survive no matter which she
uses.’7
A fourth response to these problems is relativism about truth.8 The idea is
that claims made relative to some k do not assert propositions about relations
to k but absolutist propositions that only have truth values relative to perspec-
tives characterized by indexes of k’s kind. So normative claims are not informa-
tion relative in content but instead are true or false only relative to particular
6. See also other cases below, particularly that Connie can appropriately either agree (post-
mortem) or disagree (hindsight) with Angie.
7. Suppose Sadie knows the next use of either will be lethal.
8. E.g., Kolodny and MacFarlane 2010; Kölbel 2002; MacFarlane 2007.
information perspectives. Angie and Bertie would then both address the same
simple proposition that Angie ought to use L and not R, which is not true or
false simpliciter but only relative to different perspectives. Angie correctly
asserts it, as it is true relative to her information. Bertie correctly rejects it,
as it is false relative to his. Because Bertie rejects the same proposition Angie
asserts, truth relativism secures the result that they satisfy Inconsistent Belief.
But this solution is not as straightforward as advertised. Truth relativism needs
to appeal to pragmatic resources that are also sufficient for a solution on our
contextualist semantics.
D I S A G R E E M E NT AS PR AG M ATI C
Truth relativism’s claim to account for the sense that Bertie disagrees with Angie
is based on satisfying Inconsistent Belief. But this is only sufficient for disagree-
ment given a traditional, nonrelativist view of truth. Consider:
Postmortem: Connie is a detective investigating after the fact. As it hap-

pened, Angie did not hear Bertie’s whisper and played using L, shooting
herself in the head. By an incredible fluke, the bullet passed through her
skull nonfatally, leaving Angie with a severe injury. Connie has inter-
viewed Bertie and learned all his information. She visits Angie in the hos-
pital weeks later.
On truth relativism, the proposition Angie asserted by saying ‘I ought to use

L, not R’ is false relative to Connie’s information, which is relevantly similar to
Bertie’s. Connie knows this but can agree with Angie’s claim, unlike Bertie, with
what I’ll call (figuratively) a postmortem claim:
Connie: ‘You were right. You ought to have used L, not R. In light of what
you knew it was the only sensible action.’ (Postmortem)
This case behaves as we would naively expect if contextualism rather than

truth relativism were correct. As contextualism also predicts, Connie would not
seem to be disagreeing with Bertie. She could alternatively say, ‘Bertie was right,
you ought to have used R’.
Consider also:
Less Informed: Debbie is Angie’s friend, captured with her. Debbie has less
information than Angie, not knowing how many rounds were loaded into
L or R, though she knows Angie knows.
On truth relativism, Angie’s claim that she ought to use L is false relative
to Debbie’s inferior information. Debbie knows this but does not intuitively
disagree with Angie’s claim.9 It would obviously be inappropriate for her to
say: ‘No, you’re wrong. Neither option is any better than the other.’ She can,
rather, say,
Debbie: ‘I don’t know whether you ought to use L or R.’ (Less Informed)
This identifies a general problem for truth relativism. In its least ambitious
incarnations, truth relativism holds that truth is relative to a time of assessment,
which it therefore excludes from the propositions. But consider:
Déjà Vu: One week after surviving Sadie’s game, Angie has the terrible
luck to be kidnapped by Sadie again, who compels her to play the game
with the same revolvers but reversed, so R holds the single round. Eddie is
another captive who knows the next use of R will be lethal, so he asserts,
‘You ought to use L, not R.’
According to this basic version of truth relativism, Eddie accepts as true (rel-
ative to his own temporal perspective) the same proposition Angie asserted
and rejects as false the proposition Bertie asserted. He thereby satisfies the
Inconsistent Belief criterion for agreeing with Angie’s claim and disagreeing
with Bertie’s. But intuitively he does not seem to be talking about the same
thing, as a contextualist treatment of time relativity would predict.
These cases illustrate that asserting superficially incompatible sentences like
‘s ought to φ’ and ‘s ought not to φ’, is not sufficient to trigger intuitions of
disagreement. Contextualist theories provide a simple and conservative expla-
nation: the same normative sentence is used to express different propositions.
But truth relativists are obliged to conclude that satisfying Inconsistent Belief is
not sufficient for intuitively disagreeing and therefore concede the story to be
more complex and to involve further conditions. While contextualist theories
are still in trouble provided inconsistent beliefs are necessary for disagreement,
I will argue that the missing conditions are sufficient by themselves and without
inconsistent belief.
These intuitions of disagreement are evidently responding to differences
between the contexts of Advice, Postmortem, and Less Informed rather than
9. Perhaps Debbie can infer from Angie’s claim that Angie is more likely to survive if she uses
L, which makes Angie’s claim true relative to Debbie’s information too. But suppose Debbie
also hears Bertie speak without knowing which is better informed: still she intuitively would
not disagree with Angie.
simply the sentences used. Two kinds of variability can be ruled out. First,
while the speaker’s information may vary between contexts, Bertie and Connie
possess relevantly identical information, yet Connie but not Bertie can appro-
priately agree with Angie. Second, while the speaker’s temporal relation to the
original claim may also vary (Bertie speaks prospectively, Connie retrospec-
tively), notice that Sadie can felicitously agree prospectively with Angie though
occupying the same temporal and informational perspective as Bertie:
Sadie: ‘Yes, you’re right. You ought to use L, not R.’ (Pseudoadvice)
Since Sadie does not want Angie to survive unharmed, she is not interested
in making her fuller information i available for Angie by telling her what she
ought, given i, to do. But she can still sincerely express agreement with Angie
or Connie about what ought to be done relative to Angie’s inferior information.
The relevant factor is evidently the conversational ends of the speakers, as truth
relativists themselves acknowledge: Mark Richard (2011) argues that whether
an assertion that p disagrees with an assertion that not-p depends on its point,
and Niko Kolodny and John MacFarlane (2010) maintain that a speaker like
Bertie must be disagreeing with the original claim, because he is offering advice,
which by nature aims to help the agent solve the practical problem she deliber-
ates over. In other words, the sense of disagreement here depends on recognizing
that Bertie speaks cooperatively, making it sensitive to pragmatic and not merely
semantic cues.
This role of pragmatics in generating disagreement intuitions can also be
demonstrated by observing a flaw in our naive formulation of Inconsistent Belief.
Speakers can evidently disagree without asserting contradictory propositions.
Suppose A says, ‘We ought to go stargazing at 8 pm’, and B replies, ‘Daylight sav-
ings started today.’ B is naturally interpreted as disagreeing with A, though the
propositions asserted are not inconsistent. The disagreement lies not in what is
said but in what is pragmatically expressed. Roughly, in saying that p, B disagrees
with A by pragmatically expressing the belief that not-q by virtue of what p entails
in combination with the common ground. In this example p (daylight savings
started today) implies not-q (it is not true they ought to go stargazing at 8 pm)
on the assumptions that they ought to go stargazing only after dark and that the
sky is not dark at 8 pm during daylight savings. This suggests a revised principle:
Inconsistent Belief*: B’s assertion that p disagrees (agrees) with A’s assertion

that q iff B thereby expresses the belief that not-q (q).
The objection is that relational semantics cannot accommodate normative

disagreements between advisers and agents, because the cooperative nature of
advice means that it must be intended to help the agent reach the right answer
to the question she asks in her deliberations about what she ought to do. This
cannot be a question about what the agent ought to do relative to her own infor-
mation, for which she neither needs nor receives any help, but it also cannot be
about what she ought to do relative to the adviser’s information, since this was
not her concern in deliberating. I will argue that ERT answers this objection
when we apply a single basic principle of pragmatics: that we expect people to
speak in the way they consider best for their conversational ends.10 Where con-
versational ends are shared, this implies a version of H. P. Grice’s cooperative
principle, but we will also be interested in other kinds of context.
INF OR MAT I O N R E LATI V I TY I N CO NTE XTS

OF D I R E C T I N TE R E ST
Begin with the pragmatics of agents’ claims, like Angie’s. This is a context of
deliberation, so we must identify the motivations of deliberating agents. The
objection assumes this to be the end of discovering which proposition of the
form I ought to φ is true, but ERT suggests a simpler, more natural answer: agents
deliberate with the aim of achieving their particular motivating ends. Angie’s
ultimate goal in deliberating is simply that she survives unharmed.
Knowing the truth of an ‘ought’ proposition relativized to an end is of special
instrumental interest to an agent in what I will call a context of direct inter-
est, where the agent is actively motivated toward the end. Knowledge of what
is most likely to achieve an end is an ideal basis for making a decision aimed
at achieving that end. But this depends on the quality of the information on
which this probability is based. An ‘ought’ proposition relativized to the fullest
available information will provide the instrumentally best available basis for a
decision; that is, one that most increases the probability of the end relative to
the available information. ERT therefore implies that an agent has no better
means from her own point of view for promoting her end through delibera-
tion than by identifying what she ought to do relative to the fullest information
available. So Angie would conclude her deliberation by judging what she ought
to do relative to the information available to her, in order to survive unharmed.
Now consider the pragmatics of Bertie’s utterance. This is a context of advice
where at least here11 the speaker is expected to cooperatively borrow the
agent’s motivating end with the aim of guiding the agent toward it. Bertie also
therefore speaks in a context of direct interest in Angie surviving unharmed.
10. See Finlay 2014, chap. 5, for discussion.

11. Advice sometimes guides agents toward ends they would prefer if better informed.
Because he is better informed, he knows what she ought to do in order to

survive unharmed relative to fuller information. This truth is of greater instru-
mental value relative to their shared end than the truth she asserts, and so
Bertie best promotes their shared conversational end by communicating the
proposition relativized to his information.
Since Angie’s fundamental interest is in surviving unharmed, and not in
knowing the truth of the proposition she asserted, this is the most cooperative
thing Bertie can do. He thereby makes his information available for her to act
on; because it provides a better basis for a decision aimed at Angie surviving
unharmed,12 she would prefer to act on the basis of this proposition rather than
the one she asserted. So ERT accommodates the intuition that Bertie’s advice is
cooperative, even though he addresses a different proposition. It predicts that
(i) this is the proposition he would assert; (ii) he would not waste his or Angie’s
time by addressing the proposition she asserted, which his information makes
moot; and also (iii) this is what Angie would expect him to do.
It is also now easy to explain why Connie, Debbie, and Sadie are not similarly
expected to deny that Angie ought to use L, although it is false that she ought
to use L relative to their respective information. Unlike Bertie, they would not
thereby promote their conversational ends, though for different reasons. While
the less informed Debbie has direct interest in Angie’s end of surviving unharmed,
denying the proposition relativized to her lesser information would not promote
it in any way. On the contrary, because of the expectation of cooperation, Debbie
would thereby pragmatically indicate she has more information than Angie, which
is likely to cause Angie to make a subjectively worse decision. By contrast, Connie
and Sadie each speak in contexts of indirect interest, addressing a proposition rela-
tivized to an end toward which they are not actively motivated. Connie speaks in a
postmortem context, where achieving Angie’s end is no longer at issue, while Sadie
speaks in a noncooperative context of pseudoadvice, hoping that Angie’s end does
not obtain. They therefore are not pragmatically obliged to address the proposition
relative to the fullest available information, unlike Bertie and Debbie.
This explains the appropriateness of the ‘ought’ claims in the various scenarios
but does not yet explain intuitions of disagreement (e.g., that Bertie disagrees
with Angie). But it provides the resources for a solution. In contexts of direct
interest in e, asserting that in order that e it ought to be that p (relative to the
fullest available information) will pragmatically express a preference that p.13
12. The proposition that relative to fuller information i(+) e is most likely given p itself makes
it true that e is most likely given p relative to the lesser information i(–). A subjectively better
basis for a decision does not necessarily result in an objectively better outcome of course.
13. For discussion, see Finlay 2014, chap. 5.
Angie’s utterance therefore expresses her preference that she uses L, not R, while
Bertie’s utterance expresses his preference that Angie uses R, not L. By asserting
consistent propositions, they pragmatically express a common kind of attitude
toward inconsistent propositions.
Could an expressed conflict of preferences be sufficient to trigger intuitions of
normative disagreement? Expressivists since Charles L. Stevenson (1944) have
labeled this kind of conflict a “disagreement in attitude,” in contrast to a “disagree-
ment in belief,” and have used it to explain normative disagreement. It promises
to accommodate the natural idea that disagreement involves inconsistency in
attitudes but without satisfying Inconsistent Belief, which motivates the objection.
Whereas purely descriptivist semantics are generally assumed to be committed
to explaining disagreement through some version of Inconsistent Belief,14 ERT
suggests, rather, the quasi-expressivist solution that normative disagreement can
occur through pragmatic expression of inconsistent attitudes of a different kind.15
INF OR MAT I O N R E LATI V I TY I N CO NTE XTS

OF I N D I R E C T I NTE R E ST
So far we only have an account of normative disagreement in one kind of con-

text. A complete solution requires a general theory accommodating disagree-
ment intuitions across a full range of contexts. If some normative disagreements
consist in pragmatically expressed disagreement in preference, then the sim-
plest remaining theory is that all are; that is:
Inconsistent Preference: B’s assertion that p normatively disagrees (agrees)

with A’s assertion that q iff B thereby expresses a preference inconsistent
with (equivalent to) the preference A thereby expresses.
As ERT maintains that normative utterances express beliefs in addition to any

preferences, it may seem strange if no normative disagreements consisted solely
in inconsistent beliefs; expressivists may complain that the descriptive seman-
tics is a boondoggle doing no significant work.
Another possibility is that disagreement intuitions are sensitive to inconsis-
tency either in belief or in preference—which threatens to be messy and might
seem unnecessarily complex compared to expressivism. To avoid objections of
being ad hoc, a mixed account must take the form of a general principle with
14. E.g., Horgan and Timmons 1992.

15. See also Harman (1996) and Denis Robinson (2009), who calls this “quasi-disagreement.”
testable predictions. While the theoretical space here is large, we will not need
to look far beyond the simplest, disjunctive version:
Inconsistent Belief or Preference: B’s assertion that p normatively dis-

agrees (agrees) with A’s assertion that q iff B thereby expresses either (i) an
end-relational belief inconsistent with (equivalent to) an end-relational
belief A thereby expresses or (ii) a preference inconsistent with (equivalent
to) the preference A thereby expresses.
On first glance contexts of indirect interest favor a mixed account. Our

quasi-expressivist treatment of the disagreement in preference between Angie
(Deliberation) and Bertie (Advice) appealed to two contingent features of contexts
of direct interest: the speaker aiming at the salient end and relativizing to the fullest
information available. Normative speech and thought in which either of these fea-
tures is absent, like Connie’s (Postmortem) and Sadie’s (Pseudoadvice), will not in
the same way express preferences on ERT or thereby disagreement in preference.
But Connie and Sadie both seem to agree with Angie when they say, ‘You ought to
use/have used L, not R’, which may seem to rule out Inconsistent Preference.
These results are independently plausible: in her postmortem context,
Connie presumably prefers rather that Angie did not use L, since she would
prefer Angie had been released unharmed. Because Sadie shares Bertie’s infor-
mation, she may in fact prefer that Angie uses L, but her utterance will not
express this, since her audience has no reason to interpret it as relativized to her
full information, given that she is offering uncooperative pseudoadvice. Their
expressed agreement with Angie therefore could not consist in indicating that
they share Angie’s preference for her using L and not R.
By contrast, Inconsistent Belief or Preference accommodates these as agree-
ments in belief. Connie and Sadie both know that Angie asserted a true propo-
sition relativized to her own information, and we can straightforwardly explain
the pragmatic appropriateness of agreeing with it in their contexts of indirect
interest. Since Angie surviving unharmed is not their conversational end, they
are not pragmatically expected to address a proposition relativized to the fullest
information available.
These speakers are interested instead in what deliberative conclusion Angie
should reach in her epistemic situation. Connie’s end might be to comfort
Angie by reassuring her that she could not reasonably be expected to have done
anything differently. Sadie might find perverse amusement in “helping” Angie
make a good decision within her epistemic limitations. In these contexts evalu-
ating the information-relative proposition Angie asserted is directly relevant
to the speakers’ goals, so Connie and Sadie are each naturally interpreted as
expressing agreement in belief.
Other disagreements in contexts of indirect interest are problematic for

Inconsistent Belief or Preference as well as for Inconsistent Preference, however. So
far we have found speakers in contexts of direct interest disagreeing in preference
and speakers in contexts of indirect interest disagreeing in belief. But sometimes
a speaker in a context of indirect interest intuitively disagrees with a normative
claim when our semantics predicts that no inconsistency in belief is expressed.
One illustration is what I will call a hindsight context, where a speaker addresses
retrospectively what ought to have been done in light of her present information.
Consider Connie again, who can not only appropriately agree postmortem
with Angie’s earlier claim but in the very same circumstances could appropri-
ately say instead,
Connie: ‘It turns out you were mistaken. You ought to have used R, not
L.’ (Hindsight)
(Indeed Angie herself may appropriately say at this later time: ‘In hindsight
I was mistaken. As I now know, I ought to have used R and not L.’ Intuitively,
she thereby expresses disagreement with her earlier deliberative statement.) But
according to our theory, Connie knows that the belief Angie earlier expressed
(that she ought relative to her information to use L rather than R) was true, so
these hindsight claims do not express disagreement in belief. Given that the
theory also does not predict that these claims express any preference, these dis-
agreement intuitions are unaccommodated.
The problem can also be observed for agreement from a variation on the
hindsight context:
Hindsight Agreement: Freddie is a detective who investigates the revolv-

ers after Angie survived her injury from using L, discovering that use of
R would have been fatal. Freddie is misinformed on two counts: he was
told that L was loaded with a blank so that Angie survived her ordeal
unharmed and also that Sadie lied to Angie that L held the three rounds
and R only held one. He therefore believes that Angie was badly confused
about her own information when she said she ought to use L.
In this context Freddie can appropriately say in hindsight,
Freddie: ‘Angie was right; she ought to have used L, not R. Though not
for the reason she thought!’ (Hindsight Agreement)
Freddie seems to have agreed with Angie’s statement. But by ERT he must believe
that the proposition she asserted is false, so this cannot be agreement in belief. If
we predict that he also fails to express any preference, then this agreement intu-
ition is also unaccommodated.
The assumption that normative speech does not express preferences in con-
texts of indirect interest should be reconsidered. Preferences might be expressed
in some such contexts even if our previous explanation does not apply. Indeed
the hindsight claims made by Connie and Freddie do plausibly express prefer-
ences. Regardless of their conversational ends, both can be presumed to prefer
that Angie survived unharmed. Since these hindsight claims are relativized to
the fullest information available to the speaker, they concern what was optimal
from the speaker’s present point of view for achieving an end she prefers.
This solution is not sufficiently general, however, because preference for the
salient end is contingent; for example, Sadie can also appropriately disagree in
hindsight with Angie. As she is dragged off in handcuffs she might say, ‘You
were wrong. You ought to have used R’, perhaps gloating over how she manipu-
lated Angie into harming herself. On our theory this cannot be disagreement in
belief as for the previous cases of hindsight, but since Sadie transparently pre-
fers that Angie did not survive unharmed, her statement also does not indicate
she prefers that Angie used R; if anything it indicates the contrary preference.
This case of alienated hindsight therefore still resists both Inconsistent
Preference and the disjunctive principle. Additionally, Connie’s and Sadie’s
hindsight claims seem to express the same kind of disagreement with Angie’s
deliberative claim, which suggests that this is not the right solution also for the
cases where the speaker is known to prefer the salient end. In general terms,
pragmatic expression of preferences seems to depend on end-relational claims
being relativized both to an end the speaker prefers and to the fullest infor-
mation available to the speaker, but intuitively normative disagreements arise
between speakers with different preferences and information.
AT T I T U D E S W I TH CO NDI TI O NAL C ONTENT
If some intuitive cases of normative disagreement cannot be explained as dis-

agreements in either belief or attitude, this poses a problem for any theory. The
prospects for a solution appealing entirely to inconsistent beliefs look hope-
less. But expressivists offer a way to account for these problematic cases as
disagreements in attitude. Allan Gibbard proposes that the relevant attitudes
are not directed toward options simpliciter (e.g., that Angie uses R and not L)
but toward conditionals, or options under particular conditions.16 A normative
utterance of ‘s ought to φ’ expresses the speaker’s attitude toward φ-ing on the
16. Gibbard 2003, 69.

condition of being in s’s situation or of being s herself; that is, toward the proposi-
tion expressed by ‘I φ if I am s’. Sadie’s disagreement with Angie might then be
accounted for as follows: Sadie’s hindsight statement that Angie ought to have
used R and not L expresses her attitude toward using R and not L if she herself
were Angie/in Angie’s situation C. Angie’s deliberative statement that she ought
to use L and not R expresses the same attitude toward using L and not R if in C.
This approach derives some intuitive appeal from its similarity to the follow-
ing kind of exchange, which is naturally interpreted as a case of disagreement:
Angie: ‘If I were s, I’d φ.’

Bertie: ‘Well if I were s, I’d ψ.’
It might also seem easily extended to disagreements or agreements in attitude

expressed in the contexts Advice, Postmortem, and Hindsight, which would sup-
port some version of the simpler principle Inconsistent Preference and suggest
that expressivism offers a simpler, better account of normative disagreement
than any descriptivist semantics.
This strategy requires a different kind of attitude, as preferences toward condi-
tionals generate the wrong results. Connie’s postmortem claim that Angie ought
to have used L cannot express the preference that if in Angie’s situation she uses
L, since she knows that using L is worst relative to her preferred end of surviving
unharmed.17 Gibbard therefore appeals instead to attitudes of planning or inten-
tion, which avoids this problem. Connie’s utterance more plausibly expresses her
plan to use L and not R if in Angie’s situation. Since plans are designed to be imple-
mentable in the salient situation, Connie can only make plans for what to do if in
Angie’s situation that would make sense to her if she were indeed in Angie’s situ-
ation—including being limited to Angie’s lesser information. I will temporarily
switch to talking about plans, though we will see this introduces new problems.
Another question concerns how to understand the idea of the speaker
“being s.” Taken literally, worries arise about the coherence of this. Is Sadie
being Angie not a metaphysical impossibility, and if she were Angie, would
she not do exactly the same thing as Angie? These problems can be avoided by
understanding the strategy in terms of being in s’s situation: there is nothing
absurd about Sadie contemplating herself being “in Angie’s shoes.” But this
move problematizes the claim of inconsistency in speakers’ attitudes.
One perhaps minor problem is that different speakers’ uses of ‘I’ refer to dif-
ferent people. The propositional content of Sadie’s thought, I use R and not L,
17. It does not help to say that the relevant preference is that if she were in a situation subjec-
tively like Angie’s then she uses L rather than R, since Angie is in such a situation and Connie
prefers that she uses R in it.
if I am in C, would therefore seem to be that Sadie uses L and not R if Sadie is
in C. But Angie expresses no attitude toward that proposition, and so we are
some distance from Inconsistent Preference. Accordingly, Gibbard does not ana-
lyze disagreement in attitude by appeal to logically inconsistent contents and,
rather, tries to motivate an intuition that some other kind of incompatibility
can arise between different subjects’ plans. While it is unclear whether there is
any such intuitive notion, one way of trying to identify inconsistent contents is
by appeal to de se propositions such that every person who thinks ‘If I am in C,
I φ’ entertains the same proposition. While de se propositions are controversial,
I will assume them here for the sake of argument.
A different issue about inconsistency is fatal for this strategy, however.
Being “in s’s situation” needs clarification in two respects. Does it imply hav-
ing s’s preference for ends? I will assume it does, as necessary for accommo-
dating Sadie’s disagreement in plan with Angie.18 More importantly, does
being in s’s situation imply having only s’s information? To be able to account
for all cases of normative disagreement as consisting in an expressed dis-
agreement in plan, being “in s’s situation” must sometimes involve being in
s’s epistemic or subjective situation and sometimes involve being in s’s objec-
tive situation but with different information. Connie’s postmortem state-
ment that Angie ought to have used L and not R, for example, must express
the plan to use L and not R if in C with the same information as Angie. But
Bertie’s advice that Angie ought to use R and not L must instead express the
plan to use R and not L if in C but with his own fuller information; similarly
with hindsight claims.
This scuttles any hope of explaining problematic cases of disagreement by
appeal to attitudes toward inconsistent conditionals, since what I do if I have
information i1 is not logically related to what I do if I have information i2. While
this may have some welcome implications (e.g., that Connie’s hindsight judg-
ment that Angie ought to have used R does not disagree with her postmortem
judgment that Angie ought to have used L), it also yields results incompatible
with intuitions about some basic cases of normative disagreement.
In particular, it implies that better informed advisers do not genuinely disagree
with the agents they advise; for example, Bertie does not genuinely disagree with
Angie. Bertie must express the plan to use R if in Angie’s situation but with his own
fuller information iB. But Angie cannot have expressed the conflicting plan to use
L in that very situation. Rather, she must have expressed the plan to use L in her
own actual epistemic situation with information iA. The contents of these plans are
logically consistent. Indeed these plans could not be incompatible in any plausible
18. This poses a problem for fundamental disagreement; see Finlay 2014, chap. 8.
sense, since a prudentially virtuous agent could and arguably should have both
plans simultaneously. But Bertie’s disagreement with Angie was the basic case
that motivated abandoning Inconsistent Belief for some version of Inconsistent
Preference in the first place. So even on its own terms the appeal to attitudes with
conditional content fails to accommodate intuitions of disagreement.19
CO N D I T I ON A L S W I TH ATTI TU DI NAL C ONTENT
The cases still lacking a solution involve hindsight or better informed disagree-
ment in a context of indirect interest, especially when the speaker does not
prefer the end (e.g., if Sadie says, ‘Angie ought to have used R’). Normative
claims in these contexts do not plausibly express actual preferences or similar
attitudes, toward either simple or conditional contents. ERT predicts this, but it
also predicts a subtly different kind of conditional attitude solution: that these
normative claims pragmatically indicate conditionals with attitudinal content.
We have observed that the theory predicts that somebody who asserts
ought(p) in a context of direct interest, relativized to her preferred end e and
present information, will express the preference that p. So anybody who asserts
ought(p) relativized just to her present information will pragmatically indicate
the following conditional: if her preferred end were e then she would prefer that
p. This is a counterfactual anchored in the normative belief she expressed, con-
cerning only world states where she still has that same belief and information.20
So Sadie’s hindsight statement will pragmatically indicate that if she were to
share the preference that Angie survived unharmed then she would prefer that
Angie used R. The same will be true of hindsight statements where the end is
preferred, like Connie’s and Freddie’s. I will call these hypothetical preferences in
respect of their kinship to “hypothetical imperatives” and to distinguish them
from the previous kind of “conditional attitudes.”
Might some intuitions about normative disagreement be responsive to prag-
matic expression of hypothetical preferences? A first test is to see whether
19. Gibbard indeed maintains (conversation) that there is no genuine normative disagree-
ment from unequal information, dismissing these intuitions as mistaken. He distinguishes
between the ‘ought’ of rationality and the ‘ought’ of advisability (see also Ridge 2014); how-
ever, we have also observed the need for multiple ‘ought’s of advisability.
20. Familiar worries about counterfactuals might be raised, e.g., that in the closest possi-
bilities the speaker might be “irrational” so her preference for e is not transmitted to p. The
qualification ‘if rational’ is therefore suggested, though I believe unnecessary; see Finlay
2008, 2009, 2014.
directly asserting the suggested information has the same effect. Suppose the
conversation had gone like this:
Angie: ‘I prefer to use L.’

Sadie: ‘If I were to prefer your ends, I would prefer that you use R.’
It seems natural to say that Sadie would thereby have expressed some kind
of disagreement with Angie. This is also very close to the colloquial vehicle for
normative disagreement, ‘If I were you I’d φ’. Since ‘I would φ’ is tantamount to
a description of a counterfactual preference, these utterances draw very close,
given that ‘If I were you.. ’ can be read as if I were in your objective situation
with your ends. Whereas expressivists sometimes invoke this turn of speech to
support their appeal to attitudes with conditional content, here the conditional
seems to take wide scope over the attitude.
Appeal to hypothetical preferences avoids the problems observed for the
expressivist’s appeal to attitudes with conditional content. First, since these
conditional preferences have simple and unconditionalized contents, disagree-
ing claims relativized to different information (like Angie’s and Sadie’s but also
Bertie’s advice, Connie’s hindsight, etc.) thereby express hypothetical attitudes
toward logically inconsistent contents (e.g., that A uses L and not R and that A
uses R and not L), so no unexplained kind of inconsistency is required. Second,
the relevant attitudes can be identified simply as preferences about the same
proposition rather than as plans or intentions, so we do not have to counte-
nance de se propositions or make sense of (e.g.) Sadie being Angie.
This also avoids a number of baroque consequences that Gibbard’s approach
seems unable to evade except by positing either ambiguity or noncomposi-
tionality. We can make sense of claims that s ought to φ involving remote
circumstances (e.g., ‘Caesar ought not to have crossed the Rubicon’) or nona-
gents (e.g., ‘Trees ought to have deep roots’) without saying that the speaker
must have contingency plans for being in those circumstances (deliberating
whether to cross the Rubicon with your legions to seize Rome, being a tree).
There is nothing bizarre about indicating that if you were to prefer the salient
end (e.g., that the Roman Republic was preserved, that trees grow strong),
then you would prefer that p (e.g., that Caesar did not cross the Rubicon, that
trees have deep roots).
It may seem counterintuitive that an actual disagreement could be explained
by merely hypothetical attitudes. This may also seem to predict an absurd pro-
liferation of normative disagreements. Since there are indefinitely many true
counterfactuals about what somebody would prefer under some condition or
other, if hypothetically preferring p under one set of conditions C1 were suf-
ficient for disagreeing with somebody who hypothetically prefers not-p under
different conditions C2, would not everybody at every moment both agree and
disagree with every possible normative claim?
A first response is that these hypothetical preferences and disagreements are
grounded in actual expressed beliefs, although those beliefs may be logically
consistent. They might, rather, be described as disagreements in actual disposi-
tions, grounded in beliefs. But another necessary condition for disagreement
can be adduced. Intuitively, disagreement with an attitude requires some kind
of robustness such that a disagreeing attitude resists the other.21 These prefer-
ences differ in robustness according to the completeness of the information on
which they are based. Angie’s preference to use L is provisional on no fuller
information being available, for example, so it yields without resistance to
Bertie’s preference based on fuller information. Bertie’s preference “trumps”
hers and makes it moot.
Our pragmatic principle therefore predicts that disagreement in hypo-
thetical preference will be asymmetrical on this basis, which I will call Robust
Inconsistent Hypothetical Preference (RIHP):
RIHP: B’s assertion that p normatively disagrees (agrees) with A’s assertion
that q if B thereby indicates that if he were to prefer e he would on the
basis of his information iB have a preference inconsistent with (equivalent
to) the preference A thereby indicates she would have on the basis of her
information iA if she were to prefer e, and iB≥iA.
This complication is actually serendipitous. I have described normative dis-

agreements asymmetrically, because many are intuitively asymmetrical: Bertie
intuitively disagrees with Angie, for example, but Angie does not intuitively
disagree with Bertie. This is observed but not explained by others,22 and it
poses a difficulty for any theory proposing to explain normative disagreement
by Inconsistent Belief, which implies symmetry.23 Of course a satisfactory the-
ory must also predict correct asymmetries, but ERT’s predictions are promis-
ing: Bertie’s advice disagrees with Angie’s deliberative claim but not vice versa;
Connie’s hindsight claim disagrees with Angie’s claim but not vice versa and
agrees with Bertie’s advice and vice versa. These results support our pragmatic
21. Stevenson (1944, 3–4) suggests a similar but more problematic condition: “At least one of
[the disagreeing parties] has a motive for altering or calling into question the attitude of the
other.” Disagreement does not seem to entail any motive to change others’ attitudes.
22. Dietz 2008; Ross and Schroeder 2013.
23. An alternative hypothesis is that disagreement requires responding to the other claim.
But cases are easily found where an earlier claim intuitively disagrees with a later one; cf.
Cappelen and Hawthorne 2009; Richard 2011.
account of normative disagreement and also undermine the uncharitable

hypothesis that intuitions of disagreement are simply due to blindness about
the information relativity of normative claims.
RIHP is only formulated as a sufficient condition for normative disagree-
ment, however, and at least two obstacles prevent it from being also a neces-
sary condition and thereby a general theory: postmortem claims and negations.
The former consists in normative claims in contexts of indirect interest that are
relativized to information the speaker considers inferior to her own (whether
incomplete or false): for example, Connie’s postmortem claim that Angie
ought to have used L and Sadie’s pseudoadvisory claim that Angie ought to use
L. These will not express hypothetical preferences as defined, since in the near-
est possibilities where the speaker prefers the end they still will not be relativ-
ized to the speaker’s full information. ERT easily accommodates the agreement
expressed by these claims as agreement in belief, however, since these cases
concern the same end-relational proposition Angie asserted. This suggests a
disjunctive view that generates some intuitive predictions; for example, that
Connie’s postmortem claim that Angie ought to have used L does not disagree
with Bertie’s advice or with her own hindsight claim that Angie ought to have
used R and vice versa.
Rather than simply giving a disjunctive theory, we can distinguish different
kinds of normative disagreement. Instrumental disagreement consists in a con-
flict of hypothetical preferences; it is disagreement over what to do to achieve
a particular end. Rational disagreement consists, rather, in inconsistency in
end-relational belief. But this rational disagreement also implies inconsistency
in a different kind of counterfactual preference, which we can call disagreement
in doubly hypothetical preferences: if the speaker were to prefer that e and only
had information i, then on the basis of her belief she would prefer that p. For
example, Connie’s postmortem claim that Angie ought to have used L prag-
matically indicates that if Connie were to prefer that Angie survived unharmed
and only had Angie’s information, then she would prefer that Angie had used
L. Rational disagreements can therefore also be glossed as concerning what to
do to achieve a certain end with certain information.
Asserting a normative sentence can therefore express multiple attitudes. In
saying ‘I ought to use L’, Angie expresses (i) an end-relational belief, (ii) an actual
preference to use L, (iii) a hypothetical preference to use L on the (actual) condi-
tion that she prefers to survive unharmed, and (iv) a doubly hypothetical prefer-
ence to use L on the (actual) conditions that she prefers to survive unharmed
and only has information i. This implies that two people may also stand in mul-
tiple relations of normative agreement or disagreement at the same time, which
seems another serendipitous result: while Bertie instrumentally disagrees with
Angie (in hypothetical preference), they rationally agree (in belief and doubly
hypothetical preference). He only expresses his instrumental disagreement, since

this is the felicitous thing to do in his context of direct interest. With relevantly
the same information in a context of indirect interest, Connie can rather felici-
tously express either her rational agreement (postmortem) or her instrumental
disagreement (hindsight). Intuitions of disagreement simpliciter will be sensitive
to whatever is the most salient kind of disagreement expressed in the context.
The second recalcitrant piece of data concerns negations, which pose a
well-known problem for expressivism. Recall Freddie, whose forensic investi-
gations revealed that the next use of R would have been fatal, and suppose now
that he is informed that the round in L was not a blank, causing him to believe
that Angie could not possibly have survived. He may then say in hindsight:
Freddie: ‘It was not the case that Angie ought to have used L. But not
because she ought to have used R; she simply had no good options.’
(Hindsight Negation)
Freddie seems to disagree with Angie, but we do not yet have an explana-
tion for this. It cannot be a rational disagreement (in belief), because Freddie
is concerned with different information. But neither does he express an incon-
sistent hypothetical preference, as if he said instead, ‘Angie ought not to have
used L,’ since not ought(p) does not entail ought(not-p). Negative claims do
not seem to express preferences at all, in which case they could not express
disagreement in preference.
However, our theory does predict that Freddie’s utterance indicates some-
thing about his preferences: that he does not hypothetically prefer that Angie
used L and not R.24 He pragmatically indicates that on the basis of his belief, if
he were to prefer that Angie survived unharmed, he would not prefer that she
used L.25 While this does not identify attitudes toward inconsistent contents, it
does provide a different kind of inconsistency of attitudes: preferring p and not
preferring p are logically inconsistent states of mind in a broad sense, which
also subsumes preferences toward inconsistent contents.
Might negations of normative claims express disagreement by virtue of
this attenuated kind of “inconsistent preferences”? This runs straight into the
24. Something stronger can be said in Freddie’s case: that he is (hypothetically) indifferent

about whether Angie uses L rather than R. This is not sufficiently general as an account
of believing not ought(p), however, since one may also do so on the basis of believing
ought(not-p), which we analyze as hypothetically preferring not-p.
25. What prevents Freddie from having a further desire that would cause him to prefer that
A used L in the nearest possibilities? I do not have space to discuss this here, but see Finlay
2014, chap. 8.
expressivist’s problem with negation. If Freddie’s saying ‘It wasn’t the case that
Angie ought to have used L’ expresses merely his lack of preference for Angie
using L, how can we explain the difference with Debbie’s saying ‘I don’t know
whether you ought to use L’? Or if believing not ought(p) is merely lacking the
preference that p, what distinguishes it from not believing ought(p)?26 As we
predict, Debbie also indicates she has no preference about Angie using L rather
than R, but unlike Freddie, she does not intuitively disagree with Angie. Lacking
an attitude does not in general seem to be a way of disagreeing with the atti-
tude. But unlike expressivism, ERT identifies a significant difference between
Debbie and Freddie. Whereas Debbie’s lack of preference is due to lacking any
relevant belief about whether Angie ought to use L, Freddie’s lack of preference
is grounded in his belief that it is not the case that Angie ought to have used L.27
The difference between believing not ought(p) and not believing ought(p) is
therefore that whereas in one case the lack of preference results from having a
relevant normative belief, in the other it results from lacking a relevant normative
belief. An explanation of why expressing a lack of attitude by negation can intui-
tively express normative disagreement then follows from our previous explana-
tion of why inconsistency in merely hypothetical preferences can be sufficient for
actual disagreement. As with those hypothetical preferences, Freddie’s hypotheti-
cal lack of preference is grounded in and expressed by an actual normative belief,
unlike Debbie’s. The end-relational theory therefore provides a quasi-expressivist
solution to the negation problem that is not available to (pure) expressivism itself.
T H E P R A G M ATI C S O F D I SAG R E E M ENT

F R O M D I F F ER E NT E NDS
Some of our most important normative and moral disagreements are appar-
ently fundamental, involving a conflict in basic ends.28 Relational theories are
accused of entailing that such disagreements are impossible, but this objec-
tion assumes some version of Inconsistent Belief and so no longer looks so
formidable. The prospects for extending the quasi-expressivist solution to the
26. E.g., Dreier 2006; Unwin 1999, 2001.

27. Although Debbie believes it is not the case that A ought relative to D’s information to
use L, in her context of lesser information this is not relevant, and the robustness condition
explains why she does not thereby intuitively disagree with A.
28. Attempts have been made to accommodate the absolutist character of moral disagree-
ment by analyzing moral claims as all relativized in the same way and therefore as not genu-
inely fundamental by our definition. For discussion, see Finlay 2014, chap. 8.
disagreement problem from end relativity look promising, since it is exactly

this kind of disagreement that inspires the idea of disagreements in attitude.
Above we distinguished two different kinds of normative disagreement implied
by ERT: rational disagreement in belief and doubly hypothetical preference, and
instrumental disagreement in hypothetical preference. Fundamental disagree-
ments evidently cannot be accommodated under either category, but this is not
a problem, as they seem to involve a distinct kind of disagreement in any case.
Consider first contexts of direct interest, where each speaker is actively moti-
vated toward the implicit end. (Suppose that Frank is felling a tree toward his
neighbor’s house, trying to get revenge for the poisoning of his dog; Dora says
‘You ought to stop!’, and Frank replies, ‘No, I ought to do this.’) Unlike contexts
of advice, here there are two different, incompatible ends. When Dora says ‘You
ought to stop,’ (i.e., in order not to damage your neighbor’s property, as she prefers),
she thereby also expresses her preference that Frank stop. When Frank responds
‘I ought to do this,’ he similarly expresses his preference that he continue. Here
we have attitudes toward inconsistent contents and a disagreement in preference.
This is an expressed disagreement in actual rather than hypothetical pref-
erences, which might be called outright normative disagreement.29 Not all
outright disagreements are fundamental: Angie and Bertie express prefer-
ences that are actual as well as hypothetical, for example, so Bertie disagrees
with Angie not only instrumentally (about what to do to survive) but also
outright (about what to do). But I maintain that all fundamental disagree-
ments are outright. Contexts of indirect interest may seem to provide coun-
terexamples, since here actual preferences are not always expressed. But
neither do normative claims relativized to different ends always intuitively
disagree or agree with each other. Given our definition of fundamental
disagreement as involving a basic conflict in preferred ends, only contexts
where actual preferences are expressed (whether of direct or indirect inter-
est) are relevant.
ERT does not only accommodate the existence of fundamental disagreement
but can also explain otherwise puzzling facts about its asymmetry. We have
already observed one case of asymmetrical outright disagreement, between
Bertie and Angie in light of the shared preference for some informational bases
over others. In the case of fundamental disagreements, the analogous asym-
metry is that speakers prefer some ends over others. Unlike the asymmetry of
information, the “quality” of an end will differ between speakers, predicting
29. To express an “actual preference” as intended here is to indicate one actually has it.
One may therefore express an “actual preference” one does not actually have, as in cases of
insincerity.
less asymmetry in fundamental disagreements than in instrumental disagree-

ments: Dora and Frank each prefers his or her own end, so their disagreement
in preference is symmetrical. But other cases will be asymmetrical wherever
the second speaker makes salient an end that the first speaker is also disposed
to prefer.30
Whereas the problem of disagreement is widely considered fatal to relational
semantics and to require either an expressivist or absolutist descriptivist account
of the meaning of normative language, we have seen that the end-relational
theory supports a pragmatic solution for both information relativity and end
relativity that accommodates ordinary intuitions of normative disagreement.
Despite being semantically descriptivist, it explains these intuitions as sen-
sitive to inconsistency in either belief or preference. While this might seem
undesirably messy, we have also seen that it accommodates some data that look
anomalous for nonrelational accounts despite their alleged advantages here.
Normative disagreement resists analysis exclusively in terms of inconsistent
beliefs and also exclusively in terms of inconsistent preferences. I conclude that
the end-relational theory can only be supported, and not refuted, by our prac-
tice and sense of normative disagreement.
References
Björnsson, Gunnar, and Stephen Finlay. 2010. “Metaethical Contextualism Defended.”
Ethics 121 (1): 7–36.
Cappelen, Herman, and John Hawthorne. 2009. Relativism and Monadic Truth.
Dietz, Richard. 2008. “Epistemic Modals and Correct Disagreement.” In Relative Truth,
ed. Manuel García-Carpintero and Max Kölbel, 239–263. Oxford: Oxford University
Press.
Dowell, Janice. 2013. “Flexible Contextualism about Deontic Modals.” Inquiry 56
(2–3): 149–178.
Dreier, Jamie. 2006. “Negation for Expressivists.” In Oxford Studies in Metaethics, vol. 1,
ed. Russ Shafer-Landau, 217–233. Oxford: Oxford University Press.
Finlay, Stephen. 2008. “Motivation to the Means.” In Moral Psychology Today: Values,
Rational Choice, and the Will, ed. David K. Chan, 173–191. New York: Springer.
Finlay, Stephen. 2009. “Against All Reason? Skepticism about the Instrumental Norm.”
In Hume on Motivation and Virtue, ed. Charles R. Pigden, 155–178. Basingstoke,
UK: Palgrave MacMillan.
Finlay, Stephen. 2014. Confusion of Tongues: A Theory of Normative Language.
30. This provides a solution to an objection to relational theories made by Thomson (2008)
that any normative ‘ought’ claim—even including “hypothetical imperatives” and those con-
cerning chess moves—can be a target of moral disagreement and therefore must have moral
content. For discussion, see Finlay 2014, chap. 8.
Gibbard, Allan. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Harman, Gilbert. 1996. “Moral Relativism.” In Moral Relativism and Moral Objectivity,
ed. Gilbert Harman and Judith Jarvis Thomson, 1–64. Cambridge, MA: Blackwell.
Horgan, Terence, and Mark Timmons. 1992. “Troubles on Moral Twin Earth: Moral
Queerness Revived.” Synthese 92 (2): 221–260.
Kölbel, Max. 2002. Truth without Objectivity. London: Routledge.
Kolodny, Niko, and John MacFarlane. 2010. “Ifs and Oughts.” Journal of Philosophy 107
(3): 115–143.
MacFarlane, John. 2007. “Relativism and Disagreement.” Philosophical Studies 132
(1): 17–31.
Regan, Donald. 1980. Utilitarianism and Cooperation. Oxford: Oxford University Press.
Richard, Mark. 2011. “Relativistic Content and Disagreement.” Philosophical Studies
156 (3): 421–431.
Ridge, Michael. 2014. Impassioned Belief. Oxford: Oxford University Press.
Robinson, Denis. 2009. “Moral Functionalism, Ethical Quasi-Relativism, and the
Canberra Plan.” In Conceptual Analysis and Philosophical Naturalism, ed. David
Braddon-Mitchell and Robert Nola, 315–348. Cambridge, MA: MIT Press.
Ross, Jacob, and Mark Schroeder. 2013. “Reversibility and Disagreement.” Mind 122
(485): 43–84.
Stevenson, Charles L. 1944. Ethics and Language. New Haven, CT: Yale University Press.
Thomson, Judith Jarvis. 2008. Normativity. Chicago: Open Court.
Unwin, Nicholas. 1999. “Quasi-Realism, Negation, and the Frege-Geach Problem.”
Unwin, Nicholas. 2001. “Norms and Negation: A Problem for Gibbard’s Logic.”
7
Hybrid Expressivism
How to Think about Meaning
JOHN ERIKSSON n
According to expressivism, asserting a moral sentence functions to express a

noncognitive attitude rather than a representational belief.1 According to cog-
nitivism, on the other hand, asserting a moral sentence functions to express a
representational belief rather than a noncognitive attitude. Given the nature of
moral thought and talk, it seems that the choice between these two extremes
is not very satisfying. Instead, the best theory must strike a balance between
expressivism and cognitivism—thus capturing the merits of both. This is what
so-called hybrid theories aspire to do. In Eriksson 2009 I advanced such a
theory based on R. M. Hare’s The Language of Morals (1952). A moral sentence
has both evaluative and descriptive meaning where the former is constant but
the latter standard relative. The main aim in this chapter is to explain how
I think an expressivist, as I take myself to be, should think about meaning in
general and in the light of this develop some of the details of my own view.
The outline of the chapter is as follows. In the first and second sections
I briefly present the view advanced in Eriksson 2009 and its solution to the
traditional Frege-Geach problem. In the third section I present some problems
for extant hybrid expressivist views and the standard expressivist way of think-
ing about meaning in metaethics. I then present the template that we should
1. Thanks to Mike Ridge and Guy Fletcher for organizing the conference on hybrid theories
were the main ideas of this chapter were presented. Thanks also to everyone who has pro-
vided helpful comments on the material at some point or other.
model moral language on in the fourth section and in the fifth section apply
it to a view like mine. In the sixth section I make some last remarks about
moral language, truth, and logic, which also explain why the view belongs in
the expressivist tradition.
HO MA G E T O H AR E
In The Language of Morals R. M. Hare argued that ethical terms have both
descriptive and evaluative meaning. To call an object good, according to Hare,
is to commend it. Indeed one cannot call an object “good,” in its evaluative
sense, without commending it. This is, on Hare’s view, essential to the evalu-
ative meaning of moral terms. Commendations, however, are not fortuitous.
Rather, they require grounds. We commend something because it has certain
characteristics. If someone says that Mike is a good person, it always makes
sense to ask the judger why he or she thinks that Mike is a good person. Exactly
what characteristics that we commend an object in virtue of is often not trans-
parent. Although I may not be sure why I think that Mike is a good person,
I am certain that there is something about him that makes him a good per-
son, for example, that he is generous, kind, or the like. What is crucial is that
there always is something that makes a good person good. For this reason, Hare
(1952, 145) claimed that “the judgment that a man is morally good is not logi-
cally independent of the judgment that he has other characteristics which we
may call virtues or good-making characteristics.”
The good making characteristics in turn depend on the kind of standard the
person is judging by. For instance, the good making characteristics accord-
ing to a virtue theorist may be the possession of certain virtues. It is these
characteristics that determine the descriptive meaning of an ethical term or
sentence.
The evaluative meaning of “Mike is good” asserted by our virtue ethicist
is to commend Mike, and the descriptive meaning is roughly that Mike pos-
sesses certain virtues. This idea also applies to other moral terms, for example,
“right” and “wrong.” The evaluative meaning of “Donating to charity is right”
is to commend charity. The descriptive meaning is, again, determined by the
standard judged by. For instance, in the mouth of an Aristotelian, the descrip-
tive meaning is that donating to charity is conducive to eudaimonia or would
be approved of by a phronemos agent or the like.
As I understand this, a moral sentence has the same evaluative meaning in the
mouths of different speakers. Regardless of what standard you are judging by, you
commend donating to charity by asserting that donating to charity is right. By
contrast, judging by different standards changes the descriptive meaning of the
Hybrid Expressivism 151
“Donating to charity is right.” Again, in the mouth of an Aristotelian the descrip-

tive meaning is that charity is conducive to eudiamonia or would be approved of
by a phronemos agent. In the mouth of a utilitarian, on the other hand, the descrip-
tive meaning is different. It means that donating to charity maximizes utility.
In Eriksson 2009 I advanced a hybrid expressivist theory inspired by some
of Hare’s central ideas. Indeed apart from the focus on meaning rather than
expression, Hare’s theory is no doubt very congenial to the spirit of contempo-
rary hybrid theories.2 Given that a speaker who asserts that donating to charity
is right is both commending and describing, it seems plausible to think both a
belief and a desire are expressed.3 I also argued that Hare’s framework suggests
a natural solution to the traditional Frege-Geach problem.
MO R A L I N F ER E N CE
That a moral utterance expresses approval or disapproval seems intuitively plau-

sible. For instance, asserting that Mike is good or that donating to charity is
right seem to express approval of Mike and donating to charity, respectively. This
accounts for the sense in which moral language is action guiding and explains
one of the main attractions of expressivism. It is, however, more difficult to say
what kind of state of mind is expressed when a moral sentence is embedded,
for example, if Mike is good, then it is right to donate to charity. In this context
no approval of either Mike or donating to charity seems to be expressed. Since
expressivists claim that the meaning of a moral sentence is explained by the state
of mind it expresses, one is left wondering what “good” and “right” mean in
embedded contexts. This also leads to problems having to do with moral infer-
ence. The standard example usually involves an argument like the following.
(1) Stealing is wrong.
(2) If stealing is wrong, then murder is wrong.
(3) Murder is wrong.
This argument seems to be valid. However, since no disapproval is expressed

in (2), “wrong” seems to have different meanings in (1) and (2), and the
2. For different views in this spirit, see, e.g., Barker 2000; Boisvert 2008; Copp 2001; Eriksson
2009; Finlay 2005; Ridge 2006; and Strandberg 2012. See Schroeder 2009 for a discussion of
some of these views and their respective virtues and vices.
3. Hare might agree that moral utterances express states of mind in a colloquial sense, but he
was not an expressivist in the sense to be developed below.
argument therefore commits the fallacy of equivocation (“wrong” is semanti-

cally ambiguous). I will not here try to explain or assess how expressivists have
tried to solve the problem.4 That would lead us astray. Instead, I will explain
how I think a hybrid view like the one advanced above can handle it.
As explained above, one of the reasons for an expressivist to go hybrid is to
hijack the belief part (or the descriptive meaning) associated with cognitivism to
account for moral inference. According to the view outlined above, the descrip-
tive meaning of “wrong” depends on the speaker’s standard. For instance, sup-
pose that A judges by standard S, according to which characteristics W are
wrong-making. This makes the descriptive meaning of (1) the following.
(1d): Lying is W.
The descriptive meaning of “wrong” does not change in unasserted contexts.

For instance, when the judger asks whether stealing is wrong, she or he wants to
know whether stealing has properties that are wrong-making (according to her
or his standard), that is, whether it is W.5 Given this, the descriptive meaning of
(2) will be the following.
(2d): If lying is W, then murder is W.
The term “wrong” has the same descriptive meaning in the first two premises
and the conclusion
(3d) Murder is W
follows logically. It is easy to see why A is committed to (3d) in virtue of accept-
ing (1d) and (2d). Accepting (1d) and (2d) but not (3d) would clearly involve a
kind of irrationality. It is, however, more difficult to see why A is committed to
the attitude expressed in (3).
Inspired by Hare, I argued that the explanation of why the evaluative con-
clusion follows (and why someone who accepts [1]‌and [2] is committed to
[3]) turns on the supervenience of the evaluative on the nonevaluative.6 First,
someone who accepts (1), that is, someone who judges that stealing is wrong,
4. See, e.g., Blackburn 1993, 1998; Gibbard 1990, 2003.

5. Of course this is not how the judger experiences it. We do not in general believe that moral
classifications depend on the standard we accept or our attitudes more generally. Rather, we
(or most of us) experience that the classifications are not dependent on us. The explanation
of this phenomenon has to do with the fact that we project our standard onto the world.
6. Frank Jackson (1999) advances a similar view. See also Thomas Hurka 1982.
is committed to thinking that stealing has certain wrong making characteris-

tics—given the standard in question, that stealing is W. Second, accepting (1)
requires not just thinking that stealing has certain wrong making characteris-
tics but also having a certain attitude, for example, disapproval, toward stealing.
Moreover and more importantly, accepting (1) commits the judger to disap-
proving of other acts that are judged to be W. If murder is judged to have W
(which the judger seems to be logically committed to in virtue of accepting [1]
and [2]), then he or she is committed to disapprove of it too. Failing to disap-
prove of murder therefore involves a kind of irrationality.
I still think that this is the outline of an attractive theory. It explains the
sense in which moral thought and talk is both descriptive and evaluative,
and it provides a plausible solution to the traditional Frege-Geach problem.
However, it still leaves a host of questions unanswered. For instance, it may
still be argued that the theory is guilty of a kind of equivocation in meaning.
Although “wrong” has the same descriptive meaning in (1) and (2), it seems
that (2) does not express disapproval of either stealing or murder. Given that
the theory is supposed to be a kind of expressivism, it seems that “wrong” will
not have any evaluative meaning in embedded contexts. I think it is correct that
“wrong” does not express any disapproval of either murder or stealing in (2),
but this does not show that (1) and (2) have different evaluative meaning. In
fact I have not said much at all about how expressing states of mind is relevant
to the theory advanced above. Indeed this may make one wonder in what sense
the view is expressivist to begin with. Below I hope to address these questions.
E X P R E S S I V I SM AN D M E ANI NG
Given the advent of hybrid theories, it seems as if the traditional dichotomy

between expressivism and cognitivism is no longer of much use. However, it
may still be argued that different hybrid theories are more or less expressiv-
ist or cognitivist in the traditional sense. Michael Ridge (2006), for instance,
conceives of his view as a natural heir to traditional expressivism, because it
preserves (or is partly motivated by) intuitions regarding, for example, moral
disagreement.7 Moreover, he suggests that “cognitivism” should be under-
stood as the view according to which “a moral utterance is guaranteed to be
true just in case the belief(s) it expresses is (are) true” (Ridge 2006, 307–308).
Expressivists, on the other hand, deny this. In the contemporary literature the
7. Ridge calls his view “ecumenical expressivism.” I will use “hybrid” rather than “ecumeni-
cal” for the simple reason that it is the term that is most frequently used to describe the kind
of views in question.
term “expressivism” is also commonly used to refer to a particular theory of

meaning. Allan Gibbard (2003, 7), for instance, explains that he uses “the term
‘expressivism’ . . . to cover any account of meanings that follows this indirect
path: to explain the meaning of a term, explain what state of mind the term can
be used to express.”
The view I endorse is supposed to be expressivist in both senses. It is supposed
to be consistent with the attraction of traditional expressivism, and the mean-
ing of a sentence is supposed to be explained by the state of mind it expresses.
Given the latter commitment, two questions now need to be addressed. First,
how should the expression relation be understood? Second, what kind of state
of mind does a moral sentence express?
The expression relation can be understood in a number of different ways.8 For
instance, it may be understood as a causal, intentional, or conventional notion.9
In fact the term “expression” seems to be multiply ambiguous. For example, it
is plausible to think that blushing causally expresses embarrassment. However,
it seems less plausible to think that the expression of states of mind in the sense
relevant for expressivism should be understood as a causal affair. It is possible, by
contrast to blushing, to express a state of mind by linguistic means without actu-
ally being in the state of mind. There are at least two ways of explaining this. First,
a sentence can express states of mind in virtue of conventions. Second, a speaker
can express a state of mind by intending to express a certain state of mind.10
The next question is what kind of state of mind is expressed. For instance, it
may be argued that the moral assertion “Stealing is wrong” expresses disapproval
of stealing. Although one may think that this is intuitive in assertoric contexts,
we saw above that it seems implausible when considering embedded contexts.
One may at this point think that the reason expressivism is deeply problematic
is because of the expressivists’ way of understanding the state of mind that is
expressed.11 Traditional expressivists argue that uttering a sentence like “Stealing
is wrong” expresses disapproval of stealing. If this is the kind of attitude that is
supposed to explain the meaning of the sentence, then the account seems to
run into problems when sentences are embedded. Daniel Boisvert and Ridge,
however, depart from the traditional expressivist view. “The attitude a speaker
8. Expressing a state of mind, it should be emphasized, is different from reporting that one
is in it.
9. See, e.g., Schroeder 2008 for discussion.
10. An expressivist should make room for both ways of expressing states of mind but explain
how they are related, e.g., if either sense is more fundamental. Wayne A. Davis, for instance,
distinguishes between speaker and word expression (see section four).
11. This criticism of traditional expressivism is explicit in Boisvert 2008.
expresses . . . is not directed (at least not by way of literal meaning) towards
the subject of the sentence, but towards things of a moral general kind, namely
things that have the property picked out by the ethical predicate” (Boisvert 2008,
178).“Speakers use moral sentences to express their approval of something quite
generally—either a certain sort of prescriber or simply a certain descriptive
property (or set of properties) and also express a belief whose content is partly
fixed by the object of the approval expressed” (Ridge 2009, 197).
By contrast to traditional expressivism, Ridge and Boisvert think that utter-
ing “Stealing is wrong” expresses disapproval of something quite general but not
of stealing in particular (at least not directly). Despite this similarity, Ridge and
Boisvert have radically different views about the descriptive meaning of moral
sentences. Ridge argues that the descriptive meaning of moral sentences is in
some sense speaker relative. Boisvert, on the other hand, argues that such a view
leads to insurmountable problems having to do with moral belief ascriptions.
These problems should lead us to think that the descriptive meaning is constant.
Given what I have said above, I agree with Ridge that the descriptive meaning of
a moral sentence is in some sense speaker relative. I will not here embark on a
lengthy defense of this claim but more or less take it for granted—although I think
that some of the considerations advanced below do point in this direction.12
Instead, I will explain why the move to general attitudes is problematic (in
particular for a view like Ridge’s). One may think that part of the traditional
problem with expressivism stems from the fact that expressivists think that, for
example, “Stealing is wrong” functions to express disapproval of stealing. When
the sentence is embedded, on the other hand, it does not express disapproval of
stealing. Hence it cannot be the kind of state of mind that explains the evalua-
tive meaning of the sentence. However, it may be argued that although the sen-
tence does not express any disapproval of stealing, it still expresses an attitude
toward things of a general kind.13 Consider, for instance, the following example:
(4): If donating to charity is right, then I’ll get my wallet.
A speaker asserting (4) plausibly does not express approval of donating to

charity. Nevertheless, it may seem reasonable to think the speaker expresses
approval of things of a general kind. If so, then it may be argued that it is this
12. Nor will I have space to address the kind of problems Boisvert and others think a view
according to which the descriptive meaning is speaker relative incurs. See Eriksson 2009 for
discussion of some of these issues.
13. Boisvert’s argument here relies on an analogy to pejorative terms. I will not here address
this delicate matter. However, see Eriksson and Strandberg (2013-10) for skepticism regard-
ing its value and plausibility.
kind of attitude that is relevant to explain the evaluative meaning of moral sen-
tences and moral inferences more generally.14
Although this suggestion is interesting, I think it is inadequate. First, it seems
to me that uttering a sentence like “Stealing is wrong” expresses disapproval of
stealing directly and, if at all, a more general disapproval of some property or the
like indirectly. Second and more seriously, there are examples where it seems
intuitively odd to think that an utterance expresses any approval or disapproval
toward things of a general kind. Consider the following variation on (4).
(5) If donating to charity were right, then I would get my wallet. But as you
know, I stopped believing in morality a long time ago.
It does not seem as if a speaker who utters (5) is expressing approval toward
things of a general kind. The reason that no approval (either general or specific)
is expressed is because there is no property that makes things right according
to the speaker. Here is a different example.
(6) Can I be utilitarian if I don’t believe that acts that maximize utility are
necessarily right?
Used in a pedagogical context, (6) does not seem to express approval of some-

thing quite generally. If this is right, then it seems that expressing approval
toward things of a general kind does not help explain the evaluative meaning in
all embedded contexts. At this point it may be argued that my intuitions simply
are different compared to Ridge’s and Boisvert’s or that these contexts can be
dealt with somehow. Maybe this is correct.15 There is, however, a different prob-
lem for hybrid theories, according to which the descriptive meaning or belief
part is speaker relative (as Ridge and I think is the case). Consider the following
short dialogue involving Jack and Jill.
Jack: Stealing is wrong.
Jill: Stealing is right.
According to a view like Ridge’s, it is conceivable that Jack and Jill express
beliefs with consistent contents. Perhaps Jack is an Aristotelian and Jill a
14. Importantly, it enables the expressivist to say something about the evaluative meaning in
embedded contexts.
15. It may, for instance, be argued that “expression” is a technical term that is not necessarily
intuitive (see, e.g., Schroeder 2008), that expressing disapproval of stealing can be explained
by the pragmatics of moral discourse and that propositional attitude contexts have to be
treated in a different way.
utilitarian (they approve of different ideal observers). If this is the case, Jack
will express the belief that stealing is not conducive to eudaimonia, and Jill will
express a belief that stealing is optimific. Given that this is the case, there is no
disagreement in belief.
Of course one of the attractions of expressivism, and something that Ridge
seems to think his theory inherits in virtue of being a natural heir to such views,
is the ability to account for apparent moral disagreement in the absence of dis-
agreement in belief.16 Charles Stevenson (1944, 1963) famously observed that
disagreement is not exhausted by disagreement in belief. We also sometimes
disagree in attitude. Although Jack and Jill do not disagree in belief, they dis-
agree in attitude, since both attitudes cannot be simultaneously satisfied.
However, the move to general attitudes seems to prevent Ridge from explain-
ing Jack and Jill’s disagreement as a disagreement in attitude. The reason is
that Jack (or Jack’s assertion) expresses a general disapproval of acts that a
phronemos agent would not recommend, and Jill (or Jill’s assertion) expresses
approval of acts that are optimific. The upshot of this is that Jack and Jill, who
intuitively appear to disagree about the moral status of stealing, seem to dis-
agree neither in belief nor attitude—their respective beliefs and their respective
desire-like states appear compatible. Hence nothing is left to explain the intu-
ition that Jack and Jill morally disagree.17
By contrast, according to the view I endorse, a moral utterance, for example,
“Stealing is wrong,” expresses disapproval of the subject of the sentence, namely,
stealing (and that judging that stealing is wrong is to disapprove of stealing).
My view is therefore not impaled on this horn. Instead, it seems to be impaled
on another horn, namely, it seems to inherit the expressivists’ problems regard-
ing embedded contexts. Views like Ridge’s and mine therefore seem to face a
dilemma. They cannot simultaneously explain disagreement in the standard
expressivist fashion and explain the evaluative meaning of embedded sentence.
There is, however, a way out of this dilemma. For a view like mine, the way
out turns on abandoning the standard way of thinking about the state of mind
that determines meaning and what meaning consists in. According to expres-
sivism, we explain the meaning of a sentence by explaining the state of mind
it expresses, and meaning consists in the state of mind expressed. The key to
understanding evaluative meaning is to first understand descriptive meaning.
16. See Tersman 2006 for discussion.

17. In Ridge 2013 it is argued that the Stevensonian conception of disagreement is inad-
equate, and a novel account is proposed (disagreement in prescription). However, I do not
think Ridge’s alternative suggestion improves the situation for the kind of hybrid view in
question. This argument is elaborated in Eriksson forthcoming.
As Gibbard (1992, 971) writes, “Show me how to do it in the factual realm,

and I’ll mimic you in the normative realm.” How, then, should we explain the
meaning of descriptive language? Consider a recent and influential suggestion
due to Mark Schroeder.
Each sentence . . . is associated with a condition under which it is permis-

sible (as far as semantics go) to assert it. For example, the sentence, “grass
is green” is semantically associated with the conditions that the speaker
believes that grass is green. As a shorthand, we can say that the sentence
“expresses” this belief. This belief has the content that grass is green. And
so the sentence comes to count as having this content derivatively. So it is
permissible to assert “grass is green” only if you believe that grass is green,
but not because the sentence has the content that grass is green—rather it
has the content because those are the conditions under which it is permis-
sible to assert it. (Schroeder 2008, 30–31)
First, given this suggestion, it does not seem implausible to argue that it is
permissible to assert (1)–(3) only if you disapprove of something quite gener-
ally and (4) only if you approve of something in that manner. However, it does
not seem very plausible to argue that it is permissible to assert (5) or (6) only
if one approves of something more generally (or approves of donating to
charity). Given Schroeder’s suggestion, (5) and (6) therefore do not seem
to express a more general attitude (or a more specific attitude). It therefore
seems as if we lack an explanation of the evaluative meaning of the “right” in
those contexts.
Second, Schroeder’s assertability semantics is in part motivated by the expres-
sivists’ failure to account for the most natural view, which Schroeder calls the
same-content account. For instance, it seems natural to think that the connec-
tion between “Grass is green” and “I believe that grass is green” is that “the first
sentence has the same content as the belief that the second mentions,” but this,
Schroeder (2008, 28) argues, is inconsistent with expressivism.
The reason is that descriptive sentences get their content from the beliefs
they express. But for that to happen it must first express a belief and only then
acquire a content. But the same-content account says that a sentence expresses
a belief by having the same content as that belief. So the same-content account
is not compatible with expressivism’s commitment to mentalism—expressivists
require different orders of explanation. Expressivists need to explain its hav-
ing that content by appeal to its expressing that belief, but the same-content
account explains its expressing that belief by appeal to its having that content
(Schroeder 2008, 25).
I think Schroeder’s suggestion is problematic for a number of different rea-

sons.18 The most important problem in this context is the suggestion that the
meaning of a descriptive sentence is derived from the belief expressed (and
by analogy that evaluative sentences derive their meaning from the approval
or disapproval—regardless of whether it is a general attitude or not).19 For
instance, it seems unintuitive to think that “Grass is green” expresses the belief
that grass is green when embedded.20 If this is right, then it seems that the
meaning of “Grass is green” cannot be explained by appeal to its expressing that
belief. Consider the following example.
(7) I wonder if grass is green.
It seems that “Grass is green” has the same meaning regardless of whether it
is embedded or not. However, it seems semantically permissible to assert
(7) although one does not believe that grass is green, but if (7) does not express
the belief that grass is green, then it cannot be the belief that grass is green that
(in part) explains the meaning of “Grass is green” in (7). This shows that there
is something deeply problematic about this approach to explain descriptive
language. It leads to problems distinguishing between force and content or lan-
guage use and linguistic meaning.21 Indeed it should be discarded. Expressing
a belief is what a speaker does by using a sentence in a certain way. It requires
performing a particular speech act. Similarly, expressing approval or disap-
proval of some act by using “right” or “wrong” requires making a particular
speech act. But how, then, should an expressivist think about meaning?
HOW TO THINK ABOUT MEANING: THE TEMPLATE
The most developed expression theory of meaning on the market is due to

Wayne A. Davis (2003, 2005). It is this theory that I believe should serve as a
template for understanding the meaning of moral language. Davis is mainly
18. See Eriksson 2010.

19. Confer Gibbard (2003, 75): “We explain what he means by explaining the state of mind
that he thereby expresses. Expressing we explain by analogy with prosaically factual state-
ments: Suppose Holmes instead says, ‘Moriarty will shortly arrive.’ He thereby expresses a
prosaically factual belief, his belief that Moriarty will shortly arrive. Expressing a state of
mind works the same in these two cases, but the states of mind are different.”
20. Confer Blackburn 1998, 71.
21. See, e.g., Eklund 2009.
concerned about descriptive language, and the central idea is summed up

nicely in the following passage. For the majority of expressions, “meaning
consist in the expression of ideas, and . . . meaning is determined by the idea
expressed. The word ‘dog’ means something, for example, because it is conven-
tionally used to express an idea. And it means ‘dog’ rather than ‘cat’ because
it expresses the idea of a dog rather than the idea of a cat” (Davis 2003, 553).
Davis argues that we should understand word meaning in terms of speaker
meaning. What a speaker means by a particular word depends on what idea
the speaker expresses. Before we proceed, both “idea” and “express” require
brief explanations. “Ideas, on the one hand, are understood as event-types of
a certain sort, specifically thoughts or part of thoughts” (Davis 2003, 35) and
should not be confused with images. Expression, on the other hand, is under-
stood in terms of indication. To express an idea, thought, or other mental state
is “to provide an indication that it is occurring to us” (Davis 2003, 43). What a
speaker means by a particular word thus depends on what idea the speaker is
indicating. For instance, a speaker may use “boulder” to mean a kilo of cocaine
by intending to indicate that the idea of a kilo of cocaine is occurring to him
or her. However, the word “boulder” does not literally mean a kilo of cocaine
but a large rounded stone block. But how does a word like “boulder” come to
mean what it does in English? Davis (2003, 25) argues that what “words mean,
imply or express depends not on any particular speaker’s intention, but rather
on conventional usage and the rules of language, which are determined by the
intentions of prior speakers of the language.” The explanation of why “boulder”
means large rounded stone is thus that prior speakers of English have used the
word to indicate having a particular (complex) idea occurring to them. At some
point in time this use became conventionalized. As a result the word “boulder”
conventionally came to express the idea of a large rounded stone block. We can
therefore think of the English language as “the system of expression defined by
the evolving conventions of a certain community of language users who pass
their conventions on from generation to generation” (Davis 2003, 554).
Sentences in turn are composed of words (apart from the possible excep-
tion of one-word sentences). According to Davis, the meaning of a sentence
is explained by the thought it expresses. The meaning of, for example, “Grass
is green” is explained by the thought it expresses, namely, that grass is green,
and meaning consists in the expression of the thought that grass is green.22 At
this point it is important to emphasize that thoughts should not be conflated with
beliefs. To illustrate, it is possible to entertain the thought that Smaug is guarding
the gold in Erebor even though one does not believe that he does. Smaug, after all,
22. Below I will sometimes drop talk of “expression” and merely say that meaning consists
in states of mind.
does not exist. In other words, we should not conflate thoughts and beliefs. This
is important for a number of reasons. First, the meaning of the sentence “Smaug
is guarding the gold in Erebor” consists in the thought that Smaug is guarding the
gold in Erebor. A thought is a cognitively more basic state of mind than a belief.
Second, an agent may believe, hope, wish, or fear that Smaug is guarding the gold
in Erebor. In other words, the object of these different propositional attitudes is the
same, namely, the thought that Smaug is guarding the gold in Erebor.
In relation to the second point, it is also important to emphasize that “thought”
is ambiguous. It may denote either relational objects of thoughts or acts of think-
ing. In this context “thought” is used in “the sense in which it means an object
rather than an act of occurrent thought, and a thought-type rather than a
thought-token” (Davis 2003, 342). This is important, because it explains how the
thought that Smaug is guarding the gold in Erebor can be the same object of dif-
ferent propositional attitudes. This makes the role of thoughts suspiciously similar
to the role played by propositions in standard truth-conditional theories. In fact
Davis (2003, 342) thinks that “thinking the thought that p is the same process as
conceiving the proposition that p.” An immediate objection to this suggestion is
that propositions cannot be identified with mental states—as seems to be required
for the view to be a form of expressivism in the relevant sense. Propositions are
abstract objects (Gottlob Frege) or sets of possible world (David Lewis). This is
bad news for the present proposal, because a theory needs propositions to serve
as the objects of propositional attitudes and to identify and track cognitive states
of agents. However, one may plausibly argue that the traditional conceptions of
propositions mentioned above are highly problematic. In fact in the light of these
problems one may think it better to offer a new story about propositions.
Scott Soames, for instance, advances two ways such a story can be developed.
First, “propositions are theoretical devices for tracking acts of predications by
agents” (Soames 2010, 99). On what he calls a deflationary approach, propo-
sitions are theorists’ creations and are identified with tree structures (famil-
iar form syntactic theory). Hence “the proposition expressed by a sentence
is a hierarchical structure paralleling the syntactic structure of the sentence
itself ” (Soames 2010, 70). On the alternative conception, the cognitivist-realist
approach, propositions are identified with act types rather than abstract struc-
tures. To entertain the thought that snow is white is to predicate whiteness of
snow. “This cognition is both an event involving me that occurs at a particular
time and place, and also an instance of a corresponding event type in which
an agent predicates whiteness of snow” (Soames 2010, 103). I think both ideas
offer expressivist-friendly ways of understanding propositions that are both
plausible and promising, but I will not delve into this matter any further.23
23. See Soames 2010, chaps. 5–7, for discussion.

The question we started out asking was how to explain the meaning of
descriptive language. On the proposal advanced above, we should not explain
the meaning of “Grass is green” in terms of the belief that grass is green. The
meaning of “Grass is green” is explained by the thought (event type) that grass
is green expresses. One may think this is merely an unimportant exercise in
metasemantics, that is, an explanation in virtue of what a sentence acquires
its semantic value. However, this helps gain a better understanding of how we
should think about the semantics, that is, what the meaning of a sentence con-
sists in. Once we acknowledge that the descriptive meaning of a sentence is not
determined by expressing a belief but a cognitively more basic state of mind, a
thought, we should not think that the meaning of, for example, “Grass is green”
consists in the belief that grass is green. Rather, the meaning of “Grass is green”
consists in the thought that grass is green. Moreover, this helps explain why
expressivism is not inconsistent with the same content view after all. “Grass
is green” has the same content that the sentence “I believe that grass is green”
mentions, namely, the thought that grass is green. In other words, it is not the
case that an expressivist has to explain the meaning of a sentence by appeal to
its expressing a particular belief.
This also avoids explaining the meaning in terms of the kind of speech act
that is performed. To express the belief that grass is green requires asserting
that grass is green. However, expressing the belief that grass is green is not what
determines the meaning of “Grass is green.” We can thus explain the difference
between linguistic meaning and language use.24 Finally, to believe that grass is
green is not merely to entertain the thought that grass is green. Rather, it is to
(be disposed to) endorse the thought, that is, to (be disposed to) endorse the
predication of greenness of grass. This is the kind of structure an expressivist
about evaluative language should mimic.
HY B R I D E X P R E SSI V I SM
As I explained above, moral sentences have descriptive meaning. Hence a moral

sentence derives its meaning in part from expressing a thought. If the descrip-
tive meaning of a sentence is explained by the thought it expresses, then what
explains the meaning of the evaluative or nondescriptive part must be some-
thing different from a thought. When thinking about the meaning of moral
terms and sentences on expressivist premises, focus is almost exclusively on
assertoric uses of moral sentences. The reason is presumably that it seems rather
24. See also Dorit Bar-On and Matthew Chrisman’s (2009) distinction between s-expression
(semantic expression) and a-expression (action expression).
plausible to think that, for example, “Stealing is wrong” functions to express

disapproval of stealing in such contexts (and hence, given the expressivist view,
explains the semantics in terms of disapproval). However, this is not the only
way we use moral language. Sometimes we ponder what to do; we seek advice,
reason together, and propose hypotheses that we do not (yet) endorse. In such
contexts moral sentences do not express any approval or disapproval, but it
seems strange to think the moral terms used have different meanings in these
contexts. Surely we are asking for moral advice when we ask what the right
thing to do is. We are not using “right” with a different kind of meaning. Since
this is the case, maybe trying to explain the meanings of moral sentences in
terms of expressing attitudes like approval or disapproval simply is misguided.
“Donating to charity is right” or “Stealing is wrong,” given our commitment to
expressivism, must be used to express some other kind of state of mind.
It is at this point that Davis’s suggestion regarding descriptive language
starts paying dividends. For example, the meaning of the sentence “Grass is
green” is not explained by the belief that grass is green but by the thought that
grass is green. It is by using the sentence with a certain force (i.e., by perform-
ing a certain speech act) that the belief that grass is green is expressed. The
meaning of “Grass is green” consists in the thought that grass is green. It is this
kind of structure that we want to mimic to explain the evaluative meaning of
moral sentences.
But if we should not explain the evaluative meaning of moral sentences in
virtue of approval or disapproval, then what mental state types explains their
meanings? Just as we can approach the meaning of a descriptive sentence, for
example, “Grass is green,” via thinking about the kind of thought we entertain
when we entertain the thought that grass is green, we can approach evaluative
meaning by thinking about the kind of state of mind we are entertaining when
we entertain a moral thought, for example, that stealing is wrong.25
What is it to entertain the moral thought, for example, that stealing is wrong?
Here is a suggestion. To entertain that stealing is wrong is (1) to entertain that
stealing has certain characteristics, which depend on the standard judged
by. However, it also involves (2) entertaining a state of mind the function of
which is such that if one accepts or endorses it, one is disposed to disapprove
of stealing. For example, the meaning of “Stealing is wrong” in the mouth of a
utilitarian consists in (1) the thought that stealing is not optimific and (2) the
nondescriptive state of mind the function of which is such that if one accepts it,
one is disposed to disapprove of stealing. This suggestion more or less dovetails
with the kind of view advanced in Eriksson 2009. An ethical sentence has both
25. Moral thought is supposed to denote the kind of state of mind that is entertained when
we think, e.g., stealing is wrong. It is therefore not necessarily a thought in Davis’s sense.
descriptive and evaluative meaning. The former is standard relative and the
latter is constant. Things are, however, a little more complicated. Below I will
elaborate on how to understand the different parts of this suggestion.
The descriptive meaning depends on the standard judged by. Most often, it
seems, the relevant standard is the judger’s own. However, it is perfectly pos-
sible to entertain a moral thought that one does not accept and even rejects. For
instance, a judge may entertain the moral thought that stealing is right although
he or she is firmly against it. When the judger entertains this moral thought,
he or she may do one of two things. First, the judger may entertain the thought
that stealing has properties that are wrong-making according to the judger’s
standard but not endorse the predication of the properties to stealing (because,
for instance, the judger does not actually believe that stealing has the relevant
properties). Second, the judger may use a standard other than his or her own.
For example, Jack may use a utilitarian standard when thinking about what a
utilitarian would think about a certain question. When he does this, he pre-
sumably believes that he is entertaining a moral thought. However, it may well
be the case that the moral thoughts he entertains are ones that he disagrees
with, that is, moral thoughts that he does not endorse.
This is also reflected in moral talk. Most often a speaker is concerned to
get his or her ideas across when asserting a moral sentence. When this is the
case, the descriptive meaning depends on the standard the speaker endorses.
However, the relevant standard is not always, as illustrated above, the speaker’s
own. For instance, in discussing a moral issue one may, for example, take the
perspective of someone who endorses a certain standard to try out a certain
perspective. A committed liberal may, for instance, take on the perspective of a
socialist or vice versa. In such a context an interlocutor will (hopefully) under-
stand that the standard judged by is not the speaker’s standard. If all goes well,
this will be displayed in the kind of considerations that an interlocutor counts
as right or wrong-making according to the speaker. These considerations illus-
trate the complex sense in which the descriptive meaning of a moral sentence
is standard relative.26
Moreover, to entertain the moral thought that stealing is wrong is to entertain
a mental state the function of which is such that if one accepts it, then one is dis-
posed to disapprove of stealing. It is this state of mind that makes a moral thought
evaluative. It is in virtue of this part that we sense that the meaning of a moral
sentence is not exhausted by its descriptive part. But what kind of state of mind is
this? Following Gibbard (2003, 53), we can perhaps think of this state of mind as
a plan, where a plan is “a determination of what to do in various contingencies,
26. In fact these considerations suggest that we do in fact recognize that people mean,
descriptively, quite different things when they use moral sentences.
expected or hypothetical.” The evaluative meaning of a moral sentence thus con-

sists in the state of mind involved in entertaining a determination of what to do.
It is important to emphasize that entertaining a plan or a determination of what
to do, as I understand it, is not to endorse it. One can entertain the moral thought
that stealing is wrong without endorsing the determination of what to do. The
evaluative meaning is also constant. Regardless of what kind of standard one is
judging by, the evaluative meaning of, for example, “Stealing is wrong” consists in
the state of mind involved in entertaining a determination not to steal. Moreover,
this is part of the meaning of “Stealing is wrong,” regardless of whether the sen-
tence is embedded or not. In such a context “Stealing is wrong” does not express
any disapproval of stealing, but it nevertheless has the evaluative meaning it does
in virtue of consisting in a noncognitively more basic state of mind.27
To express, for example, the belief that stealing has certain characteris-
tics and/or disapproval of stealing, the sentence must be used assertorically.
However, as has been stressed above, it is not the expression of the belief or
disapproval that determines the meaning of the sentence. Finally, to accept the
moral thought that stealing is wrong (or, in other words, to believe in a minimal
sense that stealing is wrong), the judger must endorse the determination not
to steal—where endorsing the determination not to steal can be understood as
being disposed to disapprove of stealing.
MO R A L L A N G UAG E , TR UTH, AN D LOGIC
According to the view advanced above, the meaning of a token sentence, for
example, “stealing is wrong,” consists in (1) the thought that stealing has a cer-
tain property and (2) the state of mind involved in entertaining a determination
of what to do. The kind of property in question (the wrong making character-
istics) is standard relative. Talk about standard relativity may lead one to think
that the resulting view must be either some kind of (covert) subjectivism or
contextualism according to which the truth conditions of a moral judgment
depend on the judger’s attitude or standard. Subjectivism and contextualism are
forms of cognitivism. As explained above, the advent of hybrid theories makes
the old characterization of cognitivism and expressivism difficult to apply. We
can instead, following Ridge, understand cognitivism as the view according to
which a moral utterance or moral judgment is guaranteed to be true just in case
the belief or proposition (thought) it expresses is true. Expressivists, on the
other hand, deny this (see the third section).
27. This account, it should be noted, can be appropriated by a variety of views about what
sort of noncognitive state is involved in accepting the thought in question.
What should we say about truth talk and truth conditions according to a
view like the one outlined above? We surely talk as if moral utterance and moral
judgments can be true or false, but is a moral utterance guaranteed to be true
just in case the belief it expresses is true? Begin by considering an example due
to Stephen Barker (2000, 277) featuring the racist Norm. Norm is watching a
film and gestures at the SS officer Schmidt and asserts the following:
(8) Schmidt was good.
Consider now the following possible reactions to Norm’s assertion.
(9) What Norm said was true.

(10) What Norm said was false.
Barker’s general idea is in many respects similar to mine. Roughly, “Schmidt

was good” expresses as explicature that Schmidt was F, where F depends on
what kind of things Norm approves of, which is expressed as implicature.
However, Barker thinks the truth conditions of (8) hold (given that Schmidt
was F). Similarly, according to subjectivism, (8) is guaranteed to be true if the
proposition expressed is true, namely, if the speaker, that is, Norm, approves of
Schmidt. According to contextualism, on the other hand, (8) is guaranteed to
be true if the relevant proposition or belief (where the proposition or belief is
in some sense determined by the speaker’s standard, available information, or
the like) expressed is true.
On these accounts the reason (8) is guaranteed to be true if the proposition
or belief is true is because the meaning of (8) is exhausted by the descriptive
part. However, on the account advanced in the previous section, this is not the
case. By contrast to Barker’s view, for example, the evaluative part is not merely
expressed as implicature but a genuine constituent of the meaning of a moral
sentence. Part of the meaning of “Schmidt was good” consists in the state of
mind that is involved in entertaining a determination of what to do or feel, for
example, to approve of Schmidt. Of course I am not denying that Schmidt was
F is true iff Schmidt was F, but I reject that Schmidt was good is guaranteed to
be true if Schmidt was F. Let me explain.
Consider a remark in the spirit of minimalism about truth: to say that some-
thing is true is to do little more than to assent to it. By using expressions like
“What Norm said was true” or the like, “we are confirming, underwriting,
agreeing with, what somebody said” (Strawson 1949, 93). Asserting (9) is con-
sequently to assent to what Norm said, namely, that Schmidt was good. This
seems right. However, since the meaning of (8) has both descriptive and evalu-
ative parts, the question is which part guides our use of the truth predicate.
First, we may be quite aware of the standard Norm is judging by and that the
descriptive meaning of Schmidt was good in the mouth of Norm is that Schmidt
was F (and that Norm believes Schmidt was F). This is important to understand
what kinds of judgments Norm is committed to. For instance, if Norm thinks
Schmidt is good because he was F, then he is also committed to judging that
Herman is good if Herman is F—or more generally that the set of things that
are F are good. It also explains why Norm is committed to judging that things
that fall outside the relevant class are not good (in the evaluative sense). The
descriptive meaning is, in other words, important to understand the logic of
moral language.28 However, even though we may agree that Schmidt was F, it
seems that this is not primarily what guides our use of the truth predicate in the
moral domain, that is, we do not assert (9) simply because we agree in belief
(that Schmidt is F). Rather, it is primarily guided by considerations having to
do with agreement or disagreement in attitude. In other words, asserting (9) is
not endorsing that Schmidt is F but endorsing (roughly) a determination to
commend Schmidt.
Second, suppose we think Norm is mistaken about the kind of characteristics
Schmidt has. Schmidt, we may believe, actually is not the nefarious charac-
ter that Norm thinks—perhaps only a superficial understanding of his actions
warrant thinking this. In other words, Schmidt was not F, and Norm’s belief
that Schmidt was F is assessed as false. If it is the descriptive part or Schmidt’s
belief that drives our use of the truth predicate, then it seems that this should
lead us to assess the moral judgment as false. However, suppose we judge that
Schmidt is good (because he has characteristics that are good-making accord-
ing to our standard). Given that this is our view, we would most likely assent
to what Norm says despite the fact that we think the descriptive part can be
assessed as false.29 The explanation is, again, that assenting to what somebody
says in the moral domain is primarily guided by the attitudinal rather than the
descriptive component.30,
Given the view advanced above, this is what ought to be expected. Although
the descriptive meaning of a token sentence, for example, “Schmidt was good,”
is fixed relative to the standard judged by, it does not have any fixed descrip-
tive meaning at the metasemantic level. At this level “Schmidt was good” is
consistent with indefinitely many descriptive contents. The evaluative meaning,
28. In virtue of the meaning of “good,” Norm is also of course committed to certain attitudes
toward things that have the relevant property.
29. This would probably also require explaining one’s grounds in order to dissociate oneself
from the racist standard that Norm judges by.
30. Again, this is not how our assessment is experienced. We do not think that our standard
is what determines the wrong making characteristics or that agreement in attitude is what
by contrast, is fixed. This explains why it is not a linguistic mistake to endorse

someone’s moral judgment although one rejects the descriptive part (since no
descriptive part is constant or an essential part of a moral judgment). It also
explains why the attitudinal component is constant and more fundamental.
Moral language (and thought), after all, is not primarily designed to describe or
represent but to assist us in determining how to live.
The considerations advanced above also explain why no linguistic mistake
is involved in asserting (9) and thus assenting to (8) yet disagreeing with the
descriptive part. Similarly, it is not a linguistic mistake to assert (10) and thus
disagree with (8) although one agrees that Schmidt is F while acknowledging
that this is what (8) means descriptively.31 According to both subjectivism and
contextualism, by contrast, it seems that it would indeed be a linguistic mistake
to endorse someone’s moral judgment if one rejects the descriptive part. The
reason is not that the descriptive meaning of a sentence is constant but that that
subjectivists and contextualists standardly explain meaning in terms of truth
condition. According to subjectivism, for example, the meaning of “Schmidt
was good” is explained by what it is for the sentence to be true, namely, that
the speaker approved of Schmidt. Consequently, asserting (9) thus seems to be
a way of reporting that a certain state of affairs obtained, namely, that Norm
approved of Schmidt.32 On the view sketched above, on the other hand, to say
that something is true is not to say that some state of affairs obtains. To assert
(9) is not to report that Schmidt has a certain property or that Norm approves
of Schmidt. Rather, it is to express agreement in attitude, but it is quite possible
guides the application of “is true.” The claim is merely that this is the kind of underlying
mechanism that explains truth talk. By contrast, the meaning of a prosaically descriptive sen-
tence, e.g., “Grass is green,” consists merely in the thought that grass is green. In such worldly
matters, our use of the truth predicate is guided by agreement in belief (i.e., in endorsing the
thought in question). An expressivist cognitivist could thus argue that the truth predicate in
moral matters is regulated by agreement or disagreement in belief.
31. To assert (9) but disagree in attitude on the other hand, seems linguistically more
confused.
32. Of course, an expressivist theory of meaning is not inconsistent with a subjectivist or a
contextualist understanding of moral sentences. On such views, however, the meaning of a
moral sentence would consist only in a thought. For example, according to subjectivism, the
meaning of “Schmidt was good” would consist in the thought that the speaker, e.g., Norm,
approved of Schmidt. First, this exhausts the meaning. Second, although the descriptive
meaning is not determinate at the metasemantic level since no definite speaker is picked out,
it is nevertheless the case that it picks out the speaker of the sentence. On this view assert-
ing (9) is endorsing the thought that Norm approved of Schmidt, and rejecting the thought
that the speaker approved of Schmidt would be a linguistic mistake. Consequently, a moral
sentence seems to be guaranteed to be true if the thought is true. Similar considerations, for
similar reasons, apply to contextualism.
to agree in attitude yet disagree in belief (or vice versa).33 Consequently, it

seems that “truth” itself should be understood in broadly expressivist terms.
This explains why a moral sentence is not guaranteed to be true just in case the
belief or descriptive part is true. Indeed this is part of the lesson we should draw
form Hare’s claim that the evaluative meaning is primary and the descriptive
meaning secondary.34
C O N C L U D I N G R E M AR K S
Moral sentences have both descriptive and evaluative meaning, but how should
hybrid expressivists (and expressivists more generally) explain this? In this
chapter I have argued that we should not explain descriptive meaning in terms
of beliefs or explain evaluative meaning in terms of, for example, approval or
disapproval. Instead, an expressivist should explain the descriptive and evalu-
ative meanings in terms of a more basic cognitive and noncognitive state of
mind. This allows an expressivist to distinguish between linguistic meaning and
language use. I have also indicated how a view like the one I endorse can be
developed against this background and how to think about truth in relation to
it. Although I find the resulting view attractive, a lot of details still need to be
worked out more properly, but that is beyond the scope of this chapter.
References
Barker, S. 2000. “Is Value Content a Component of Conventional Implicature?” Analysis
60: 268–279.
Metaethics, vol. 4, ed. R. Shafer-Landau, 133–166. Oxford: Oxford University Press.
Blackburn, S. 1993. Essays in Quasi-Realism. Oxford: Oxford University Press.
Blackburn, S. 1998. Ruling Passions. Oxford: Clarendon.
Boisvert, D. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89: 169–203.
Davis, W. 2003. Meaning, Expression, and Thought. Cambridge: Cambridge University
Press.
Davis, W. 2005. Nondescriptive Meaning and Reference. Cambridge: Cambridge
University Press.
Eklund, M. 2009. “The Frege-Geach Problem and Kalderon’s Moral Fictionalism.”
Philosophical Quarterly 59: 705–712.
33. In other words, one important function of the truth predicate is that it functions as a (dis)
agreement (in attitude) marker.
34. The ideas in this section are, it should be emphasized, programmatic and very preliminary.
Eriksson, J. 2009. “Homage to Hare: Ecumenism and the Frege-Geach Problem.” Ethics

120: 8–35.
Eriksson, J. 2010. “Being For: Evaluating the Semantic Program of Expressivism, Mark
Schroeder.” The Philosophical Quarterly 60: 878–882.
Eriksson, J. Forthcoming. “Explaining Disagreement: A Problem for (Some) Hybrid
Expressivists.” Pacific Philosophical Quarterly.
Eriksson, J., and C. Strandberg. 2013-10 “Hybrid Expressivism, Pejoratives, and General
Attitudes: An Argument for License for Pessimism.”
8:205–223.
Gibbard, A. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press.
Gibbard, A. 1992. “Reply to Blackburn, Carson, Hill, and Railton.” Philosophy and
Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Hare, R. 1952. The Language of Morals. Oxford: Clarendon.
Hurka, T. 1982. “The Speech Act Fallacy Fallacy.” Canadian Journal of Philosophy
12: 509–526.
Jackson, F. 1999. “Non-cognitivism, Validity, and Conditionals.” In Singer and His
Critics, ed. Dale Jameison, 18–37. Oxford: Blackwell.
Ridge, M. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116:302–336.
Ridge, M. 2009. “Moral Assertion for Expressivists.” Philosophical Issues 19:182–204.
Ridge, M. 2013. “Disagreement.” Philosophy and Phenomenological Research 86:41–63.
Schroeder, M. 2008. Being For. Oxford: Oxford University Press.
Schroeder, M. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119:257–309.
Soames, S. 1999. Understanding Truth. Oxford: Oxford University Press.
Soames, S. 2010. What Is Meaning? Princeton, NJ: Princeton University Press.
Stevenson, C. 1944. Ethics and Language. New Haven, CT: Yale University Press.
Stevenson, C. 1963. Facts and Values. New Haven, CT: Yale University Press.
Phenomenological Research 84: 87–122.
Strawson, P. 1949. “Truth.” Analysis 9:83–97.
Tersman, F. 2006. Moral Disagreement. Cambridge: Cambridge University Press.
PART TWO
Pessimism About Hybrid

Theories
8
Moral Utterances, Attitude

Expression, and Implicature
GUY FLETCHER n
A recent trend in metaethics has been the development of so-called hybrid the-
ories of moral thought and language. Hybrid theories are those that subscribe
to at least one of these claims:
(i) moral thought: Moral judgments have belief and desire-like aspects or

elements.
(ii) moral language: Moral utterances both ascribe properties and express
desire-like attitudes.1
One interesting subclass of hybrid views and one of the most developed is
implicaturist hybrid views of moral language. Proponents of these views claim
that moral utterances express or convey the speaker’s desire-like attitudes via
implicature. Thus whatever their attitude toward (i), “implicaturists” hold (ii)
along with the claim that the desire-like attitudes are expressed by moral utter-
ances via implicature.
1. Theories that hold at least one of these claims include Bar-On and Chrisman 2009; Boisvert
2008; Copp 2001, 2009; Ridge 2006, 2007. Note that a “besire” view would also conform to
(i), given that (i) is formulated in terms of aspects or elements and that “ascribe” is not used
as a success term here. One could hold an error-theoretic hybrid theory.
It is unsurprising that some have claimed that moral utterances convey atti-
tudes via implicature. It seems clear that when we engage in moral discourse
we learn about the desire-like attitudes of our interlocutors. But at the same
time many want to resist holding that moral utterances simply report desire-like
attitudes (à la subjectivism) or simply express them (à la pure expressivism). An
implicature view thus suggests a way to explain the connection between moral
utterances and corresponding desire-like states while avoiding well-known
difficulties associated with subjectivism and pure expressivism.2 And implica-
turists—most notably Stephen Barker, David Copp, Stephen Finlay, and Caj
Strandberg—have been keen to highlight such advantages.3
Nonetheless, there is considerable work left to do in determining whether the
expression of moral attitudes by moral utterances is a matter of implicature. In part
this is simply because this is a relatively new line of research, but it is also because
implicaturists have tended to focus on some aspects of implicature and how atti-
tude expression fits them but have passed over other details and issues concerning
implicature. Furthermore, implicaturists have often engaged in the task of arguing
for one form of implicature view over another, a context that tends to downplay
the question of how well moral attitude expression fits implicature in general.4
This chapter rectifies this by outlining a comprehensive set of tests for each
form of implicature before assessing how well moral attitude expression fits
them. My conclusions are (1) that generalized conversational implicature (GCI)
is the most promising implicature theory of moral attitude expression but
(2) that this view faces a threat from a more minimal pragmatic theory, out-
lined at the end of the chapter, that shares all the same virtues but that is able to
explain more cases while positing less.
Here is how the discussion proceeds. After dealing with some preliminaries,
I begin (second section) by outlining some of the Gricean theory of implicature
needed to frame the discussion. I then compare the expression of desire-like
attitudes by moral utterances to the three main forms of implicature (conven-
tional, particular conversational, and generalized conversational) using five
standard criteria for implicature: indeterminacy (third section), reinforceabil-
ity (fourth section), nondetachability (fifth section), cancelability (sixth sec-
tion), and calculability (seventh section). I close (eighth section) by considering
some problems for the GCI view and outlining a competitor.
2. Examples of such pure expressivist theories include Ayer (1952); Blackburn (1998).
3. Finlay (2004, 2005) and Strandberg (2011) defend conversational implicature views,
Barker (2000) and Copp (2001, 2009) defend conventional implicature views.
4. Why? Because there might be good arguments for why attitude expression is more like
type 1 than type 2 that do not address how closely it resembles type 1 generally.
Moral Utterances, Attitude Expression, and Implicature 175
Before getting started it will be useful to clarify some terminology. I use “moral
judgment(s)” to refer to the mental state(s) a subject is in when he or she has a
view about some moral issue rather than the process of forming such views. By
“moral utterance(s)” I mean a speaker uttering a sentence that applies some moral
predicate. Paradigm moral utterances are “torture is wrong,” “giving to charity
is permissible,” “saving lives is obligatory.” I also assume, along with implicature
theorists, that at least one of the attitudes standardly conveyed by moral utter-
ances is desire-like (unmodified uses of “attitude” mean “desire-like attitude”).
The general issue here is how desire-like attitudes are conveyed by moral
utterances. For the sake of brevity and because of its intuitive plausibility, I do
not argue for the claim that moral utterances at least typically convey informa-
tion about the desire-like attitudes of the speaker.5 That claim is weak, plau-
sible, and accepted on all sides of the debate here. The relevant alternative to
the implicaturist view cannot be that desire-like attitudes are not conveyed
by moral utterances. That is implausibly strong. The relevant alternative to an
implicaturist view must be that implicature is not how desire-like attitudes are
standardly conveyed by moral utterances.
Finally, there are many differences (both actual and possible) between spe-
cific implicature theories. For example, implicature views can differ about what
is implicated, and also on how the implicature is produced.6 For the most part,
however, my discussion abstracts from these specific features of the views to
concentrate on more general features. This is to keep the discussion tractable
but also because these differences do not make a significant difference to the
general viability of implicature views.
GR I C E A N I M PLI C ATU R E
H. P. Grice distinguishes two species of implicature: (i) conversational impli-

cature and (ii) conventional implicature. Conversational implicatures are gen-
erated by general principles of rational behavior as speakers exploit common
rational expectations to communicate propositions without (strictly speaking)
asserting them.7 Grice’s (1989, 30–31) fullest characterization of conversational
implicature is as follows:
5. I use “convey” as a placeholder for whatever this relation is as opposed to thinking that
“convey” refers to something determinate.
6. For example, see the differences between Strandberg’s and Finlay’s views and between
Barker’s and Copp’s views.
7. Grice 1989, 28: “As one of my avowed aims is to see talking as a special case or variety of
purposive, indeed rational, behavior, it may be worth noting that the specific expectations
A man who, by (in, when) saying (or making as if to say) that p has implicated
that q, may be said to have conversationally implicated that q, provided that
(1) he is presumed to be following the conversational maxims, or at least
the Cooperative Principle;
(2) the supposition that he is aware that, or thinks that, q is required to
make his saying or making as if to say p (or doing so in those terms) con-
sistent with this presumption; and
(3) the speaker thinks (and would expect the hearer to think that the
speaker thinks) that it is within the competence of the hearer to work out,
or grasp intuitively, that the supposition mentioned in (2) is required.
The aforementioned Cooperative Principle is thus: “Make your conversational

contribution such as is required, at the stage at which it occurs, by the accepted
purpose or direction of the talk exchange in which you are engaged” (Grice
1989, 26).
Grice’s maxims are nicely summarized by Wayne A. Davis (1998, 11–12) thus:
Quality: Make your contribution true; so do not convey what you believe

to be false or unjustified.
Quantity: Be as informative as required.
Relevance: Be relevant.
Manner: Be perspicuous; so avoid obscurity and ambiguity, and strive for
brevity and order.
Here is how this works. Audiences expect speakers to adhere to the cooperative
principle. Speakers generate implicatures by either flouting or adhering to the max-
ims, given the background assumption of adherence to the cooperative principle.
An example of this, involving respecting the relevance maxim, is the
following:
A: “I am out of petrol.”

B: “There is a station around the corner.”8
B’s utterance would only accord with the maxim of relevance if the station he
refers to could provide A with petrol (it is a gas station, it is open, etc.), and so
B’s utterance generates a conversational implicature that the station around the
corner sells petrol.
or presumptions connected with at least some of the . . . maxims have their analogues in the
sphere of transactions that are not talk exchanges.”
8. Example from Grice 1989, 32.
A second example of implicature arising from flouting the relevance

maxim is:
A: “What do you think of the new boss?”

B: “It sure is sunny at this time of year, isn’t it!”
B’s utterance openly violates the maxim of relevance, and B’s openly violat-
ing the maxim enables A to infer, given the assumption that B is obeying the
cooperative principle, that B means to implicate something like that the boss is
incompetent or that B is unwilling to discuss the issue for some reason.
We should also note Grice’s (1989, 56) distinction between particular and
generalized conversational implicatures:
Particular Conversational Implicature: [A]‌n implicature is carried by say-

ing that P on a particular occasion in virtue of special features of the con-
text, cases in which there is no room for the idea that an implicature of this
sort is normally carried by saying that P.
Generalised Conversational Implicature: Sometimes one can say that the
use of a certain form of words in an utterance would normally (in the
absence of special circumstances) carry such-and-such an implicature or
type of implicature.9
Notice the contrast between:

Particular Conversational Implicature

[Implicature ≈ B’s boss is not competent.]
Generalized Conversational Implicature
C: “Some of the guests have gone.”

[Implicature = Not all of the guests have gone.]
9. Examples of generalized conversational implicatures suggested by Grice take the form

“a/an [Noun]”: “a car,” “a finger,” “an oven,” “a man.” Without any special contextual back-
ground, the use of these expressions implicates that the subject of the sentence was not (e.g.)
the owner of the car, or was not the person whose finger, or oven, is being referred to, or did
not know the man. These examples fall under the maxim of quantity, because the audience
supposes that if the subject of the sentence bore some closer relation to the car/finger/oven/
man, then the speaker would have said “his car,” “his finger,” “his oven,” “his friend/partner.”
B’s utterance only generates an implicature because of specific features of the

context, namely its coming in response to A’s specific question. By contrast,
C’s utterance generates the implicature without needing special contextual fea-
tures, because sentences of the form “Some of the [noun] have [verb]” generally
produce analogous implicatures.
Conventional implicatures are quite different from their conversational cous-
ins. Conventional implicatures are standardly treated as being part of the con-
ventional meaning of the word(s) that carry them while not being part of their
truth-conditional meaning.10 Grice (1989, 25) describes them thus:
In some cases the conventional meaning of the words used will determine
what is implicated, besides helping to determine what is said. If I say (smugly),
He is an Englishman; he is, therefore, brave, I have certainly committed myself,
by virtue of the meaning of my words, to its being the case that his being brave
is a consequence of (follows from) his being an Englishman. But while I have
said that he is an Englishman, and said that he is brave, I do not want to say that
I have said (in the favored sense) that it follows from his being an Englishman
that he is brave, although I have certainly indicated, and so implicated that this
is so. I do not want to say that my utterance of this sentence would be, strictly
speaking, false should the consequence in question fail to hold.
This passages outlines the main features of conventional implicatures, summa-

rized by Christopher Potts (2005, 11) thus:
a. Conventional Implicatures are part of the conventional meaning of

words.
b. Conventional Implicatures are commitments, and thus give rise to
entailments.
c. These commitments are made by the speaker of the utterance “by virtue
of the meaning” of the words he chooses.
d. Conventional Implicatures are logically and compositionally indepen-
dent of the at-issue entailments.
One might think Grice’s claims false of the example he uses (“therefore”) and
for some of examples commonly used in the literature (such as “but”). Potts
(2005, 13–17) suggests some perhaps better candidates for conventional impli-
catures in the form of “supplemental expressions” and “expressives” as well as
epithets or slurs, which gives us three candidate examples thus:
10. Conventional implicatures are controversial for a number of reasons. For discussion, see
Bach 1999.
“As you well know, John is cheating on his wife.”

“This fucking printer cost £200.”
“Chelsea has signed that kraut Ballack.”
The list of features outlined above, coupled with the controversy sur-
rounding putative examples of conventional implicatures such as “but” and
slur terms, brings out the fact that the category of conventional implicatures
is vexed. This is because conventional implicatures are, or would be, part of
the conventional meaning of expressions while nonetheless not affecting truth
conditions and falling within pragmatics. Though there is much to say on this
issue, I hope to sidestep it here. I aim to show that even if conventional impli-
cature has good credentials generally, there is sufficient reason to reject it as
a model of moral attitude expression.11 Thus I will proceed as if conventional
implicature is well motivated.
Having outlined Grice’s account of the three kinds of implicature, I move
to the main business of the chapter, that of determining how closely attitude
expression fits them.12 To do this I take five tests for kinds of implicatures and
apply them to attitude expression to see how closely they correspond to it.
The tests are:
(Test 1) Conversational implicatures can be indeterminate. Conventional

implicatures are not.
(Test 2) Conversational implicatures are nonredundantly reinforceable.
Conventional implicatures are not.
(Test 3) Conversational implicatures are nondetachable. Conventional
implicatures are detachable.
(Test 4) Conversational implicatures are cancelable. Conventional implica-
tures are not cancelable.
(Test 5) Conversational implicatures are calculable. Conventional implica-
tures are not calculable.13
11. For discussion of conventional implicature generally, see Bach 1999; Potts 2005, 2007.
12. Judgments about the results of the tests are not always strong and are susceptible to
manipulation from (e.g.) tonal stress (with the possible exception of test 3). It is thus impor-
tant to stress that the evidentiary basis for choosing one implicature model over another is
the result of how they fare across the tests considered as a whole.
13. Four of these are from Grice’s original discussion, while one (test 2) comes from Sadock
1991, 374. Following Jerrold M. Sadock (1991, 367), I ditch Grice’s “part of conventional
meaning” and “not what is said but by the saying of what is said” tests.
T E S T 1 : D E T E R M I NAC Y
Grice claims that because there are multiple ways to make what a speaker says
consistent with the cooperative principle, a conversational implicature can be
highly indeterminate. Grice’s point only really applies to particular conversa-
tional implicatures, though. Taking the earlier example:

There is a fairly high degree of indeterminacy in the implicature generated by B’s

utterance due to the different ways of making B’s utterance fit with the cooperative
principle (B could be implicating that the boss is nearby, that the boss is incom-
petent, that the boss is a tyrant, or simply that he or she is unwilling to discuss the
issue for whatever reason). This is typical of particular conversational implicatures.
Moral utterances, by contrast, convey fairly determinate information about
the speaker’s desire-like attitudes. Utterances such as
O: “Waterboarding terror suspects is wrong.”
convey negative desire-like attitudes against torturing, specifically negative

reactive attitudes, such as resentment toward those who torture. In this way,
then, attitude expression differs from particular conversational implicature by
being fairly determinate across contexts.
Those who favor a generalized conversational implicature model of attitude
expression have a ready explanation of this, given that generalized conversational
implicatures are highly determinate, a fact connected to their being generated
without special contextual setup. (For example, the word “some” carries the deter-
minate implicature not all, and it does so across a wide range of contexts.) Thus the
determinacy of attitude expression fits well with a GCI theory also.
While it was important to introduce particular conversational implicatures
above to explain the Gricean theory, it is unlikely that moral attitude expres-
sion is a form of particular conversational implicature. Moral utterances convey
information about a speaker’s attitudes that is fairly determinate and do so in
a way that is much less context dependent than it would be on a particular
conversational implicature model. I will therefore focus only on conventional
implicature and generalized conversational implicature from here.
Because conventional implicatures are part of the conventional meaning
of the expressions that carry them, conventional implicatures are also highly
determinate. For example, these utterances
“Bill is rich but honest.”

convey determinate implicatures, namely that being rich is likely to reduce

one’s likelihood of being honest and a derogatory attitude toward German
people. Thus a conventional implicature model of attitude expression can also
well explain the determinacy of the attitude expression generated by O’s moral
utterance. Thus the conventional implicature understanding of attitude expres-
sion fares well on the determinacy criterion.
T E S T 2 : R E I N FO R C E AB I LI TY
As noted above, the main contrast between conventional implicatures and

conversational implicatures is that conventional implicatures are part of the
conventional meaning of utterances while conversational implicatures are not.
As noted by Jerrold M. Sadock (1991, 374), this explains the fact that one can
felicitously make the implicated information explicit with a conversational
implicature but not with a conventional implicature, where doing so generates
redundancy.14 Compare these two pairs of cases:
Conventional Implicature
# “Chelsea has signed that kraut Ballack. I have derogatory attitudes toward
German people.”15

# “This fucking printer cost £200. I am angry about the printer.”
Conversational Implicature
“Some of the guests have gone.”
“Some of the guests have gone. Not all of them.”
14. One complication is that this contrast is most easily seen when comparing conventional
implicatures with particularized conversational implicatures. To avoid confusion, I do not
reintroduce particularized conversational implicatures in the main text.
15. One oddity with this sentence is that the person uses a nonslurring “German people”
after using a slur referring to the same group.
“You can have wine or beer.”

“You can have wine or beer. You can’t have both.”
What of moral attitude expression? Despite moral utterances conveying infor-

mation about the speaker’s desire-like attitudes, it does not seem redundant to
make these attitudes explicit.

O: “Waterboarding terror suspects is wrong. I’d resent anyone who did so and
feel guilty if I did.”16
I concede, however, that one might think that O’s second utterance is somewhat
redundant.17 And some might think that there is at least some redundancy in the
reinforcement of the generalized conversational implicature above (a phenom-
enon that is unsurprising given that generalized conversational implicatures often
seem to be part of the meaning of the expressions that trigger them).18 In light of
these issues, I make only the weaker claim that, with respect to reinforceability,
even if there is no clear winner on this test, attitude expression seems more like
generalized conversational implicature than conventional implicature. Following a
moral utterance by making an explicit claim about one’s desire-like attitudes seems
less redundant than reinforcing a conventional implicature and no more redun-
dant than reinforcing a generalized conversational implicature.
T E S T 3 : N ON DE TAC HAB I LI TY
The third test is a high degree of “nondetachability.” An implicature is detachable

when the speaker can assert the same at-issue proposition without generating
the implicature (holding the speaker’s tone and manner fixed).19 Generalized
conversational implicatures are highly nondetachable, whereas conventional
implicatures are detachable, because they are carried either by expressions
16. I grant that intuitions about redundancy are liable to be affected by the level of specificity
in the speaker’s claim. At least one alternative to O’s second utterance—“Waterboarding ter-
ror suspects is wrong. I’m against it”—sounds more redundant than the claim about guilt and
resentment in particular. This is one place where a piecemeal focus on individual proposals
for the relevant implicatum would be beneficial, though that would be too much to do here.
17. This test is particularly vulnerable to tonal focus. For discussion, see footnote 18 above.
18. Someone suggested this point to me and I have subsequently forgotten whom. Many
thanks, and apologies, to that person.
19. Grice (1989, 39) points out the need for the parenthetical clause.
that can be omitted without affecting the at-issue proposition (as in the case of
expressives and supplemental expressions) or by the choice of one word over
another that lacks the implicature.20 Take sentences S1 and S2:
S1: “Chelsea has signed that kraut Ballack.”

S2: “Chelsea has signed that German Ballack.”
The conventional implicature in S1 arising from “kraut” is detachable, because

there is an alternative way of stating the at-issue proposition (that Chelsea has
signed Ballack, a German) that does not generate the implicature.21
For a generalized conversational implicature, take the example:
B: “Some of the guests have gone.”
B’s utterance generates an implicature in this case. The implicature is unaffected

by the form of B’s utterance. There are no means by which B can state the same
proposition without generating the implicature.
Is attitude expression nondetachable? It seems highly nondetachable. Take O’s
reply in this exchange:
B: “Waterboarding is permissible, it’s the only means to prevent death to

innocents.”
This conveys that O has negative desire-like attitudes toward torturing terror sus-
pects. The same attitude expression occurs with
O: “It’s wrong to waterboard terror suspects.”
and with any other way of stating the same at-issue proposition. Thus moral atti-
tude expression, like conversational implicature, is highly nondetachable. With
respect to detachability, attitude expression fits conversational rather than conven-
tional implicature.
20. Lycan 2008, 198.

21. It is common for people to use “say the same thing” where I am using “state the same
at-issue proposition.” But both “say” and “the same thing” are vague, and it is difficult to
use them without seeming to contradict the analysis of conventional implicatures (as part
of the conventional meaning of expressions that carry them). Thus for this reason I choose
this other terminology from Potts 2005. Others, such as Davis (2010, 3), get around this by
construing “say” narrowly. But this may still cause confusion.
Copp, who defends a conventional implicature model of attitude expression,

objects. He grants, “One might wonder why there is no familiar predicate that stands
to wrong as ‘Italian’ stands to ‘wop,’ ” by which he means an implicature-detaching
equivalent term (Copp 2009, 188). But Copp thinks a conventional implicaturist
can establish the detachability of attitude expression in the following way. He or
she can claim that it is only a contingent truth that there is no attitudinally neutral
equivalent term in English for “morally wrong” and that we can construct such
neutral equivalents using scare quotes or descriptions such as “capital punishment
has the property that would be ascribed by saying it is ‘morally wrong.’ ” In earlier
work Copp (2001, 36) suggested the following inverted commas proposal: “On
this model, the amoralist could have said ‘Cursing is ‘morally wrong’,’ placing ‘mor-
ally wrong’ in inverted commas, or she could have said ‘Cursing is so-called mor-
ally wrong’; either method would have detached the implication that she has a
policy against cursing.”
One problem, pointed out by Finlay (2005, 14), is that scare quotes do not
do what Copp wants, namely merely preventing moral terms from conveying
desire-like attitudes. Inverted commas sometimes introduce reference to what
some (contextually relevant) group of people believe (so, for example, the speaker
is merely reporting that most people around here believe the action to be wrong).22
Other times inverted commas signal that one thinks that the predicate has no
extension or that the subject does not exist. Examples of this are the hard deter-
minist, who says “We’re all ‘free’ to vote in the election,” or the atheist, who asks her
religious friend “Who or what made ‘God’?” Thus, as Finlay observes, it is not the
amoralist but the error theorist who would use inverted commas in this way, and
she would do so to signal her view that there is no such thing as moral wrongness.
Copp’s later suggestion for why attitude expression is detachable is that a
speaker can use the complex description “has the property that would be
ascribed by saying that it is ‘morally wrong.” Though I am not sure, I doubt that
this achieves detachability. The complex description will likely be interpreted as
the claim that the action is wrong, with attendant attitude expression, or alter-
natively (more likely) the use of the complex description would be interpreted
as a way of using obscurity of expression to signal either that this action is
simply one that other people call wrong or that the speaker rejects the presup-
position that anything is morally wrong.
To illustrate the latter possibility, suppose I say “Chocolate cake has the prop-
erty that would be ascribed by saying that it is nice.” I think someone would
interpret this as not saying that chocolate cake is nice but rather as saying
something else, such as that other people would like it. Alternatively but less
likely, it will be interpreted as the claim that the cake is nice, albeit expressed in
22. This fits with Copp’s formulation using “so-called.”

an unusual way (in which case any attitude expression that would occur with
“Chocolate cake is nice” will also occur).
In light of these problems, it seems clear that attitude expression does not
exhibit the high degree of detachability of a conventional implicature. Just to
emphasize this, compare the difficulties Copp’s proposal has with how obvious the
implicature-detaching alternatives to prototypical conventional implicatures are.

“This printer cost £200.”

“Chelsea has signed that German Ballack.”
Conventional Implicatures are easily detached by omitting expressions or by using

an alternative term.23 This brings out what is problematic in Copp’s proposal—it
is so unobvious how exactly Copp’s definite description would be interpreted by
audiences. Thus, pace Copp, on the test of detachability, attitude expression does
not fit conventional implicature but rather conversational implicature.24
Before moving on, I briefly pause to strengthen my case against the conven-
tional implicature model of attitude expression by examining the speaker orien-
tation of conventional implicatures. (Those already convinced that conventional
implicature is the wrong model can jump to the start of the next section.)
Copp (2009, 188) asks, “One might wonder why there is no familiar predicate
that stands to wrong as ‘Italian’ stands to ‘wop.’ ” In keeping with Copp’s view,
this assumes that moral terms are conventional implicature carrying terms. But
the speaker orientation of conventional implicatures, namely, the fact that the
commitments are made by the speaker of the utterance,25 provides compelling
evidence that moral terms do not express attitudes through conventional impli-
cature. To see why, compare the following belief attributions:
“Bachmann thinks that homosexuality is wrong.”

“Bachmann thinks that this fucking printer cost £200.”
“Bachmann thinks that Chelsea has signed that kraut Ballack.”
23. For discussion of this, see, for example, Hay 2013.

24. This is not to suggest that other theories cannot explain this feature of moral terms.
A view that holds that there is a semantic connection between moral terms and attitudes
would also fit the detachability data. Such views are outside the scope of this chapter, though.
Thanks to Mike Ridge here.
25. Term from Potts 2005.
As pointed out by Ryan Hay (2013) and Mark Schroeder (2009, 301), with the
moral belief attribution it is obvious that there is no commitment on the part of
the speaker with respect to homosexuality. We learn nothing of the speaker’s views
about homosexuality or his desire-like attitudes more generally. By contrast, in
the belief attributions that deploy conventional implicature terms, it is clear that
the commitments are taken on by the speaker. We learn about the speaker’s (and
not Bachmann’s) attitudes toward the printer and toward German people. The
speaker orientation of conventional implicatures is thus at odds with how moral
terms behave in belief attributions. The same is true of indirect quotations:
“Bachmann said that homosexuality is wrong.”

“Bachmann said that this fucking printer cost £200.”
“Bachmann said that Chelsea has signed that kraut Ballack.”
These cases are strong evidence that attitude expression does not occur via
conventional implicature. Thus in response to Copp’s question—why there is
no familiar predicate that stands to wrong as “Italian” stands to “wop”—we
should reject the assumption that “wrong” is in fact like “wop.” “Wop” (like all
conventional implicatures) is speaker-oriented. “Wrong” is not.
Someone might try to defend Copp’s view by building on a recent discussion
by Hay and distinguishing between slurs (such as “wop”) and general pejora-
tives (such as “jerk”). He or she might then argue that general pejoratives are
the correct model for moral terms, because general pejoratives, unlike slurs,
behave like moral terms in belief attributions and indirect quotations.
The trouble with this suggestion is that even if general pejoratives are a good
model for moral terms, this would not vindicate the conventional implicature
view. This is because general pejoratives are not plausible candidates for a con-
ventional implicature treatment precisely because they are not speaker-oriented
and lack “implicature”-detaching equivalent terms. Thus this suggestion would
be of no help to Copp.26
Having rejected particular conversational implicature and conventional
implicature, the rest of the chapter examines the claims of (generalized) con-
versational implicature to be the vehicle for attitude expression.
T E S T 4 : C A N C E LAB I LI TY
The fourth test to consider is Grice’s test of cancelability. Sometimes this is glossed
as simply that some implicatures “can be canceled.” But this leaves the idea too
26. Copp (2009, 186–187) himself flags this possibility. See also Hay 2013.
imprecise to do much work. Finlay (2005, 14) provides a more precise formula-

tion of cancelability thus: “An implicature is explicitly cancellable iff a speaker can
(without abusing the meaning of his words) legitimately deny that he means to
express that content; it is contextually cancellable iff there are contexts of use in
which the utterance of that sentence wouldn’t even seem to suggest that content.”
An implicature is cancelable when a speaker can signal that she does not
accept the (otherwise) implicated proposition without thereby misusing words
or contradicting herself. If one tries to withdraw a conventional implicature,
one’s utterance is incoherent or contradictory:
“As you well know, John is cheating on his wife.”

# “As you well know, John is cheating on his wife. Not that you knew that.”

# “Chelsea has signed that kraut Ballack. I have nothing against German
people.”
By contrast, generalized conversational implicatures are not part of the mean-

ing of the sentence (/word/expression), so a speaker can withdraw them with-
out having misused a word or contradicting herself. A speaker might say, “The
prize is in the garden or the garage.”27 Her use of “or” generates an implicature
that she does not know which is true (because one would expect her to assert
the more informative proposition if she could). But one can felicitously cancel
the generalized implicature by adding “but I’m not telling you which” without
being guilty of having misused the word “or” or thereby contradicting oneself.
Can attitude expression be canceled? It is difficult to decide what to make
of examples of putative cancelation, and one can see the temptation to think
that they really involve inverted-comma uses of “morally wrong.” Nevertheless,
attitude expression is cancelable, but to see this we need to clear away some dis-
tractions that often afflict discussions of cancelability. I will first highlight these
problems and then show that when we clear these up we can find examples
where cancelation is coherent (albeit highly unusual).
Finlay (2005, 15) claims that it seems “clear that we can appropriately utter
moral judgements without attitudinal expression” and that evidence of this is
the example of the amoralist
who sincerely makes assertions about the moral value of things but doesn’t
subscribe to those moral standards herself and doesn’t express approval (etc.)
27. Example from Grice 1989, 44–45.

by her moral speech acts. Attitudinal content can be cancelled explicitly, if

she merely explains that she is an amoralist, or that she is contemptuous
or indifferent towards morality. It is contextually cancelled if her audience
already knows of her amoralism (particularly if they themselves are amoral-
ists, rendering amoralism contextually uncontroversial). (My italics)
One minor issue with Finlay’s formulation is its reference to “those” moral stan-
dards. This is potentially misleading, because it suggests that the amoralist is
only indifferent to some particular set(s) of moral standards. Finlay’s example
needs to be one in which the putative amoralist lacks desire-like attitudes con-
nected to the set of moral standards that he or she regards as correct. This is nec-
essary to make clear that the example is one where the person is an amoralist
(he or she judges that things are morally wrong while being indifferent to this
fact) rather than either an error theorist or simply one who rejects some moral
standards (namely those of others).
A second problem to deal with is highlighted by Copp’s claims against the
cancelability of attitude expression. Copp (2009, 187) claims: “It is ‘semanti-
cally inappropriate’ to use the term [‘morally wrong’] in making a moral judge-
ment unless one disapproves of actions that have the property of being wrong.”
(Before coming to the main issue, Copp’s use of “semantically” is liable to mis-
lead by suggesting that he means that a person who uses “wrong” in a moral
utterance but who does not disapprove of wrong actions is violating the truth
conditional element of the term’s meaning. I presume that Copp calls this puta-
tive infelicitous use of “wrong” a kind of semantic impropriety, because the
implicature is a conventional element of the word.)
Moving to Copp’s cancelability claim itself, to assess it we need to know more
about what disapproval is (as Copp himself points out, it is easy to think of dis-
approving of something as believing that it is morally wrong). Copp (2009, 170),
however, is deliberately and understandably noncommittal about the nature of
the desire-like attitude implicated by moral terms. A downside to proceeding
in this fashion is that it makes cancelability harder to test. If we want to test
whether attitude expression is cancelable and do not specify the relevant atti-
tude, we must instead describe the hypothetical cases as ones in which either
(a) the agent lacks all desire-like attitudes or (b) the subject lacks whatever
desire-like attitude is standardly connected to the moral utterance (where this is
to be read de dicto).
Alternative (b) is so vague that it is doubtful that our intuitions about cases
thus described should carry any weight. Alternative (a) is problematic, because it
would involve someone who lacks any desire-like attitude but who is nonetheless
motivated to engage in moral discussion. This might play a role in our finding
putative cancelation bizarre and thus being tempted to treat it as a contradiction.
To avoid this problem we should try to fix on plausible candidates for the
desire-like attitudes conveyed by moral utterances and test whether speakers
can cancel these. The desire-like attitudes standardly conveyed by moral utter-
ances declaring actions to be morally wrong include, at least, the disposition to
resentment toward others and guilt toward oneself for performing such actions.
Therefore these are attitudes that we can use in trying to determine whether
attitude expression is cancelable.
A third issue for assessing cancelation is evident in the following example:
E: “It would be wrong to do that. I don’t care if you do though.”
It is unclear what to make of this. If E has no negative desire-like attitudes about

it, then we would be reluctant to ascribe to her the judgment that the action is
wrong. But one thing that might be at work here is that without any background
context being described, we are puzzled as to why E is making any moral utter-
ance if, by hypothesis, she lacks the relevant desire-like attitudes.
Plausibly, our puzzlement is explained by two facts. First, most people seem-
ingly care about morality (amoralists are rare or at least discreet). Second, a
more general fact about claims that invoke standards, namely, that unless some-
thing in the context signals otherwise, a person’s engaging in discussion about
what is required, permitted, prohibited, or recommended by some standard
makes it likely that he or she cares about conformity to that standard. If that is
even approximately right, then filling in the context to make clear why the per-
son is making a moral utterance, despite lacking the corresponding desire-like
attitudes, should help make cancelation more coherent.
Here, then, is an example designed to take care of these kinds of considerations.
Suppose F is on the board of a financial services company, and one day the board
is discussing some tax avoidance scheme, one that exploits a regulatory loophole
in profitable but socially harmful ways. Suppose one of F’s colleagues has pointed
out to her that it would be morally wrong to exploit this loophole, and F replies:
F: “You say that it’s morally wrong and you’re absolutely right about that. But
I don’t care about what’s morally right and wrong; I just want to make as much
money and have as much fun as possible. It’s not illegal. So lets do it and keep
it quiet. Anyone who passed up this kind of opportunity is a sucker.”28
To my ears, F’s utterance is more coherent and intelligible than E’s. The extra
details of the example better enable us to understand why F is talking about the
28. This speaker does not talk explicitly about the reactive attitudes. Explicitly referring to
them tends to make the speaker sound too much like a philosopher.
moral status of things despite her professed indifference.29 I think this should
make us confident that attitude expression is cancelable even if it takes some
work to do so in most contexts (and more work than in the case of standards,
such as those of a particular religion or legal standards, to which indifference is
more widespread).
I conceded above that it is difficult to determine whether attitude expression
is coherently cancelable. My approach to making this more plausible has been
to show how some discussions of cancelability, such as Finlay’s and Copp’s, have
misdescribed what is involved in cancelation or procededed in a way can make
cancelability seem more implausible or problematic than it really is. I have also
suggested that when we formulate cases of putative cancelability without the
problematic features identified above, we find speakers can cancel the presence
of relevant desire-like attitudes without seemingly contradicting themselves
or generating incoherence, thus matching conversational implicature. And
for reasons outlined by Strandberg (2011), if moral attitude expression were
a matter of generalized conversational implicature, then it is highly likely that
we would find exactly this pattern of data, namely, that cancelation is possible
though highly unusual.
T E S T 5 : C A L C ULAB I LI TY
Conversational implicatures must be capable of being calculated on the basis

of what the speaker asserts. In asserting that P, a speaker conversationally
implicates that Q only if the audience is able to work out that the presumption
that the speaker believes Q is necessary to make his saying that P consistent
with observing the cooperative principle. Thus a conversational implicature
is generated only if the audience is able to work out what is being implicated
by applying the cooperative principle and their background knowledge to
the situation (included in which is the meaning of the words the speaker
used).30
If moral attitude expression is a form of conversational implicature, it must
then be possible to calculate these implicatures on the basis of what is said.
One such suggestion for how such implicatures would be calculable stems from
29. Film and literature present us with even more vivid examples of agents who seem to
know the right and the good but lack corresponding desire-like attitudes.
30. See the second section for a spelled-out version of the inference. See also Potts 2007, 669.
With generalized conversational implicature, because such information is carried by the use
of certain forms of words, in general this means that the implicatures might not need to be
calculated in this way. This is no problem. All that is important is that they are calculable.
the remarks made above, namely, that a speaker’s engaging in discussion about
what is required, permitted, prohibited, or recommended by some standard
makes it likely that she cares about conformity to that standard.
A similar proposal comes from Strandberg (2011, 106), who defends a gener-
alized conversational implicature view of attitude expression and who suggests
that attitude expression is calculated as follows:
The person has uttered a sentence to the effect that certain actions are
wrong. There are no grounds for believing that she is not observing the
cooperative principle; that is, there are no reasons to believe that she does
not try to contribute to the moral conversation in a way that fulfills its
mutually accepted purposes, which, among other things, is to influence
behavior. . . . Given these assumptions, she would not have uttered the sen-
tence unless she wants that such actions are not carried out. . . . Therefore
she wants that such actions are not performed.
This is plausible, especially if we focus on conversations where individuals are

jointly deliberating about what to do (individually or collectively). Take the fol-
lowing exchange:
b: “I’m not sure whether I should invest in X or Y.”
a: “It’d be morally wrong to invest in X.”
b: “Ok, I’ll invest in Y.”
a: “No. You should invest in X.”
b: “But you just said it was morally wrong to invest in X.”
a: “It would be wrong to invest in X, and you should.”
b: “You’re trying to get me to do what you think is morally wrong?”
a: “Yes!”
A is (initially) uncooperative in this exchange. The original utterance suggests
that B should not invest in X (even if this implicature is canceled later).31 This
suggests that Strandberg’s suggestion for how attitude expression can be cal-
culated fits the data here well. On the calculability test, then, attitude expres-
sion fits a conversational implicature model, as there seem to be numerous
ways one could calculate an implicature (that the speaker has the relevant
negative desire-like attitude) from their asserting that something is wrong.32
31. A plausible story for why this is so is that we take moral requirements to be overrid-
ing. For discussion of this see my ‘Moral Judgement Internalism: Its Scope, Status and
Significance’ (manuscript).
32. Finlay (2004, 2005) has a different plausible theory about how the implicature is calcu-
lable. I only discuss Strandberg’s purely for reasons of space.
S O ME P R OB L E M S FO R THE G CI V I E W
A N D A N A LT E R NATI V E
As I hope to have shown, there is much plausibility in the view that attitude expres-
sion is a kind of conversational implicature. If any form of implicature theory is
the truth about how moral utterances convey desire-like attitudes, then GCI is it.
However, I close on a more pessimistic note by highlighting a collection of related
problems for this type of view and by suggesting an alternative pragmatic story
capable of “stealing its thunder” by explaining more while postulating less.
As argued by Dorit Bar-On and Matthew Chrisman (2009), the conversa-
tional implicature view relies on a conversational context for attitude expres-
sion to take place.33 One question this view therefore leaves unanswered is the
connection between moral thought and desire-like attitudes.34 And this is quite
an important omission. If one of the motivations for an implicature view is a
close connection between people’s desire-like attitudes and their moral utter-
ances, one might think that was explained in turn by the connection between
people’s desire-like attitudes and their moral judgments.35 But on this later issue
the conversational implicature view is itself silent.36
I move on now to a different problem for the conversational implicature
view—that it cannot account for some ways moral utterances convey desire-like
attitudes. In particular, it cannot apply to moral utterances that convey such
desire-like attitudes as the following:
Eavesdropping Cy: Trixi is struggling with a difficult dilemma. Her friend

Alma must decide whether to marry Ellesworth, the honorable and sen-
sible move, or convince Bullock to leave his wife. Wanting privacy to think
about the matter, Trixi books a room at Cy’s lodge. Trixi believes herself to
be alone. But she is mistaken, and Cy overhears her say aloud: “It would be
wrong for Alma to steal Bullock away.”
33. Strandberg’s (2011) description of how one calculates a conversational implicature is

explicit on this point.
34. Bar-On and Chrisman 2009, 153.
35. For an up-to-date account of this issue, see Björklund et al. 2012.
36. Stranberg (2011, 115) suggests in a footnote that “it is plausible to think that conversational
implicatures often take place in thought. Roughly put, in our thinking we follow the conversa-
tional implicatures of the utterances we would have used to communicate our thoughts.” This
proposal is only mentioned briefly, but I think the details of implicature prevent its being suc-
cessfully applied to the thought case. I think a better route would be to say that the connection
in the thought case is explained by something else and that this separation can be motivated.
For discussion of conversational implicatures and thought, see Speaks 2008.
It is clear that Cy learns something about Trixi’s desire-like attitudes toward

the action of stealing Bullock away.37 Can this be explained by a general conver-
sational implicature model? The answer is “no.” Remember how Grice (1989,
30–31) characterizes conversational implicature:
A man who, by (in, when) saying (or making as if to say) that p has impli-
cated that q, may be said to have conversationally implicated that q, pro-
vided that
(1) he is presumed to be following the conversational maxims, or at least
the Cooperative Principle;
(2) the supposition that he is aware that, or thinks that, q is required to
make his saying or making as if to say p (or doing so in those terms) con-
sistent with this presumption; and
(3) the speaker thinks (and would expect the hearer to think that the
speaker thinks) that it is within the competence of the hearer to work out,
or grasp intuitively, that the supposition mentioned in (2) is required.
The issue is whether (1)–(3) hold in the case of Eavesdropping Cy. It is pretty
clear that (1) and (2) do not. There is no reason to expect people to adhere to
the cooperative principle when, by their lights, they are not communicating
with anyone. So (1) and therefore (2) look doubtful. More importantly, (3) is
clearly not satisfied. Trixi has no beliefs about what her hearer can work out.
She believes that she is alone and may not believe anything about Cy’s com-
petencies. Thus implicature cannot be the means by which Cy learns of Trixi’s
desire-like attitudes in this case.38
How might an implicature theorist reply? For the reasons just given, there is
no plausibility in treating this as a case of implicature. The only plausible option
response is to carve off cases like Trixi’s utterance and give a separate account
of how attitudes are conveyed therein. The trouble with such a response is
that it leaves a hostage to fortune if an alternative view explains what occurs
in both Eavesdropping Cy cases and standard conversational cases. I present
37. Of course Trixi might also have the positive desire-like attitude of wanting Alma to steal
Bullock away. My description of the case is not supposed to suggest that she only has a nega-
tive desire-like attitude toward the action.
38. On an understanding of implicature that makes a speaker’s intentions central, such as
Davis 1998, the problem is even worse, given that Trixi has no intention to be heard by anyone
(and may indeed have the opposite intention). However, for reasons given by J. M. Saul (2001,
2002), it is controversial (at least) whether speakers’ intentions matter to the generation of
conversational implicatures. Thus I mention this claim only for those who think that inten-
tions matter to implicature and to show that my main argument does not rely on this claim.
such an alternative view shortly, after outlining what I take to be the funda-
mental problem for GCI that underlies the case of Eavesdropping Cy.
This more general problem is that with conversational implicatures the
speaker’s utterance act plays a substantial role in their production, whereas this
is not true of the expression of desire-like attitudes by moral utterances. To see
why, take the following two cases:
Utterance: A and B often give C, D, E, and F a ride home from parties. A–F all
go to a party. At the end of the party B says to A: “C and D need a ride home.”
Implant: A has a device implanted in B’s brain that reliably informs her
when the following judgment is formed by B: <C and D need a ride home>.
A consults the device and sees that B has this belief.
B’s utterance generates an implicature q, namely, (B believes that) E and F

do not need a ride home. By contrast, A merely learning via the implant that B
has a judgment with the same content is not good evidence that E and F do not
need a ride home (or that B believes this). After all, B might believe that C and
D need a ride home from the party as well as believing that E and F need a ride
home from the party. Or B might not have considered whether E and F need a
ride home. Thus A learns something from B’s utterance “C and D need a ride
home” that she would not learn from the fact that B believes <C and D need a
ride home>.
The point here is that an utterance that p can generate the implicature and
provide good evidence that q (or that the speaker believes that q) even where
learning that the person has a belief with the same content would be little
or no evidence that q (or that the person believes that q). This is because the
implicature is generated specifically by the speaker’s act of asserting p, given
the presumption that the speaker is adhering to the cooperative principle.
Things are different, however, when it comes to learning about someone’s
desire-like attitudes. For compare these two cases:
Utterance (moral): G says to H, “Torture is wrong.”

Implant (moral): H has a device implanted in G’s brain that reliably informs
H when the following judgment is formed by G: <torture is wrong>. H
consults the device and sees that G has this belief.
G’s utterance to H conveys that G has desire-like attitudes against torture. But
in contrast with the first pair of cases, the information from the implant in the
final case gives H good (though defeasible) evidence that G has desire-like atti-
tudes against torture.
My claim is not that the implant gives H equally good evidence of G’s
desire-like attitudes as an utterance of “torture is wrong” would. Rather, my
claim is that in the nonmoral case the presence of the judgment is very weak
evidence, if any, for what the corresponding assertion would implicate, whereas
in the moral case the presence of the judgment is good (even if defeasible) evi-
dence for what the corresponding assertion would convey.
This difference between the cases points to a problem for GCI. For it seems
to be a general feature of conversational implicature–generating utterances that
one learns something from the utterance that p that one cannot know merely
from the presence of the corresponding belief that p. With implicatures, the act
of uttering that p has an important role over and above being a mere sign of the
presence of the belief that p in the speaker. But this does not seem to be true
in the moral case, where the utterance act seems much less important. As we
see in the case of Eavesdropping Cy, what seems important in the moral case is
merely determining the presence of the judgment.
I claimed above that the GCI view would be vulnerable if an alternative view
could explain what goes on in both cases like Eavesdropping Cy and standard
conversational cases. A simpler view that I think is able to do this consists of
the following three claims:
The Simple Pragmatic Story (SPS)
(1) People commonly have desire-like attitudes in accordance with their

moral judgments.
(2) People’s moral utterances voice their moral judgments accurately, at
least for the most part.
(3) (1) and (2) are common knowledge.
Three brief points are in order here. First, there will be various explanations
of why (1) is true. It could be explained by a widespread desire to do the
right thing or some less fetishistic story, as suggested by Sigrún Svavarsdóttir
(1999, 198–199). Second, it is important to note that (1) is not a form of
“moral judgment internalism.” Rather, (1) is a generalization that moral
judgment internalism might be used to explain. Third, someone who favors
a conventional implicature story of moral utterances might claim that his
view is well placed to explain (1). I doubt this but sadly lack the space to go
into this issue here.
If (1)–(3) are true, moral utterances provide strong defeasible evidence of
a speaker’s desire-like attitudes. So speakers convey attitudes by their moral
utterances, because audiences simply infer the presence of the attitudes on the
basis of the utterance (and their background beliefs).
Despite resting on three weak and highly plausible claims, the SPS shares all of
the features of the conversational implicature view that seem to correspond to atti-
tude expression. The SPS thus has an equal share in all of the positive evidence for
that view. It also avoids certain problems. For example, it can explain how utter-
ances such as Trixi’s convey information about desire-like attitudes without need-
ing to somehow treat this as a kind of conversational implicature or by appeal to
a separate mechanism for what goes on in conversations. It does this by simply
relying on the fact that speakers tend to voice their own judgments accurately both
when communicating with others and when thinking and talking aloud and these
judgments are at least commonly accompanied by desire-like attitudes.
Strandberg (2011, 108) in fact suggests that the truth of something like (1)–
(3) from the SPS is (part of) the explanation of why moral utterances carry
generalized conversational implicatures. And it is hard to imagine a plausible
story of how moral utterances generate generalized conversational implicatures
that does not incorporate the subclaims of the SPS.
The pertinent question then is do we have any reason to go further and adopt
the GCI view over the SPS? Constraints of space preclude a full comparison of
the two views here, but I think the answer is “no” for at least the following rea-
sons. The SPS is plausibly going to be part of any explanation that a GCI view
offers of how such implicatures arise, the SPS is supported by all of the same
evidence that supports the GCI view, and the SPS explains more and gives a
unified explanation that also encompasses cases like Eavesdropping Cy while
incurring fewer commitments.
There are, then, a number of reasons for choosing the SPS over the GCI view.
By contrast, there seem to be no reasons for choosing the GCI view over or in
addition to the SPS.
CO N C L U S I ON
In this chapter I have considered the extent to which attitude expression by

moral utterances corresponds to implicature. I drew three main conclusions.
First, moral utterances do not convey desire-like attitudes through either par-
ticular conversational implicature or conventional implicature. Second, there is
a good fit between generalized conversational implicature and attitude expres-
sion, giving this theory a strong claim to being the right model of attitude
expression. But third, a simpler pragmatic story seems to share all of the virtues
of the GCI theory while avoiding its problems and while committing us to less,
so we do not yet have reason to adopt GCI over SPS.39
39. Many thanks to Matthew Chrisman, Terence Cuneo, Alex Gregory, Allan Hazlett, Chris
Heathwood, Jonas Olson, Mike Ridge, Debbie Roberts, Caj Strandberg, anonymous referees,
and audiences in Edinburgh, Gothenburg, and York.
References
Ayer, A. 1952. Language, Truth, and Logic. New York: Dover.
Bach, K. 1999. “The Myth of Conventional Implicature.” Linguistics and Philosophy 22
(4): 327–366.
Barker, S. 2000. “Is Value Content a Component of Conventional Implicature? Analysis
60 (267): 268–279.
Bar-On, D., and M. Chrisman. 2009. “Ethical Neo-Expressivism. In Oxford Studies in
Björklund, F., G. Björnsson, J. Eriksson, R. Olinder, and C. Strandberg. 2012. “Recent
Work on Motivational Internalism.” Analysis 72 (1): 124–137.
Blackburn, S. 1998. Ruling Passions: A Theory of Practical Reasoning. New York: Oxford
University Press.
Boisvert, D. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89
(2): 169–203.
Philosophy and Policy 18 (2): 1–43.
———. 2009. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies in
Davis, W. 1998. Implicature: Intention, Convention, and Principle in the Failure of Gricean
Theory. Cambridge: Cambridge University Press.
———. 2010. “Implicature.” In The Stanford Encyclopedia of Philosophy, ed. E. Zalta.
http://plato.stanford.edu/archives/win2010/entries/implicature/.
8 (3): 205–223.
———. 2005. “Value and Implicature.” Philosophers’ Imprint 5 (4): 1–20. http://bit.ly/
uxrgbk.
Fletcher, G. 2014. “Moral Judgement Internalism: Its Scope, Status and Significance”
(manuscript). Accessed April 22, 2014. <https://edinburgh.academia.edu/
GuyFletcher>.
Grice, H. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press.
Language.” European Journal of Philosophy 21 (3): 450–474.
Potts, C. 2005. The Logic of Conventional Implicatures. Oxford: Oxford University Press.
——— 2007. “Into the conventional-implicature dimension.” Philosophy Compass 4
(2): 665–679.
Ridge, M. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116 (2): 302–336.
———. 2007. “Ecumenical Expressivism: The Best of Both Worlds?” In Oxford Studies
in Metaethics, vol. 2, ed. R. Shafer-Landau, 51–76. Oxford: Oxford University Press.
Sadock, J. 1991. “On Testing for Conversational Implicatures.” In Pragmatics: A Reader,
ed. S. Davis, 365–376. Oxford: Oxford University Press.
Saul, J. 2001. “Critical Study of Wayne Davis, Implicature: Intention, Convention, and
Principle in the Failure of Gricean Theory.” Noûs 35 (4): 630–641.
———. 2002. “Speaker Meaning, What Is Said and What Is Implicated.” Noûs 36
(2): 228–248.
Schroeder, M. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119 (2): 257–309.
Speaks, J. 2008. “Conversational Implicature, Thought, and Communication.” Mind and

Language 23 (1): 107–122.
Phenomenological Research 84 (1): 87–122.
Svavarsdóttir, S. 1999. “Moral Cognitivism and Motivation.” Philosophical Review 108
(2): 161–219.
9
Pure versus Hybrid

Expressivism and the
Enigma of Conventional
Implicature
STEPHEN BARKER n
Conventional implicatures are meanings encoded by particles like even, but,

nevertheless, therefore, and so on. Take sentence (1) below with even:
(1) Even Granny got drunk.
The truth condition of (1)—its said content—is that Granny got drunk. The
contribution of even is to indicate or signal a probability scale relative to some
contextually given class of individuals in which Granny is the least expected to
be drunk and all got drunk. The dual-content form of (1) is then:
Said content: Granny got drunk.

Implicature: low subjective probability that Granny got drunk relative to a
contextually determined class of individuals C.
Conventional implicature is conventional in the sense that it is a function

of linguistic meaning and context in the way that truth-conditional content is.
Nevertheless, although conventional in this sense, it makes no contribution to
truth conditions. Rather, it introduces a pragmatic presupposition. The delivery
of content in implicature is in the mode of the unsaid. How exactly we should
understand this mode will concern us greatly below.
The value hybridist can use the putative features of conventional-

implicature-bearing sentences to fashion a theory of value sentences. Call this
approach the conventional implicature theory or IT.1 For IT, value sentences are
dual-content sentences, just like (1) is. In producing a value utterance, like O
is good, the speaker U says that O has F for some natural property F and con-
ventionally implicates approval of F-things—or instantiation of F and so on.
So the dual-content form of such sentences is:
Said content: O is F.
Implicature: approval of instantiation of F.
IT can also be work for slurs, pejoratives, and so forth, sentences like O is a
pom, which seems to carry both descriptive and evaluative content. So the basic
form for O is a pom, would be:
Said content: O is English.

Implicature: disapproval of instantiation of Englishness.
IT can be supplemented with a theory of conversational pragmatics. In

self-standing utterances of O is good, additional communicated content may
be present due to the mechanisms of conversational implicature. The speaker U
may conversationally implicate that she or he approves of O being F or approves
of O. Such content, however, would only be present under some circumstances,
for example, self-standing utterances of O is good.
In this chapter I argue that once we accept that conventional implicature
exists and develop an adequate theory of how it works, then IT can be main-
tained as an adequate theory of value language. However, making sense of
conventional implicature is surprisingly challenging and indeed touches on
expressivism in a way that moves us beyond mere hybridism. That is because to
make sense of conventional implicature as a mode of unsaid content, we need
already to accept a form of pure, nonhybrid expressivism about a range of locu-
tions and constructions in natural language, or so I will argue. Moreover, the
form of expressivism has to be a cognitive expressivism, one according to which
(expressive) talk is truth apt and belief expressing. Once this expressivism is in
place, we can have a hybrid theory of value sentences, IT, for free. However, we
may have lost some of our motivation for it along the way, since pure cognitivist
expressivism becomes available.
1. I defended this view in Barker 2000. See Copp 2001 for a related proposal.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 201
I MP L I C AT U R E AND THE I NTE R SUB J ECTIVE

D I ME N S I ON
Why believe that conventional implicature operators deliver non-truth-

conditional content? Evidence for this contention is that we do not judge to be
false sentences whose implicatures fail. For example, a sentence like (2) below
does not appear to be false:
(2) Even Elvis was famous.
Although we want to judge (2) incorrect, we do not want to say that it is false.
The fact that we do not want to affirm the falsity of (2) does not establish in
itself that even contributes an implicature (see Finlay 2005). We do not want to
affirm that (2) is true either, that is:
(3) It is true that even Elvis was famous.
That is consistent with even introducing a semantic presupposition—a con-

tent whose failure generates a truth-value gap or simply a failure of bivalence,
whereas implicature is meant to leave truth value untouched.
However, there is a fairly obvious reason why even cannot introduce a semantic
presupposition. The said content of an even sentence like Even Elvis was famous is
that Elvis was famous. Even’s implicature has no semantic connection to this con-
tent of the kind that, if the implicature content fails in the context, then the said
content of the utterance is undermined. (2) says what it says whether the impli-
cature fails or not. If so, we ought to conclude that (2) is true even if infelicitous.
Even does not introduce semantic presupposition. The unassertability of (3) can
be explained through conversational implicature. That is, a truth attribution car-
ries a generalized conversational implicature that the conventional implicatures
of the sentence are fulfilled (see Barker 2003 for discussion of this point).
Multipropositional Sentences
If conventional implicature is not semantic presupposition, it might neverthe-

less be some other kind of semantic phenomenon. Some authors, for example
Kent Bach (1999), attempt to assimilate conventional implicature to second-
ary propositional content. Bach’s view is that implicature-bearing sentences are
multipropositional sentences in the following sense. (1) Even Granny is drunk,
following Bach’s proposal, comes out something like this:
(4) Granny, who, by the way, is extremely unlikely to get drunk, is drunk.
So, the content is:
Primary Proposition: <Granny is drunk>

Secondary Proposition: <Granny is extremely unlikely to get drunk>
According to Bach, the whole sentence (4) will be true or false according to

whether the primary proposition is true. If the secondary content fails to obtain,
then the resulting evaluation of the whole sentence is upset. It is not clear that
we want to say (4) as a whole is true. Though we can make sense of the judgment
that part of what is said is true, part of what is said (secondary) is false.
Bach’s approach is not tenable. The issue I want to concentrate on is the sup-
posed proposition that is the content of the implicature, the one to do with prob-
ability. What kind of probability are we concerned with? There are broadly two
options: objective probability or subjective probability. Both options are problem-
atic for Bach’s propositional treatment of implicature. Let us look at the alternatives.
Are we conveying facts about objective probability with even statements?
That does not seem likely. People produce even sentences on evidence that is
insufficient to ground objective probability claims. Suppose I am at a party and
see people around me of varying ages, but all are drunk. I might issue: Even
Grandma over there is drunk. To be able to make this claim, I do not wait for the
kind of evidence that I would require to convey that it is objectively improbable
that a certain woman at the party is drunk. My even sentence reflects my epis-
temic state, which, I admit, may fall short of reflecting reality perfectly.
That leaves us with the second alternative, that even sentences express sub-
jective probabilities. However, if that is what they express, what is, according
to the multiproposition hypothesis, the secondary proposition concerning this
subjective probability? One idea is that the proposition is about the speaker
possessing a subjective probability state. In that case the even sentence really
has the form:
(5) Granny, who my credence state assigns low probability of drunkenness

to, is drunk.
But if this is the form of (1), we face another problem concerning how speakers
judge each others’ utterances of even sentences. This takes a little explaining.
Evaluating the correctness of speech acts has two dimensions. The first dimen-
sion I call production correctness. In judging whether an utterance is production
correct I am concerned with whether the speaker, given the information in her
or his ken, made no error, broke no linguistic rule, in issuing her or his utter-
ance. So if someone utters (2) Even Elvis was famous, I can judge that the speaker’s
utterance is production correct if I find out that the utterer U thought Elvis was a
marginal case of fame. The speaker then makes no linguistic error in producing
her or his utterance. Nevertheless, in this case we still judge that U’s utterance is
wrong but in another respect. This is the other dimension of correctness: recep-
tion correctness. In judging a sentence reception correct, I am committing myself
to issuing the sentence (or some content equivalent utterance) in my own case. I
accept the utterance in the sense that I am sincerely willing to produce it myself as
an utterance expressing whatever mental states its utterance typically expresses. In
the case of (2), assuming typical beliefs about Elvis, I am not inclined to produce
the utterance. Although I judge its said content true—it is reception correct at
least with respect to truth assessment—I am not willing to accept the implicature
content.
What, however, guides reception-correctness judgment with respect to the prob-
ability attitude supplied by even? I think the following general thesis captures it:
Intersubjective Dimension for even (ID-Even): H judges U’s utterance of

Event O is F correct iff H believes O is F and H has a subjective probability
state assigning low credence to O being F (and O believes others in the
contextually given class are F).
In other words, in judging an even sentence correct, as produced by another

speaker, I am not concerned with whether she or he has the subjective prob-
ability state but with whether I have the state. I call this the intersubjective
dimension of implicature evaluation, since speakers do not judge reception
correctness on the basis of whether the speaker has the state but on whether
they, the assessors, have the state.2
Here is the problem for the multiproposition hypothesis. If H judges U’s utter-
ance reception correct, then U and H agree about an utterance. But if ID-even
is correct, we are left with a puzzle about what the common proposition is that
U and H agree about. On the multiproposition view, assuming (5) captures the
content of the even sentence, U utters the even sentence encoding the proposition
that she or he, U, has such and such probability state, and H agrees but on the
basis that she or he, H, has a subjective probability state. U and H are talking at
cross purposes, since there are two distinct propositions concerning probability:
<U’s credence state assigns low probability to O being F>

<H’s credence state assigns low probability to O being F>
2. See Barker 2003 and 2004 for discussions of the intersubjective dimension in various
contexts.
What is the proposition about subjective probability that they agree on when
one speaker agrees with the other that even Granny got drunk?
One might propose that the illusive secondary proposition about probability
is the proposition that both speaker and hearer have a certain subjective prob-
ability state. But what about a third speaker, S, who overhears the conversation
though is not included in it? S can agree or disagree with the even sentence pro-
duced, but that will not be on the basis that S thinks both parties have the prob-
ability state. Rather, again in line with ID-even above, S will search her or his
own bosom to determine whether she or he has the subjective-probability state.
Another proposal about the probability in secondary propositional content—
<Granny is extremely unlikely to get drunk>—unlikelihood is irreducible. It is nei-
ther objective probability nor subjective probability but somehow between these
two poles. This is certainly a position one might propose. For example, consider:
(6) It’s surprising that Granny got drunk.
It does not seem unreasonable to hold that utterance of (6) is neither a report
about some objective fact of surprisingness nor a report about a mere subjec-
tive state. That certainly seems to be correct. But it is nevertheless problematic,
since how can we invoke without explication such putative forms of probability
or surprisingness? Moreover, it leaves us with a puzzle about the intersubjective
dimension (ID-even) as I have called it.3 How does the fact that speakers con-
sult their own subjective-probability states relate to judgments involving this
supposed probability that is neither objective nor subjective?
Even if we could accept this kind of probability, I think the dual-proposition
view still misrepresents implicature. The dual-propositional approach takes it
as axiomatic that sentence content is propositional. So although the implica-
ture is not part of the truth conditions of sentence (1), it still has propositional
content. So on this view, in saying (1) we say something true, that Granny got
drunk, and something else true, that it is unlikely that she got drunk, even if
this latter truth does not enter into the truth conditions of the sentence in the
way the first proposition does. But I do not think one is communicating a sec-
ondary truth with (1) Even Granny got drunk. For example, it is inappropriate
to reply to utterance of (1) along these lines: What you are saying is true. It’s sur-
prising she got drunk. This misrepresents the mode of content delivery provided
by implicature. But one could so respond to utterance of (4). In other words,
(4) is not synonymous with (1).
3. Could relativism about reception correctness conditions help us with ID-even? I think
there are all sorts of problems with relativism in relation to phenomena like the intersubjec-
tive dimension. See Barker 2011b.
Pragmatic Presupposition
Our attempts at reducing conventional implicature to semantic presupposition

and to secondary propositional content have not worked.4 What will work?
One broad idea is that conventional implicature is pragmatic presupposition in
Robert Stalnaker’s (1973) sense. On Stalnaker’s view, if a sentence S carries a
pragmatic presupposition P, then when S is uttered speakers would normally
expect to hold P as common ground between discourse participants. We can
allow that P is an attitude, for example, a subjective probability state. So U in
uttering Even Granny got drunk would normally expect her or his discourse
participants to have a subject-probability state assigning low likelihood to
Granny getting drunk. Stalnaker’s view, however, leaves a puzzle about how
the pragmatic presupposition functions as a mode of content. Suppose U utters
Even Granny got drunk. Then it seems part of the conventional meaning is that
she or he normally expects that her or his audience has the probability attitude.
But is this part of the truth-conditional content of the utterance? Clearly we do
not want it to be. The speaker is not asserting that normally speakers have such
and such attitudes. So what kind of speech act is the speaker producing in com-
municating the content that the attitudes are background? Stalnaker’s approach
does not tell us. Rather, it presupposes some as yet untheorized speech act.
I conclude that Stalnaker’s approach fails, because it conflates the state (of a
conversation) in which propositions are held as uncontroversial with a mode of
delivery of content that is non–truth conditional and presuppositional. It is the
latter we are seeking to understand and not the former if we want to illuminate
conventional implicature.
I MP L I C AT U R E AND THE E SSE NCE OF A S S ER TION
I now want to turn to the view of implicature that I think works. This view
provides an account of non-said-content delivery through an account of asser-
tion that clearly distinguishes asserting from implicating. In what follows
I treat implicating as a speech act of a nonassertoric kind—I will often call it
an implicative act. If so, in uttering an implicature-bearing sentence like (1) in
a self-standing illocutionary act, a speaker U performs a compound illocu-
tionary act containing two speech acts that are intimately connected but dis-
tinct. U performs two acts in tandem: (i) an assertion and (ii) an implicative
4. I therefore cannot accept David Copp’s (2009) approach, which accepts Bach’s framework
of secondary propositions for his version of the implicature approach.
act. Utterance of the whole sentence is an assertion but also an implicative

act, both performed simultaneously. So utterance of (1) has a dual speech-act
purpose. There is (1) insofar as it is produced in the assertion, denote that by
Even Granny got drunkASS, and (1) insofar as it is produced in the implicative
act, Even Granny got drunkIMP. If so, (1) can be judged true or false, because it is
used to make an assertion, but also judged felicitous or infelicitous, because it is
used to perform an implicative act.
The task now is to distinguish the illocutionary act of assertion from the illo-
cutionary act of implicating. You might think the usual accounts of assertion
will distinguish assertion from implicating, but they fail to do so. Typically,
people distinguish assertion as a belief-manifesting act, or one aimed at truth,
or that act that is governed by the norm of knowledge. But such approaches
will not distinguish assertion from implicating. So they cannot be correct
accounts of assertion.
Take knowledge. According to the knowledge norm, assertion is that act,
governed by the norm: assert that P only if you know that P. It seems that
knowledge can apply to implicature-bearing sentences. Take:
(7) Buggsy knows that even the best philosophical minds can get confused.
In (7) implicature comes within the scope of knows that—there is no rea-

son to think it does not. If so, implicature-bearing sentences can be known.
Moreover, a case can be made that one should only produce Even the best philo-
sophical minds can get confused in a self-standing act if one knows that even
the best philosophical minds can get confused. If so, a case can be made that
implicature-bearing sentences conform to a norm of knowledge with respect to
their production in illocutionary acts. But these acts are not assertions, so the
knowledge-norm treatment of assertion cannot be right.
One might object that implicating cannot be governed by the norm of knowl-
edge, because implicatures are not propositional. Take implicature-bearing
sentence (1). Nothing propositional is communicated by (1) regarding the
implicature—at least that is what I have argued. One might contend, then, that
this is why implicating is not asserting. Assertions are displays of commitment
to propositions, whereas implicative acts cannot be that since there is no propo-
sition to endorse.
This is all very well, but it just assumes the distinction between propositional
versus nonpropositional speech acts. And what is the ground of that distinc-
tion? For example, take the truth-apt sentence (8):
(8) It’s very unlikely that Granny got drunk.

(8) presumably encodes a proposition, whereas Even Granny got drunkIMP does
not. If we are really going to explain the difference between the two sentences, we
need to give an account of the nature of the proposition in the case of (8). But as
we saw above in examining Bach’s multiproposition view of implicature-bearing
sentences, that is hard to do.
One might reply that truth aptness is the answer. Sentence (8) is truth
apt—it is open to truth/false assessability—whereas Even Granny got drunk-
IMP
is not truth apt. But what account of truth aptness will deliver that result?
I have argued elsewhere (Barker 2011a) that the usual accounts of truth apt-
ness, such a disciplined syntacticism, do not work in the context of languages
with implicature.
Brandom on Assertion
What of Robert Brandom’s (1998) celebrated theory of assertion? Brandom

attempts to carve out the assertions through the phenomenon of inference.
For Brandom, assertion is a move in the game of giving and taking reasons.
Very roughly:
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.
However, Brandom’s theory cannot deal with languages featuring implicature,

since uttered implicature-bearing sentences meet both of his conditions for
assertions. Let us see how.
Implicative acts meet condition (i). Suppose someone utters (1): Even Granny
got drunk. It is obvious that one could be called on to justify one’s grounds for
the implicature component. For example, one’s audience might say: Whaddaya
mean, even Granny got drunk! Her or his response is asking for justification
with respect to the implicature. Implicating involves commitment, as does any
speech act. If so, one might be called on to justify that commitment. Condition
(i) is met.
Condition (ii) of Brandom’s approach involves inference. If one asserts, one
issues that utterance as a potential premise in further arguments. But note now
that implicature-bearing sentences can enter into inferential relations. For
example, the piece of reasoning in (9) is fine and involves implicature:
(9) Even Granny was drunk. If even Granny was drunk, it must have been
a wild party. Therefore it was a wild party.
Even plays a role in this inference. Remove even from the argument, and one
loses a sense of what is being conveyed. Compare:
(10) Granny was drunk. If Granny was drunk, it must have been a wild
party. Therefore it was a wild party.
The difference is that (10) is open to interpretations not available to (9). For
example, (10) sustains the interpretation that the reason parties are wild if
Granny gets drunk is that in her drunken state she commands that everyone
smoke vast amounts of marijuana, generating general hilarity and high jinks
but nonalcoholic wildness. One cannot produce (9) in that context. Even, then,
is doing work in the overall constraining interpretation of the ground for (9).
I take that to mean it enters into inferential relations.
Implicating meets both of Brandom’s conditions (i) and (ii) for being an
assertion. But implicating is not asserting. So Brandom has not isolated the
essence of assertion.
The Positive Proposal
Nevertheless, we can take the spirit of Brandom’s theory to forge an account that
will distinguish asserting from implicating. My proposal is to shift the expla-
nation from utterances to mental states. Consider again (8): It’s very unlikely
that Granny got drunk. (8) is a declarative sentence whose standard use is to
make an assertion. It is a truth-apt sentence. But there is an intimate connec-
tion between (8) and (1), in particular the implicature component of (1), Even
Granny got drunkIMP. Both involve the same mental state, a subjective prob-
ability with respect to Granny getting drunk. In other words, exactly the same
mental state P can underpin an assertion and a distinct implicative act. So what
makes the difference between asserting and implicating? The difference is how
the state P enters into the act. My proposal is this. In an assertion the speaker’s
purpose is to manifest a defensive stance with respect to P, and in an implicative
act it is merely to manifest the state. Here is the core idea:
Assert: Assertion is a speech act comprising utterance of S with the pur-

pose of manifesting a disposition to defend a mental state P.
Implicate: Implicating is a speech act comprising utterance of S with the
purpose of manifesting a state P.
Defense means the speaker is disposed to indicate reasons for the state P.
Although any rational being with a state will have reasons for it—it will fit
into a (relatively) rational array of mental structures involving other states—

assertion is the act where the purpose of the act is to manifest a disposition
to display reasons, that is, the rational structure in which the state embeds.
By contrast, implicating involves the purpose of merely manifesting the
state. It is this difference of purpose that is crucial to the difference between
asserting and implicating.
Suppose a speaker produces an assertion and so utters a sentence with the pur-
pose of manifesting a disposition to defend possession of a mental state P. Then the
audience in processing and accepting the utterance has to reconstruct it in her or
his own case and recognize its purpose. In recognizing that the purpose is dialecti-
cal engagement, the audience registers that that acceptance of P is not being taken
for granted. On the other hand, implicating is utterance whose purpose is merely
expressing/manifesting P. Although one may be called on to provide reasons for
the implicative speech act, this is not built into the purpose of the speech act. The
purpose is not to defend the state P. There is, in other words, no call to dialectical
arms. If so, P is being “taken for granted” in that case. The proposed theory then
explains why implicative acts are presuppositional, whereas assertions are not.
Although in implicating a speaker does not utter a sentence with the purpose
of manifesting a defensive stance with respect to a state P, implicature-bearing
sentences can enter into inferences. Entering into inference is not constitutive,
pace Brandom, of these sentences being asserted. Moreover, although implicat-
ing involves no purpose of manifesting a dialectical stance, implicating is still
open to the whaddaya mean response—calling for justification. That is because
implicating is still open to reception-correctness judgment, because it involves
commitment, like any speech act.
The defensive/nondefensive expression distinction explains implicating’s pre-
suppositional status. But what about non–truth aptness? Sentence (8) is truth
apt since it is associated with the production of an assertion. However, (1) taken
in its implicative role, used merely to express a state P, is non–truth apt. But why
is defensive expression associated with truth aptness and nondefensive expres-
sion not? There is no deep explanation here. Truth is just the reception correct-
ness term for strings used in assertions. (More about that below.)
You might object to this account of truth aptness in sentences because it
links truth aptness to assertion. What about unasserted sentences in logical
compounds? They are unasserted, but can they be truth apt? When simple
declarative sentences are embedded, they are not asserted, but they are still
associated with assertion types. It is their association with assertion types that
renders them truth apt. So when we think of content, propositional, we really
mean assertion type (Barker 2004). An implicature, the content of a sentence
used in an implicative act SIMP, is an implicative type. The content of SIMP is not
propositional. An implicative type is not an assertion type. You cannot capture
the content of an implicature content through a that clause. That is why Bach’s
multiproposition account of implicature-bearing sentences is untenable. It con-
fuses two kinds of acts.
Mental States Defended and Expressed
Let us note that on the present account sentences do not inherit their truth apt-
ness from the states that are defensively expressed through utterance of those
sentences. Subjective probability states are not truth apt. They are not belief
states. We might say they are predoxastic states—states prior to belief. Still, sen-
tence (8) is truth apt. Its truth aptness comes from being associated with an
assertion, a speech act whose essence is linked to the defensive expression of
states.5
Although subjective probability states are not truth apt, they can neverthe-
less enter into relations of rationality. They can be grounded in evidence or
sensory experience and meet formal requirements. If so, one can have a defen-
sive, reason-manifesting stance with respect to them. A general question at this
point is whether this picture of truth aptness and assertion generalizes to all
assertions. Should we analyze all assertions as acts of expressive defense with
respect to mental states that are not in themselves truth apt? This is the view that
I have argued is indeed correct. Articulating and defending that goes beyond
the bounds of this chapter and is not required as such to make the main claims
presented here.6
Both (1) and (8) can be believed. For example, one can believe that even
Granny got drunk. So we can say that utterance of (1) can be belief manifest-
ing in that sense and likewise with (8). So although these sentences involve
expressing predoxastic states, nevertheless they are associated with beliefs. We
shall look at the nature of belief states that makes sense of this idea below in the
fifth section.
Pure Expressivism
The account we have given of assertion is essentially expressivist. Take (8). (8)
can be used to make assertions, but the state expressed is a predoxastic state.
5. See Barker 2007 but also Barker 2011b. In the latter I argue that to make sense of faultless
disagreement we have to think of assertion generally as expressive of predoxastic states.
6. That is somewhat hand-waving, but look at Barker 2004, 2007 for more details.
This expressivism is nevertheless cognitivist in that sentences like (8) are treated
as truth apt and belief manifesting and not just in some minimalist way.
This pure expressivism allows us to make sense of the puzzle we had with
reception correctness and the intersubjective dimension, which we wrestled
with in the first section. Say H judges that U’s production of an even sentence is
correct. What are U and H agreeing about according to ID-even? One answer
is that they are agreeing about a speech-act type, one in which a speaker utters
a sentence and one defensively expressing some state P and nondefensively
expressing a distinct but related subjectively probability state S. In short, the
bearer of agreement is not a proposition in the sense of a representation of a
state of affairs but a speech-act kind identified by mental properties. In other
words, it is a compound illocutionary act type of the kind we have character-
ized above. In saying that the sentence is correct, H is expressing her or his
disposition to perform the corresponding speech act.
The only way to make sense of that is through an expressivist treatment of
correctness claims:
In uttering S is correct, U defensively expresses her or his disposition to

produce S sincerely and clearheadedly.
In other words, statements of correctness defensively express dispositions to

reissue utterances (judged to be correct). All of this goes to show that to make
sense of conventional implicature we need some form of pure expressivism.
Similar comments apply to the term true as a predicate used in assessment of
utterances. I leave details aside here.
I N T E R P R E TATI O N AN D THE I NTE R SUBJECTIVE

D I ME N S I ON
That is the basic picture of conventional implicature and the kind of expres-
sivist treatment that it seems to require. Conventional implicature as a mode
of nonpropositional content requires a serious rethinking about the nature
of assertion. So accepting conventional implicature at all is a serious business
from the point of view of semantic foundations. Let us now return to IT, the
conventional implicature theory of value sentences.
According to IT, in producing an utterance of O is good, U asserts that O is
F and implicates approval of F being instantiated (in general). The property F
predicated of O is the property in the associated moral attitude or F-attitude.
I will assume that good has distinct evaluative senses: moral, aesthetic, func-
tional—as in good hammer and so on. The different evaluative senses are a
function of the kinds of attitudes involved. A moral sense of good presumably

is associated with attitudes about human conduct directed toward others.
These very general constraints on attitudes, however, will not fix in them-
selves any specific natural property F. So what does? What property F does the
speaker’s utterance come to pick out? One hypothesis is that what fixes a par-
ticular F is the specific F-attitudes of the speaker. Such a proposal would be an
indexical theory, according to which each speaker’s utterance of O is good picks
out the F whose instantiation the speaker approves of.7 This cannot be correct.
Consider the conversation:
Schmidt: Hitler was good.
Brown: Hitler was not good.
If the indexical approach were correct, then each speaker, given her or his
distinct F-attitudes, would assign different properties F to good. Schmidt and
Brown then would not be disagreeing with each other. They would be talking at
cross-purposes. That would be a disaster for IT.
An alternative is that the property F is fixed by the audiences’ attitudes. So
Brown judges racist Schmidt’s Hitler was good not in terms of Schmidt’s twisted
attitudes, approval of race supremacism, but her, Brown’s, own approval of toler-
ance and antiracism. Brown judges Schmidt’s statement false accordingly, since
on this interpretation Schmidt asserts that Adolf Hitler was tolerant and antira-
cist. This interpretationist theory moves us from a speaker-centered indexical-
ity to an audience/interpreter-centered indexicality. Although it improves the
position of matters to some degree, it is still unattractive. What does good in
the conversation refer to? One would have to answer that it is a relative mat-
ter. Relative to Schmidt qua audience of Brown’s utterances, it refers to FS, and
relative to Brown qua audience of Schmidt’s utterances, it refers to FB. But then,
again, it makes it hard to see how Schmidt and Brown are having a real debate.
Of course one might respond that they are not really having a debate. That is
par for the course, given the general assumption of expressivism. That is a kind
of defeatism we want to avoid if possible.
This looks like a serious position for IT. Where do we go from here? The
answer lies again in the intersubjective dimension. I suggest that in thinking
of the reference of good in the context of the implicature theory (and expres-
sivism in general) we ought to move from considering the question What fixes
the reference of the term “good” to considering How does talk about the reference
of “good” work? That is similar to our move from asking What does reception
correctness consist in to asking How do judgments of reception correctness work?
7. This is how Daniel Boisvert (2008), Stephen Finlay (2005), and Mark Schroeder (2009)
take the view.
In judging what a speaker is referring to with good, the speaker H appeals to

her or his own F-attitudes. After all, these attitudes capture her or his concep-
tion of the good. So H should use that conception to judge what people are
referring to with good (whatever their attitudes are). Instead, we are proposing:
Ref-Exp: H’s judgment that good refers to F is grounded in H’s F-attitudes:

in judging that U refers to Fx with good, H expressively defends a disposi-
tion to use good with F-attitudes involving Fx.
This is not a theory of what conditions there are for good to refer to some F
but a theory of what lies behind the production of attributions of reference to
a term, good. This is not a speaker-indexical theory or an audience-indexical
theory. It is not a theory of reference at all. It is a theory of reference talk.
Call this an expressivist treatment of reference. How does it help us? Consider
Schmidt and Brown’s exchange. Are they taking at cross-purposes? No. We, as
observers of their debate, make our judgment about what they are referring
to. Having broadly identified that good in their mouths involves expression of
moral attitudes, we then judge that they are referring to FUs based on our own
conceptions of the good. FUs is the F-property fixed by our (moral) F-attitudes.
This use property of our utterances, the fact that we have specific F-attitudes
and they lie behind production of our utterances using good, does not fix the
reference of our term good. We are not here in the realm of a theory of what fixes
reference. Rather, the claim is this: our judgment that Schmidt and Brown refer
to some natural property is nothing but a defensive expression of a disposition
to use good with those F-attitudes. We judge that they are talking about good-
ness qua FUs. That is what we take goodness to be. Indeed we can assert: that is
what goodness is. It is not a relative matter.
This approach requires that we now countenance the idea that speakers can
express attitudes that in fact they do not have from the point of view of what,
psychologically speaking, their attitudes are. So Schmidt’s utterance is taken by
us as using good to refer to tolerance and antiracism and so on. Obviously, as a
matter of psychology, Schmidt lacks any such F-attitudes. In short, expressing
attitudes is not a psychological matter. Does this denude the implicature theory
of any interest or empirical bite? I do not see why. It seems, on the contrary,
what expressivism should be aiming for. As a kind of antisubjectivism, expres-
sivism seeks to separate value utterance from any kind of autobiography.
Now, Schmidt might cotton on to the fact that he is being interpreted by us as
picking out FUs. Could he say that this is not what he means by good? He could.
But that would just boil down to his saying that this is not what his conception
of the good is. For example, Schmidt and Brown can engage in a debate about
what goodness is. That is the same as a debate about what the term good really
refers to. In short, the expressivist account of reference fits in with reference
being an absolute matter. There is no place in the account for any relativism.
The distinct F-attitudes that speakers have do not determine distinct reference,
just what we might call conceptions of the good. But the price we pay for this
is that expressing an attitude ceases to be a psychological matter. This does not
mean that it is a mysterious emergent property either. It is just as our expressiv-
ist treatment of the reference of good requires that we cease thinking of good’s
reference as a psychological property fixed by empirical attitudes of agents,
given that F-attitudes move in tandem with the referent of good.
These proposals about reference attributions to good give rise to quite a few
general philosophical questions. How does the expressivism about reference
for good fit in with attributions of reference more generally? Is it part of a gen-
eral expressivism about attribution of reference to predicates? I think it is—it
is not attractive to suppose that expressivism about reference merely applies
to value predicates, though one might take this line. One might put forward a
general case that in assigning reference to any predicate F, a speaker expresses
a disposition to use F in a certain way. That certain way cannot be specified in
terms that presuppose the reference of F. Rather, the certain way would have
to be a use property that involves capacities to interact with the environment,
a differential sensitivity to entities in the world, but such facts of use would not
amount to fixing reference.8
To illustrate the thought, take color predicates. An expressivism about reference
would have to suppose that a capacity to use red is underpinned by, say, men-
tal modules that react differentially to surfaces in the visible environment. But
we do not have to suppose that these mental modules enabling one to use red
together, given embedding in an environment, fix an extension. Maybe the mental
machinery of humans simply cannot fix something as precise as that.9 Still, distinct
speakers largely agree in the module structures that underlie their uses of red. That
explains stability and agreement in practice, including indecision about borderline
cases. But an expressivist about reference does not have to say this similarity in
underlying functional states fixes an extension of the term as used by that group.
Nor do they have to say that it constitutes that fact that such and such speakers all
mean the same by red. Rather, at best the similarity of underlying functional states
across certain groups of speakers explains why the meaning judgments of such
speakers and their applications of red to given surfaces tend to agree.
The expressivist about reference can still say that someone who sincerely
claims that red as a word in a public language denotes yellowness betrays
8. This view is discussed most recently in some detail in Barker forthcoming.

9. Epistemicists, who think that somehow human cognitive systems are able to pick out utterly
precise extension, are placing almost Olympian standards on human cognitive capacities.
semantic confusion. We do not have to bring in conventions of language com-

munities to explicate the notion of semantic confusion or correctness. Rather,
the predicates, semantically confused or correctly using the term, will them-
selves be predicates given expressive explication.
I do not think that the proposals about expressivism concerning refers in
relation to value language necessarily require the sort of generalization I just
sketched. Still, knowing that it is a possibility might support the specific local
proposal about good I have made.
Finally, the move we are making shows we need to be careful with the term
express. There is a sense of express that holds of Schmidt insofar as he produces
utterances whose antecedents are certain attitudes, F-attitudes directed toward FS.
But there is also a sense of express in which (given our judgment) he expresses
approval of FUS based in the expressivist account of reference attribution. Call the
first sense expressPrime and the second expressExp.10 Both are important for under-
standing what is going on. When we analyze speech acts—the underlying mecha-
nisms of production of sentences—we are concerned with expressPrime. But from
the point of view of reception by others, we are usually concerned with expressExp.
I will make the sense of express explicit when it is required in the discussion below.
MO T I VAT I ON AL E X TE R N ALI SM
I have argued that countenancing conventional implicature in a language

requires adopting a cognitive expressivist treatment of assertion and a pure
(nonhybrid) expressivism concerning certain locutions, such as the predicate
correct—conveying reception correctness. It can be shown easily enough that
we can deal with the so-called Frege-Geach embedding problem using this
framework. I have already indicated that sentences in embedded contexts are
associated with speech-act types, both assertive and implicative. That provides
the bedrock for explaining what underlies our judgments of validity of argu-
ments featuring implicature-bearing sentences. I will not go into that here.11
Instead, I shall consider another contentious issue affecting expressivism. Take
the amoralist who lacks any of the feelings of approval or disapproval. Can she
or he sincerely make moral claims? On a simple expressivist story, the problem
10. One might say that we are being expressivist (in the sense of expressPrime) about expressing
in the second sense. That is because claims like Schmidt is expressing such and such attitude
(which he psychologically lacks) can only be understood in terms of an expressivist treatment
of reference attribution. That expressivism about reference infects the expressivism about
attributions of expressing attitudes.
11. See Barker 2003, 2004, and 2007 for some details.
is that if sincere moral assertion is to have moral beliefs and having moral beliefs
is in the end about having desires (motivational states), then to really sincerely
assert, a speaker has to have moral desires, but that is what amoralists (psycho-
paths and so on) lack. There appears to be a serious issue for expressivism here.
One might attempt to circumvent the issue by denying that amoralists can
make sincere value assertions. I do not want to consider this avenue here. Let us
suppose that amoralists can make sincere assertions about what is good, right,
required, and so on. The amoralist gains her or his moral beliefs from oth-
ers through testimony. She or he lacks internal mechanisms, attitudes, through
which she or he can launch her or his own judgments. So how can the expres-
sivist allow for amoralists who nevertheless, on the basis of testimony, make
moral claims? This problem is not specific to hybrid theories, since hybrid
expressivists have it too.12 However, as I will now argue, the implicature theory,
IT, has resources enabling it to deal with the amoralist, since, as we have seen,
it can allow that desires are implicated by agents who lack the desires. We say
above that Schmidt should be interpreted by us as implicating and expressing
attitudes that, from a psychological point of view, he lacks. (This is expressingExp
sense—see the second section above.) Think, then, of the amoralist as analo-
gous to Schmidt (from the third section). The amoralist implicates F-attitudes
that she or he lacks. She or he lacks them because she or he simply lacks the
psychological resources or inclinations to develop any such attitudes.
To get a rich sense of what might be going, on we need to turn to analo-
gous cases in relation to conventional implicature operators, like even, where
we shall see that comparable problems to the problem of the amoralist and the
expressivist arise.
Implicature and Subjective Probability
Consider the person Robbo who is told that even Timbo got drunk. Let us sup-
pose that Robbo knows who Timbo is but has no posterior probability scale
about the relative unlikeliness of Timbo getting drunk (compared with mem-
bers in some contextually given class). Robbo does not know about Timbo’s
history, drinking tendencies, any statements he has made about the evils of
alcohol, and so on, that is, the sort of information that typically would underlie
a subjective probability state. Despite lacking such information, Robbo might
come to believe his audience’s claims. Robbo could say: Jacko told me that even
12. Indeed the amoralist is the reason some hybrid expressivists, like Finlay (2005) and Dorit
Bar-On and Matthew Chrisman (2009), treat attitudes as only ever conversationally impli-
cated by moral utterances and thus not as essential components of such utterances.
Timbo got drunk. I believe Jacko. So it would seem even Timbo got drunk. Robbo’s
claim is not insincere or confused.
The same source of testimony that furnishes Robbo with the even sentence
also furnishes him with a subjective-probability claim. Robbo can affirm Timbo
is unlikely to drink based on testimony. Here again, we can wonder, for the rea-
sons already given, whether Robbo can have the subjective probability state
supposedly expressed by sincere clearheaded utterance of Timbo is unlikely to
get drunk. One might suppose that Robbo could have a credence state that,
although lacking any evidential base, would be manifested through betting
behavior. That is, as a result of the testimony Robbo is disposed to accept odds
of Timbo’s getting drunk in accord with the a high degree of credence that
he will not get drunk. But one might doubt that Robbo could feel informed
enough to make a bet. For Robbo, there could be no estimation or weighing of
possibilities, since he lacks evidence, that is, general information about Timbo.
You might object. Suppose Robbo is forced to bet on whether at the next
party Timbo is likely to get drunk. Given that Robbo has been informed that it
is very unlikely that Timbo will get drunk, then surely the only rational course
for Robbo is to accept very low odds on Timbo’s inebriation. I am inclined to
think Robbo should feel a bit queer about this. Robbo should think along these
lines: I have to make a bet. OK. How would I act if I really had the subjective
probability state that my informer Jacko (who really knows Timbo) has? I will bet
like that: accept low odds. But I am not really accepting these odds as a reflection
of my credence state. I am really only acting as a kind of proxy for Jacko. It is not
really my gambling behavior.13
My claim is that although informed through testimony that even Timbo got
drunk, Robbo really lacks the subjective probability state. Nevertheless, Robbo
can sincerely and clearheadedly produce the implicature-bearing utterance.
There are various locutions that can be added to the utterances to indicate
testimony-based status. Robbo might say, Apparently, even Timbo got drunk,
or Timbo is very unlikely to drink, as I have been informed, and so on. Robbo is
not merely engaging in indirect discourse in adding such locutions. Robbo is
committing himself to the correctness of the sentences.
It seems that speakers can sincerely and clearheadedly produce Even O is F or
It’s unlikely that O is F without possessing a subjective-probability state. We did
say that speakers can produce utterances and express states they lack—Schmidt
was a case of that. Still, there is a problem. If speakers can produce these utter-
ances without these states, what exactly is the link between these sentences and
13. Does this mean the gambling behavior is a good measure of “degrees of belief ” when the
subject actually has degrees of belief to be measured? The dispositions to gambling behavior
do not constitute the credence states.
probability states? Why do we associate them at all with probability states? What
exactly is the link? At least part of the issue here is what underpins the fact that
they are being sincere and clearheaded. In other words, what mental states are
the antecedents of the production of their sentences? We have been supposing
that in uttering, sincerely and clearheadedly, in a self-standing illocutionary act
Even T is G, a speaker is doing this—here the sense of express is expressPrime:
Defensively expresses P—the state associated with T is G.

Nondefensively expresses a low subjective probability regarding (T is F).
But if we are going to allow for testimony-based utterance, this cannot be

right. If we are to retain expressivism, then we need another kind of state X
that is expressed in the implicature component of such sentences. However,
we cannot deny that X, whatever it is, while not identical to a subjective prob-
ability state, is nevertheless somehow intimately related to a subjective prob-
ability state.
Here is the basic proposal, though it has to remain sketchy for reasons of
space. What speakers primarily express are states X that have subjective prob-
ability states as their canonical causes (or outputs). Most of the time speakers
have both the canonical state C and X. But sometimes speakers lack the canoni-
cal states and gain X via some other path. Testimony is one such path. What
is this state X? The answer is that it is a functionally abstract state defined by
its inputs and outputs. But we are not supposing that it is a state that any folk
psychological state can be identified with. It is not a belief or a desire. Here in
figure 9.1 is at least part of its functional character of the pathway that underlies
production of sentences:
In other words, the full-blooded states are only canonical causes of the
thinner, relatively functionally abstract state X. But X is the proximal cause of
utterances, which will be defensively or nondefensively expressed of them or
Canonical Input
Unlikely [O is F]
Token [X] Utterance S
Testimony state
Noncanonical
Figure 9.1 Utterance Production Paths

nondefensive. That depends on what kind of speech act the speaker is going to
generate with S.
We can suppose that such states X have a compositional character. The state
corresponding to Even O is F might be structured in a way that reflects the
grammatical structure of the sentence. It will have constituents. These struc-
tures are not meanings. They are part of what underlies the production of
sentences. Expressivists should not say that the desire expressed by a moral
utterance is its meaning. But they want to say that central to the speech act of
uttering the value sentence is the desire.
This modification of the cognitive expressivist theory, in terms of the dis-
tinction between canonical causes and functionally abstract outputs of such
canonical causes, looks almost unavoidable if we are going to get a viable theory
working at all. Take our expressivism about correctness claims. One way I can
make a reception correctness claim is actually being disposed to produce S. But
that is not the only way. Another way is that I am told by someone that some
utterance, which I may not have been encountered, is correct. If so, we have to
allow, again, cognitive pathways, a state X that can be caused by a disposition to
utter a sentence but that might also be caused through a path of testimony, as
when I am told that the first utterance made today is correct.
So how, then, do we deal with the amoralist? It should be clear that the
amoralist is someone who acquires beliefs about the good and so on through
testimony—she or he has no other source for making such claims, since we are
assuming lack of affect. However, that does not prevent her or his functional
system from employing pathways of production that go through noncanonical
routes. The amoralist is just like the person informed about subjective proba-
bility purely through testimony. Like Robbo’s lack of real gambling behavior in
relation to Timbo, the amoralist lacks motivation. Just as Robbo goes through
the motions with placing bets, the amoralist might go through the motion of
acting morally. But there is no real engagement. To conclude: IT is compatible
with morally knowledgeable amoralists and motivational externalism.
B E L I E F, B E L I E F STATE S, AN D O B J E CTS OF BELIEF
We now come to the question of belief. It appears that there are beliefs about
value such that being nice to kids is a good thing. Moral utterances seem to
express beliefs. Can the value expressivist accommodate this fact? Can the
hybrid theorist accommodate it? Again, my strategy in this chapter is to
ask about implicature-bearing sentences and their relation to belief. We can
approach that question by asking if implicature-bearing sentences can appear
in the scope of believe that? They apparently can, as in:
(11) Timbo believes that even Granny got drunk.

(12) Schmidt believes that even Elvis was famous.
The implicatures may or may not project from such a position. To project is to
become a commitment of the whole utterance.14 So, for example, (11) could be a
case of projection. In that case the belief attribution carries an implicature about
subjective probability. Even if there is projection, it does not follow that the con-
tent of the attributed belief does not include the implicature. The implicature
can still be part of the content of the belief. This is clearer in the case of (12). In
this case there is no projection. (12) is a case in which the speaker is attributing
an infelicitous belief to Schmidt. Schmidt has a weird belief. But we do not want
to say it is a false belief. Rather, its particular form of defectiveness is infelicity.
Beliefs can be true or false. But on the picture developed here, they can also
be infelicitous or felicitous. That kind of observation naturally leads to the
idea that the objects of attitudes are utterance types, which (potentially) have
truth-conditional and implicature content. In other words, beliefs are not, as
such, propositional attitudes. In short, the objects of belief as utterance types
and not propositions, understood as sets of worlds, or what have you.15 Rather,
they are speech-act type attitudes. Such a view is not entirely unattractive and
certainly fits in with the earlier claims of this chapter.
If that is right, what is the relation between believer and utterance type? What is
the belief state? The obvious answer is: the belief state is being disposed to token,
in a sincere and clearheaded way, the utterance type, whether that tokening is a
private matter—a sentence in the head—or a public utterance. Let us consider
the speaker who has access to canonical states for production of sentences. Take
believing that Even O is F. This comprises, the sense of expressing is again expressPrime:
Disposition to defensively express P.

Disposition to nondefensively express S.
Here P is the state associated with O is F and S a subjective probability state.

Belief on this picture is irreducibly expressive. It is only at the level of expressive
acts or dispositions thereto that one encounters belief. Prior to that the underlying
states have complicated functional natures, but they are not belief states.
So apply this kind of account to believing that O is good. The belief state is:
Disposition to defensively express P.
14. For one, there is the complicated issue of projection. See Langendoen and Harris 1971.
15. In Barker 2003, 2004, and 2011a I argue for a speech-act theoretic understanding mean-
ings as speech-act types.
Disposition to nondefensively express D.
Here P is the state associated with O is F and D an affect state. The states are cou-
pled in that ultimately they are directed toward the same property F. On this view
the value belief is not a desire state, nor is it a pair of states, one of which is a desire.
No, the coupled states are dispositions to expressive acts, one of which involves
expression of affect. In this respect the belief state for O is good is no less unified
than the belief state for Even O is F.
However, some evaluative beliefs are even more unified. Take this belief:
(13) It’s desirable that Fred leaves.
Sentence (13) involves the defensive expression of a certain affect state—desiring that
Fred leaves. This assertion is a genuine, robust, truth-apt assertion, like any other in
that respect. In this case the belief state is simple, a disposition to defensively express
the affect state. Again, we are not proposing that belief here is a desire. No belief is a
desire. Rather, belief as always is a disposition to a certain kind of expressive act. The
desire enters into this disposition, but that does not make the disposition a desire.16
A more developed characterization of the theory being developed here would
have to take into account what we have said about motivational externalism and
the need for a distinction between canonical causes and functionally abstract
states. But I have been leaving that aside at this stage. The basic point should be
obvious. The present approach is not saying that beliefs for the expressivist are
desires or states that have desires as some conjunctive component.17
P U R E E X P R ESSI V I SM V E R SU S H YB RID
I have tried to make a case for conventional implicature as a phenomenon, have

described some of the structures that I think are required to make sense of it,
have applied this framework to slurs and epithets and thin ethical terms—like
good—basically following Barker 2000. Leaving aside the question of whether this
framework is ultimately tenable, one salient issue needs to be briefly addressed.
That is the role a pure form of cognitivist expressivism plays in the account. We
need this pure expressivism to make sense of conventional implicature, or so
I have argued. But if we need pure expressivism, why not offer a pure cognitivist
16. Nor is it a besire. We are not simply postulating that there are beliefs that are somehow
desires. The belief state corresponding to (13) is not a desire, just internally connected to one.
17. It is interesting to compare the framework being described here and the theories of what
belief states are for a hybridist considered by Michael Ridge (2009).
expressivism of thin value utterances, such as utterances of O is good? Why not

indeed. It is here that a new debate can begin. Indeed a range of possible ways of
understanding value sentences becomes available. The simplest would be to say
that a claim O is good is simply a defensive expression of an affect state (or the
canonical output of such a state). That is, it is very much like the kind of asser-
tion produced in uttering sentences like (13). Given the cognitive expressivism
sketched here, such utterances would be entirely truth apt and belief manifest-
ing. No need for conventional implicature and hybridism. It could be that in the
end, although we can make sense of hybridism—it is a tenable view—it is made
redundant by the very machinery that clears a space for it.
References
Bach, Kent. 1999 “The Myth of Conventional Implicature.” Linguistics and Philosophy
22:327–366.
Barker, Stephen. 2000. “Is Value-Content a Component of Conventional Implicature?”
Analysis 60:268–279.
———. 2003. “Truth and Conventional Implicature.” Mind 112:1–33.
———. 2004. Renewing Meaning. Oxford: Oxford University Press.
———. 2007. “Semantics beyond the Distinction between Sense and Force.” In Illocutions,
Institutions, and Intentionality: Essays on Themes from the Philosophy of John Searle,
ed. Savvas Tsohatzidis. Oxford: Oxford University Press, pp. 190–211.
———. 2011a. “Bearers of Truth and the Unsaid.” In Making Semantics Pragmatic, ed.
Ken Turner, 81–102. Bingley, UK: Emerald.
———. 2011b. “Faultless Disagreement, Cognitive Expressivism, and Absolute but
Non-objective Truth.” Proceedings of the Aristotelian Society 110:183–199.
———. Forthcoming. “Expressivism about Reference: How to Quantify over Non-existent
Entities without Meinongian Metaphysics.” Erkenntnis.
Bar-On, Dorit, and Matthew Chrisman. 2009. “Ethical Neo-Expressivism.” In Oxford
Studies in Metaethics, vol. 4, ed. Russ Shafer-Landau, 133–166. Oxford: Oxford
University Press.
89:169–203.
Brandom, Robert. 1998. Making It Explicit. Cambridge, MA: Harvard University Press.
Copp, David. 2001. “Realist-Expressivism: A Neglected Option for Moral Realism.”
Social Philosophy and Policy 18:1–43.
Metaethics, vol. 4, ed. Russ Schafer-Landau, 167–202. Oxford: Clarendon.
Finlay, Stephen. 2005. “Value and Implicature.” Philosopher’s Imprint 5:1–20.
Langendoen, D. Terence, and Savin Harris. 1971. “The Projection Problem for
Presuppositions.” In Studies in Linguistic Semantics, ed. Charles J. Fillmore and D.
Terence Langendoen, 373–388. New York: Holt, Reinhardt, and Winston.
Ridge, Michael. 2009. “The Truth in Ecumenical Expressivism.” In Reasons for Action,
ed. David Sobel and Steven Wall, 198–220. Cambridge: Cambridge University Press.
Schroeder, Mark. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119:257–309.
Stalnaker, Robert. 1973. “Presuppositions.” Journal of Philosophical Logic 2:447–457.
10
(How) Is Ethical Neo-Expressivism

a Hybrid View?
D O R I T B A R - O N , M AT T H E W C H R I S M A N , A N D J A M E S S I A S n
Mainline metaethical expressivists of the 1960s through the 1990s generally

conceived of expressivism as a view in philosophical semantics. Roughly speak-
ing, the idea is that the meaning of ethical sentences is to be given not in terms
of what they represent or describe but rather in terms of the noncognitive atti-
tudes they express.1 But if this idea is combined with the traditional thought
that the meanings of most other declarative sentences are given by the proposi-
tions they express, then it turns out to be impossible to give a systematic seman-
tics for logically complex sentences with mixed ethical and nonethical parts.2
In response many expressivists have endorsed an “ideationalist” conception of
meaning across the board. That is, they suggest that all sentences mean what
1. It is not obvious that this was how protoexpressivists such as A. J. Ayer (1946), Rudolf
Carnap (1935), and Charles L. Stevenson (1937, 1944) thought of the view. However, under
the influence of R. M. Hare (1952, 1963), who treated ethical sentences semantically as pre-
scriptions and sought to align satisfaction conditions with ethical sentences as one might
align truth conditions with descriptive sentences, the view became a semantic one. Many
contemporary expressivists are quite explicit in locating (the expression of) conative atti-
tudes in the semantics of ethical terms or sentences. See, for example, Gibbard 2003, 75.
Even expressivism’s most prominent critics typically conceive it as a semantic view. See, for
example, Jackson2001,10; Schroeder 2008, 87.
2. This we take to be the main upshot of P. T. Geach’s seminal 1965 paper.
they do in virtue of the “idea” (mental state type) they express; it is just that
ethical sentences express a different kind of mental state from descriptive sen-
tences.3 Recently, a further epicycle of this debate has seen the articulation of
various “hybrid” views that in some way seek to get the best of both views that
ethical sentences express conative attitudes and that ethical sentences express
beliefs by giving the meaning of ethical sentences in terms of both cognitive
and conative states.4
We find the initial ideationalist thought and its development in some hybrid
views rather odd.5 The main source of our puzzlement is that we find it odd
to say that sentences—ethical or not—are in the business of expressing mental
state types (whether cognitive, conative-motivational, or some suitable hybrid).
We think the more natural and conservative idea is that sentences express
propositions. At least this is often seen as the neutral framework in which to
investigate how sentences are semantically composed and what it takes for
good translations of a sentence in one language into other languages. Moreover,
independently of metaethics we would be inclined to say that it is acts of mak-
ing claims that express various mental states (or better, that it is people making
the claims who express their mental states). Appreciating these points in the
context of trying to capture some special internal connection between ethi-
cal claims and motivation leads, we think, to something reasonably regarded
3. For a recent exposition of this idea, see Richard forthcoming.

4. It is important to note that, regardless of the innovations “hybrid” views purport to add to
ethical expressivism, they nonetheless have tended to persist in conceiving of expressivism
as first and foremost a take on the semantic content of ethical claims. According to Daniel
Boisvert’s “Expressive-Assertivism” (2008), for instance, the claim “Tormenting the cat is
bad” expresses both the belief that tormenting the cat has a certain (non-speaker-relative)
property and a negative attitude toward things with that property. The claim therefore has
the following meaning, according to Boisvert(2008, 172): “Tormenting the cat is F; boo
for things that are F!” Likewise, MichaelRidge (2007, 63) once described his “Ecumenical
Expressivism” as “offering a systematic and unified semantics for both asserted and unas-
serted uses of normative predicates” (italics added; see also Ridge 2006). (Though more
recently Ridge [2014] has moved to casting his hybrid expressivism as a view in metaseman-
tics, i.e., as a view about that in virtue of which claims have the semantic contents that they
have. See Chrisman 2014 for an argument that expressivism as a metasemantic thesis is both
a better version of expressivism and a plausible interpretation of the kind of quasi-realist
proposal made in Blackburn 1984).
5. It is especially odd in light of the widely received view among philosophers of language
that the contemporary paradigm of ideationalist semantics, understood as a theory of mean-
ing for natural language—Grice’s theory—faces insurmountable difficulties. (For an attempt
to develop a Gricean semantic theory that overcomes these difficulties, see Davis 2003.)
For an alternative way of construing Grice’s proposal (which is, however, consistent with a
non-Gricean, propositionalist semantics), see Bar-On 1995.
(How) Is Ethical Neo-Expressivism a Hybrid View? 225
as a hybrid expressivist view. But it is a view that—if it hybridizes anything—

hybridizes the notion of expression rather than the types of mental state seman-
tically expressed by ethical sentences. This is the view that two of us previously
defended under the label “ethical neo-expressivism.”6
According to this view, (declarative) ethical sentences do have propositions as
their semantic contents—in this they are like any other (declarative) sentence.
However, it is acts of making an ethical claim that are properly said to express
mental states, and they do so “directly” rather than by expressing somehow
the proposition that one is in a particular mental state. Accordingly, the previ-
ous paper argues for two claims: (i) there are two different but antecedently
quite plausible notions of expression, called s-expression (for “semantic”) and
a-expression (for “action”);7 (ii) if one wishes to capture a special internal con-
nection between ethical claims and motivation in a way that accommodates
the logico-semantic behavior of ethical sentences, it is best to articulate the
idea that ethical claims express motivational attitudes in terms of the notion of
a-expression rather than the notion of s-expression.
This raises two important questions:
• First, the traditional reason for denying that ethical sentences express
propositions was that propositions determine truth conditions, which
are thought to be ways the world could be; so if we reject all analyses
of ethical terms in terms of natural properties, is not anyone who says
that ethical sentences express propositions (e.g., the neo-expressivist)
committed to the Moorean conclusion that ethical sentences describe
sui generis “nonnatural” ways the world could be?
• Second, if we reject the claim that motivational attitudes constitute any
part of the literal or implicated meaning of ethical sentences, are we not
committed to denying the intuition behind motivational internalism
after all? (Or another way of putting it: is not a-expression unsuitable
for capturing internalism?)
We want to use this chapter as an occasion to return to some of the main ideas
in “Ethical Neo-Expressivism” in light of these questions. In particular we want
to explain in more detail the way ethical neo-expressivism adopts an appealing,
metaphysically neutral framework within which to think about the semantics
of ethical sentences. In addition we want to flesh out the notion of a-expression,
explaining why we see it as more basic and more relevant to issues in metaethics
6. In Bar-On and Chrisman 2009.

7. The distinction takes its inspiration from Sellars 1969.
than anything having to do with expressing or conveying the proposition that

one is in a particular mental state. This will also put us in a position to address
a number of critical questions we have encountered.
A P L E A F O R C O N SE R VATI SM I N SE M A NTICS
Independently of any metaethical debate, it is natural to say of any declarative

sentence “S” that it expresses the proposition that-S (bracketing the subtle but
at present irrelevant issues about context sensitivity). The idea of a sentence
expressing a proposition provides an attractive framework for beginning to
think about the literal meaning of sentences and several related issues.8 This is
for at least the following connected reasons, all of which have been traditionally
invoked in support of the need for invoking propositions.
First, the literal meaning of a sentence is surely preserved in a good transla-
tion of that sentence into another language. This is why it is natural to say that
a sentence S1 in one language is a good translation of another sentence S2 in
another language only if S1 and S2 express the same proposition. Expressing
the same proposition may in some cases not be sufficient for good translation,
but it is at least necessary.9 So, for example, good translation cannot be achieved
by simply preserving truth conditions. Clearly, two sentences can have the same
truth conditions and not be good translations of one another.
Second, a declarative sentence type can be tokened on its own in making
an assertion, but it can also be tokened alongside other semantic elements or
embedded in various sentential contexts, and we need some generic way to
keep track of the commonality of content across different contexts. For exam-
ple, the sentence “It is raining” can be used to make an assertion, but it can
also occur as the antecedent of a conditional or be appended to an epistemic
modal—namely, “If it’s raining, then I’ll take an umbrella” and “It might be
raining.” What do all of these tokens of the sentence type “It’s raining” have in
8. The notion of “proposition” we have in mind here is the ontologically neutral notion
of locutionary content rather than the more specific and ontologically committal Fregean
notion (viz., a Platonic object in a “third realm”). In a model theoretical context it might be
represented as a function from the worlds postulated in a semantic model to truth values (or
by some more complex function or set-theoretical object). In our view, whether using such
theoretical models commits one ontologically is a further question beyond the question of
whether such models are useful for understanding various compositional phenomena of the
semantics of a language.
9. Though see Bar-On 1993 for some qualifications and complications that do not bear on
the issues of concern to us here.
common? It may not tell us much, but it surely provides a convenient frame-
work to say that they all share propositional content.
Third, declarative sentences can be used to articulate the object of various
attitudes. For example, you may suspect that it is sunny in Edinburgh, and
I might hope that it is sunny in Edinburgh, in which case the sentence “It is
sunny in Edinburgh” articulates what you suspect and what I hope. How does it
do that? The straightforward answer is that it does that by expressing the propo-
sition that it is sunny in Edinburgh, which is the object of both your suspicion
and my hope.10
Now consider ethical sentences like “Charity is good,” “Not giving to charity
is wrong,” and “Middle-class citizens are obligated to give to charity.” Should we
say that these sentences also express propositions? Well, as far as syntax goes,
these sentences are surely declarative sentences, and we can easily produce
nonethical sentences that would appear to have exactly the same logical form:
for example, “Charity is common,” “Not giving to charity is legal,”“Middle-
class citizens are likely to give to charity.” This suggests that like the nonethi-
cal sentences, the ethical sentences fall under the generalization mentioned
above that, for any declarative sentence “S,” “S” expresses the proposition that-S
(again, bracketing context sensitivity). Moreover, ethical sentences in one lan-
guage seem to admit of good translations into other languages just as much as
nonethical sentences. So insofar as natural criteria for good translation deploy
the notion of two sentences expressing the same proposition, we will want
that notion to apply in the ethical case just as much as in the nonethical case.
Likewise, it seems that ethical sentences can function just like nonethical sen-
tences in semantic embeddings and to articulate the objects of diverse attitudes.
We can say, for instance, “Not giving to charity is wrong” but also “If not giving
to charity is wrong, then I’ll endeavor to give to charity” or “Not giving to char-
ity might be wrong.” Similarly, it seems that you may suspect that middle-class
10. It does not seem plausible to us to suggest that the sentence “It’s raining” expresses the
belief (type) that it is raining without expressing the belief (token) of any particular indi-
vidual. What is in common among attitude ascriptions like (i) John believes that it’s raining,
(ii) John hopes that it’s raining, (iii) John fears that it’s raining, (iv) John doubts that it’s rain-
ing, and (v) John suspects that it’s raining, it seems, is some content and not a type of belief
(or any other attitude) with that content. (Moreover, one might wonder: why single out belief
as the relevant type of attitude that is held constant across attitudes?) What one wants is a
notion of content that is attitude neutral, one that abstracts away from attitude type (as well
as, relatedly, abstracting from speech-act type)—precisely something like the conventional
notion of a proposition understood as Austinian locutionary content or a Fregean thought
(minus the Platonist ontological commitment). (Thanks here to Dean Pettit.) It is revealing
that philosophers otherwise drawn to “mentalist” conceptions of meaning have nonetheless
found reason to resort to essentially abstract and so in that sense nonmental notions of con-
tent (for recent examples, see Davis 2003; Soames 2010).
citizens are obligated to give to charity, while I doubt that middle-class citizens
are obligated to give to charity, in which case the sentence “Middle-class citi-
zens are obligated to give to charity” is a pretty good way to articulate what you
suspect and what I doubt. Again, the straightforward account of how it does
this is that this sentence expresses the proposition that middle-class citizens are
obligated to give to charity.
Does not all of this speak strongly in favor of treating declarative ethical
sentences just as we treat declarative nonethical sentences, in terms of their
expressing propositions? Obviously, yes. But to be clear, we do not view a posi-
tive answer to this question as the end of philosophical semantics but rather as
the insistence on working within a relatively neutral framework when it comes
to the metaphysics and psychology of particular areas of discourse and for
beginning to think more systematically about the literal meaning of declarative
sentences of whatever category. There will surely remain interesting questions
in semantic theory about how various syntactic, contextual, and logical factors
contribute systematically to the propositions expressed by various sentences.
There will still be questions about what semantic contribution is made by rel-
evant subsentential components. There will also remain interesting questions in
developmental linguistics/psychology about how beings like us acquire compe-
tence with the literal meanings of the relevant sentences and their components.
Moreover, there may be further interesting questions in the metaphysics of
semantics regarding what (if anything) a literal meaning is (abstract/concrete,
structured/unstructured, external/internal, etc.). Pursuing these further ques-
tions is perfectly consistent with the idea that declarative sentences, including
those with ethical content, express propositions.11
Note that, on the conservative approach to theorizing about the meanings
of ethical sentences that we are recommending, there is no expectation that it
should be possible to provide a meaning analysis of ordinary declarative sen-
tences by paraphrasing them in other terms. A long history of failures in areas
other than ethics to give paraphrases of sentences containing simple terms
and analyses of atomic concepts should make us leery of any attempt to para-
phrase sentences containing ethical terms in other terms (normative or not).
Ethical sentences are not unique in this regard. According to the semantic
conservatism we advocate, the unavailability of such paraphrastic analyses in
the ethical case is not by itself indicative of some elusive “is-ought” gap but
11. Huw Price (1994; 2013, chap. 1) defends a view he calls “semantic minimalism” that we
find congenial, but we do not think (as he does) that it supports global nonfactualism. The
framework Price offers—in contrast to ethical neo-expressivism—is not designed to address,
specifically, the motivational asymmetry between ethical and nonethical discourse. We leave
discussion of this and other differences for another occasion.
rather of the more general fact that meaningfulness does not require the avail-
ability of paraphrastic analysis. Nor do we need to suppose that the acquisition
and competent use of such terms involves mastery of paraphrastic analyses.
Sentences mean what they do in virtue of expressing propositions, and in
many cases the propositions they express can be specified disquotationally,
at least when the object language and the language in which the meaning is
specified are the same. “John loves Mary” expresses the proposition that John
loves Mary. And it is no different with ethical sentences: “Tormenting the cat
is wrong” expresses the proposition that tormenting the cat is wrong. Thus the
semantic conservatism we adopt allows us to be semantically neutral in the
sense of not being committed to the availability of a reductive meaning analy-
sis for ethical terms and sentences. In general we think that ordinary declara-
tive sentences S can be said to express the proposition that-S, and this includes
declarative sentences with ethical content.12 Of course a full compositional
semantics must say much more about how the parts of ordinary declarative
sentences interact to determine the proposition expressed, which will involve
more than simple disquotation. But the view that declarative sentences express
propositions that can be initially specified disquotationally comports with the
most neutral framework for thinking about the compositional possibilities for
ethical and nonethical terms and sentences.13
All in all we think there are good linguistic reasons not to make any spe-
cial pleading for ethical sentences (or even normative sentences more gener-
ally). Such sentences do not constitute a distinct semantic category, that is, one
deserving a radically different treatment from other kinds of sentences, simply
in virtue of containing ethical terms. (Specifically, there is nothing in the ordi-
nary use of these sentences that marks them as disguised nondeclaratives and
thus as diverging in their linguistic behavior from declarative sentences that
are plausibly taken to have propositional meaning. On the contrary.) Given
the similarities in linguistic behavior between ethical and nonethical terms
and given that ethical terms embed seamlessly in “mixed” contexts as well as
figuring in various nonethical contexts, we ought to prefer adopting the same
basic semantic framework for all declarative sentences, ethical or not. And for
reasons articulated above, we think the propositionalist framework is a good
12. Note that this is consistent with maintaining that meaningfulness requires much more
than the mere possibility of disquotation or syntactic well-formedness.
13. See Chrisman 2012 and Pettit 2010 for suggestions along these lines with respect to par-
ticular terms. The key point is to deny that semantic analysis must, e.g., result in an analytic
paraphrase of some sort, involving lexical decomposition of the relevant terms, or spell out
(nondisquotationally) necessary and sufficient conditions.
place to start. (We note in passing that there is additional linguistic support
for taking the link between descriptiveness and embedability—and the requi-
site notion of propositionality—to be relatively superficial. For many sentences
that are uncontroversially used (at least in part) to make genuinely descriptive
claims—for example, “What a loyal friend you are”—cannot be embedded in
conditionals; “If what a loyal friend you are, then I can trust you” seems no less
illformed than “If Loyal friend! then I can trust you.”)
So what has motivated expressivists of many different stripes to depart from
propositional semantics and undertake wholesale semantic revisionism? In the
main they have assumed this to be the only plausible way to accomplish two
important things expressivism is intended to accomplish:
(a) Avoid commitment to “spooky” irreducibly normative facts

(b) Capture the “internal” connection between ethical claims and
motivation
Re (a): Another traditional role assigned to propositions—one we have not

mentioned above—is being bearers of truth value. Given that and given an
assumed connection between a sentence being truth evaluable and it serving
to describe “the way the world is,” it might be thought that, if one concedes that
an ethical sentence “S” expresses the proposition that-S, then the only way one
could be an antirealist would be to endorse some kind of unpalatable ethical
error theory or fictionalism.14 Re (b): One of the key virtues of expressivism is
supposed to be that it accommodates the motivational internalist’s claim of a
close connection between sincere ethical claims and motivation to act. But to
do so, it is thought, one must build the relevant motivational attitudes into the
meanings of ethical sentences.
However, we view the propositionalist framework we recommend as allowing
one to be neutral on the metaphysical issue underlying (a) above—and neutral
in a way that accommodates antirealism. (See the fourth section of the chapter.)
Moreover, as regards (b) above, we think that whatever “internal” connection
there may be between ethical discourse and motivation, it should be captured
not through the semantic content of ethical sentences—what sentences express,
semantically speaking—but rather through what speakers (or thinkers) express
in acts of ethical claim making. In what follows we take a step back and explain
more carefully how we think of these two different expression relations and
how they figure in defusing the motivations provided by (a) and (b) for expres-
sivists to depart from standard propositionalist semantics. This puts us in a
14. See, e.g., Kalderon 2005; Mackie 1977.

position to offer anyone moved by desiderata (a) and (b) new resources for
meeting them without going in for anything like an expressivist semantics for
ethical sentences.
A -E X P R E S S I O N V E R SUS S- E X PR E SSION
The class of behaviors ordinarily described as “expressive” spans a wide range.15

At one end of the spectrum, we have so-called natural expressions—yelps, gri-
maces, and various gestures—where both the behavior and its connection to
the expressed states are supposed to be inculcated by nature. There are also
mimicked or acquired facial expressions or gestures that become “second
nature,” such as shrugging shoulders or tut-tutting. Then we have conventional
nonverbal expressions, such as tipping one’s hat or sticking out one’s tongue.
(The line here is not sharp; giving a hug, jumping for joy, stomping your feet, for
example, all seem to fall somewhere in between, exhibiting both “natural” and
acquired elements.) Still in the conventional realm, we have expressive verbal
utterances, such as “Darn it!” “Ouch!” “Sorry!” “This is great!” “I hate you!” and
so on. We also find in the verbal domain utterances such as “There’s a crow on
the telephone pole,” which (if sincere) expresses a speaker’s present belief, or
“Let it rain,” which expresses the speaker’s wish for rain. Finally, at the far end
of the conventional side of the spectrum, we have speech acts, such as assertion
or promising, which are alleged to have the expression of certain mental states
as part of their felicity conditions.16
Still, we can discern the following commonality among the expressive
behaviors mentioned so far: they all express states of minds, as opposed to, say,
propositions, concepts, or ideas. This is why we follow Wilfrid Sellars (1969) in
conceiving of this kind of expression as a relation that holds between performers
of acts and the mental states these acts directly express—“a-expression”—and
contrast this with the relation that holds between meaningful strings (e.g., sen-
tence tokens) and their semantic contents. As we are thinking of it, a-expression
is something a minded creature does, be it through bodily demeanor, facial
expression or gesture, or speech, using a natural, conventional, and even idio-
syncratic expressive vehicle.
Thus, for example, we say that the sentence “Snow is white” s-expresses the
proposition that snow is white, and we say that the word “justice” s-expresses
the abstract concept of justice. On the other hand, your dog, when he or she
15. Parts of the present section overlap with Bar-On forthcoming.

16. See Green 2007, 140–143.
gets up and walks over to give you a lick, is a-expressing his or her affection-
ate feeling. And when you give a friend a hug or say “It’s so great to see you”
or alternatively “I’m so glad to see you,” you a-express your joy at seeing your
friend through an intentionally produced act (where some of the acts utilize
sentences that in turn s-express propositions).
In general when one a-expresses a state of mind using a sentence, the sentence
uttered retains its linguistic meaning. “It’s great to see you” and “I’m so glad to
see you” each have their own meaning in virtue of the linguistic rules governing
the lexical items and their respective modes of composition. Each s-expresses a
certain proposition. What proposition? Well, setting aside some nuances about
the context sensitivity of indexicals, it is most natural to say that the former sen-
tence expresses the proposition that it is great to see the addressee and the latter
sentence expresses the proposition that the speaker is happy to see his or her
addressee. It is because they express propositions that these sentences can par-
take in logical inferences and stand in systematic logico-semantic relations to
other sentences (and in particular can be embedded in negation, conditionals,
intensional contexts, etc.). For all that, we think that normal cases of producing
unembedded tokens of these sentence types, in speech or in thought, are cases
of one directly expressing one’s joy.
Intuitively, a-expression is more basic than s-expression. It is certainly more
ubiquitous; nonhuman animals and prelinguistic children express states of
mind through a variety of nonlinguistic means despite not having in their rep-
ertoires the ability to token sentences that s-express propositions. Now it may
be that ultimately (in “the causal order of being”) s-expression has some (com-
plicated, to be sure) relation to a-expression. Perhaps, for example, the conven-
tional meanings of English sentences ultimately originate in Gricean speaker
meanings that have “fossilized.”17 But once conventional meaning is in place,
it is clear that we can separate what a given sentence s-expresses from what
mental states speakers who use the sentence a-express on a given occasion or
even regularly.18
Among linguistic creatures, a familiar acquisition process leads to the
increasing use of more or less conventional means—both words and gestures—
to give vent to present states of mind (making a disapproving face, thinking out
17. As suggested in Blackburn 1984, chap. 3. See also Bar-On 1995.

18. An analogous point can be made regarding thought tokens. Such tokens have semantic
contents. So they too can be said to s-express propositions. (What the tokens s-express is
something that an adequate [psycho]semantics will aim to specify.) But when a given thought
token is produced,we can separate what the individual producing the token a-expresses from
what the token s-expresses.
loud, airing opinions, and so on). The distinction between expressive acts, on
the one hand, and the expressive vehicles used in them, on the other, allows us
to capture underlying action-theoretical similarities between expressing one’s
annoyance through a gesture, an inarticulate sound, or a full sentence while still
acknowledging significant differences.
Now, there is an intuitive contrast between acts of expressing states of mind
and acts of merely telling about them. Anyone can say truly and some can even
tell reliably that DB is feeling sad. But presumably only DB is in a position to
express her sad feeling—for example, by letting tears roll down her cheeks or
saying “This is so sad.” To use earlier terminology, we can say that when you say
“DB is feeling sad” you are employing a sentence that s-expresses the proposi-
tion that DB is feeling sad, and if you are sincere, you are a-expressing your
belief that DB is feeling sad, whereas DB’s tears s-express nothing, though in
letting them roll down she a-expresses the sad feeling itself; and her utterance
“This is so sad” s-expresses the proposition that something sad is happening,
and she can use it to a-express her sadness.
What about DB’s avowal: “I’m feeling so sad”? On the neo-expressivist
view of avowals defended in Bar-On 2004, these are different from evidential
reports on the presence and character of states of mind (whether others’ or
our own) in that they are acts in which we a-express the very state that is
attributed to us by the proposition that is s-expressed by the sentences we use.
One of the central points of defending this idea was that it can help explain
the epistemic asymmetry between avowals and third-person reports of the
same states without compromising the semantic continuity between avowals
and other statements. The asymmetry is not between types of sentences with
certain semantic contents. Rather, the contrast in play here is between acts that
directly express one’s mental state and reports of that state (whoever produces
them and however reliably). (For notably the sentence “I am feeling sad” can
be used by DB as a mere evidential report of her own state, say, at the conclu-
sion of a therapy session.)
For present purposes, what matters is that the distinction between
s-expressing and a-expressing be recognized as a distinction that applies across
all areas of discourse and regardless of what semantic, epistemological, or meta-
physical analysis we adopt for the relevant domain. On the view we advocate,
mental states are indeed the relata of an expression relation. And it is a relation
that may well be relevant for understanding of various linguistic acts we per-
form. But the relation in question is what we are calling a-expression. It is not
the expression relation that holds between sentences and propositions or words
and concepts (i.e., s-expression).
The relevance of this here is that we think the expressive character of ethical
claims and their apparently tight connection to motivation can be explained
in partial analogy to the neo-expressivist treatment of avowals mentioned ear-

lier.19 According to avowal neo-expressivism, the epistemic asymmetry between
avowals and (even self-)reports is due to the fact that in acts of avowing one
gives direct vent to the very state that the sentence produced self-ascribes.
Similarly, ethical neo-expressivism maintains that ethical sentences s-express
straightforward propositions (which can be specified, at least initially, disquo-
tationally), and this allows us to preserve their semantic continuity with other
sentences. However, what we might describe as the motivational asymmetry
between ethical discourse and ordinary descriptive discourse can—on anal-
ogy to the epistemic asymmetry between avowals and reports—be captured
(according to ethical neo-expressivism) by appeal to the expressive character of
acts of making ethical claims.
So on our view expressivists have been correct to identify the expressive
function of ethical discourse and reflection; if internalism is correct, then
it is in fact part of the “job” of ethical claims (whether made in speech or in
thought) to express certain of our motivational attitudes.20 Armed with the dis-
tinction between s-expression and a-expression, we think we can retain this
key idea while avoiding a host of difficulties faced by traditional expressiv-
ism and later developments of it. As in the case of avowals, we can distinguish
between the act of making an ethical claim and the vehicle used in making it.
The vehicle used in claiming “Tormenting cats is morally wrong” is a sentence
(or thought) token that employs an ethical term (or concept); the token can be
said to s-express a true or false proposition. However, as with avowals, what
is s-expressed by a claim does not settle even what mental state is character-
istically a-expressed by acts of making the claim let alone what mental state is
a-expressed on a given occasion of producing the token. With Humeans, one
may think that purely cognitive states (such as beliefs) cannot by themselves
motivate or explain action, and furthermore one may think there is an intimate
connection between sincerely making an ethical claim and being motivated to
act (or refrain from acting) in accordance with it. The neo-expressivist main-
tains that in making ethical claims, we a-express the very same states whose
presence is required for understanding the perceived motivational force of such
claims.21 As Ayer already saw,22 this expressivist insight is best captured without
19. Following Bar-On and Chrisman 2009.

20. Below we try to spell out how the neo-expressivist can respect the internalist take on this
claim.
21. They do so whether or not they also express a belief whose content is given by the
s-expressed proposition. For some discussion, see Bar-On and Chrisman 2009.
22. Ayer 1946, 104–108.
supposing that the vehicles used in making ethical claims s-express proposi-
tions that self-ascribe those states. (So it is important to note that we are not
claiming that ethical claims are themselves avowals.)
Thus as long as we are talking about expressive acts, we can agree with Ayer
that ethical claims betray or show motivational attitudes not because the claims
report them23 but because those who make the claims (a-)express the attitudes
directly. What the neo-expressivist goes on to add is that we need not endorse
Ayer’s view that ethical claims are not truth evaluable, because they do not (s-)
express any true or false propositions. According to ethical neo-expressivism,
then
(i) As a species of evaluative claims, ethical claims understood as

acts are different from ordinary descriptive claims in that agents
making them (in speech or in thought) a-express motivational
attitudes.
(ii) The vehicles used in making ethical claims—typically ethical
sentences—are semantically continuous with ordinary descriptive
sentences in being truth evaluable, embeddable in truth-functional
as well as intensional contexts, logical inferences, and so on. This is
because they s-express true or false propositions.
E T H I C A L N E O - E X PR E SSI V I SM :
A N T I R E A L I S M , I NTE R N ALI SM
In the previous section we argued that the idea that ethical claims are distinc-
tive in that they express motivational attitudes can be captured without build-
ing attitude expression into the semantic content of ethical sentences. But doing
so was of course supposed to allow expressivists to accommodate ethical anti-
realism and to capture motivational internalism. These were desiderata (a) and
(b) from the end of the second section. How does a neo-expressivist view pro-
vide resources for satisfying them?
Re (a): Recall that above we distinguished propositions from truth condi-
tions but said that propositions may determine the truth conditions of the sen-
tences that express them. There is a pervasive tendency in metaethical debate
to combine this idea with a metaphysically inflated conception of truth and
23. And we might add, not even because they imply that one has them. (This in contrast with
the implicature-style hybrid views in Copp 2001, 2009; Finlay 2005. For discussion of the
contrast, see Bar-On and Chrisman 2009.)
to interpret anyone who thinks a sentence expresses a proposition as commit-

ted to there being a way the world might be that would make this proposi-
tion true. While we recognize that many metaphysically inclined philosophers
are attracted to this picture, we deny that it is essential to the propositional-
ist framework for philosophical semantics. Notice that none of the reasons we
gave earlier for assigning propositions to sentences as their semantic contents
(translation, embedding, common content of diverse attitudes) trades on any
particular conception of truth as correspondence with the world or of truth-
conditions as ways the world might be. In our view semantics is not the place to
explain how reality is or might be, it is the place to systematize our understand-
ing of things like sameness of meaning across different languages, common-
alities of content under embeddings in force-stripping contexts, and the way
sentences can articulate contents toward which diverse attitudes may be taken.
Because of this we want to allow that it makes sense to speak of an ethical
sentence expressing a proposition prior to settling on the correct metaphys-
ics for ethical discourse. Thus we are assuming that it can make sense to
assign, for example, “Tormenting the cat is wrong” the meaning that tor-
menting the cat is wrong, for logico-semantic purposes, independently of
determining what, if anything, constitutes the nature or essence of moral
wrongness (or torment or cats for that matter) or whether moral wrongness
constitutes a metaphysically genuine property (natural or otherwise). The
important point is just this: an expressivist can accept the sort of semantic
conservatism we are here encouraging and still avoid any commitment to
“spooky” irreducibly normative properties. Semantic conservatism is consis-
tent with ethical antirealism, since it implies nothing about the existence or
nature of moral reality. Given this neutrality, it is left for further metaphysi-
cal investigation to determine whether there are ethical properties or states
of affairs and, if there are, what their nature is and to what extent they are
mind (or judgment) dependent.24
Some, however, will worry that the sort of semantic conservatism that we
here encourage offers scant resources for explaining what sentences in a given
area of discourse are about. To say that ethical sentences s-express “innocent”
disquoted propositions is to offer a very thin notion of the sentences’ mean-
ings. If one wanted to understand what such sentences “really say” or what
they are “really about,” then one would apparently have to beef up the account
24. We are claiming that, in the ethical case (unlike in the avowal case; see Bar-On 2012), the
possibility that there are no “truthmakers” for the relevant propositions should be left open
to allow for a meaningful dispute between ethical realists and antirealists who can agree
on the availability of a disquotationally specified proposition as a semantic starting point.
(Note that to say this is not to commit to a disquotational theory of the truth of ethical—or
other—claims.)
of the sentence’s s-expressed content. But then it is no longer possible to main-

tain the sort of semantic conservatism that we prefer—either one goes real-
ist or one goes traditional expressivist, or subjectivist, or whatever. Or so the
objection might go.25
In response we would begin by pointing out that, as far as the semantics of
ethical sentences is concerned, it is not as if all we can say is that they s-express
“innocent” disqouted propositions. As explained earlier, there is still plenty of
interesting work left for the semanticist to do: arguably she ought to explain
how the proposition expressed by an ethical sentence is a systematic function
of the semantic values of the parts of the sentence, how these are composed in
its logical form, and so forth. She can seek systematic integration of that piece
of her semantics with other pieces, such as her view about embedding in indica-
tive conditionals, under epistemic modals, or in propositional attitude reports.
She can posit covert context sensitivity or deny it. She can debate the logical
form of the sentence and ones syntactically like it. What she cannot do, how-
ever—without abandoning the sort of ontological innocence that we would like
to maintain—is to say what in reality the sentence is about. Our view, however,
is that the thought that a propositionalist semantics carries specific ontological
commitments is a significant step beyond the truth-conditionalist framework
that has proven so fruitful in compositional semantics.
We join Simon Blackburn (1998, 79) in thinking that “the ethical proposition
is what it is and not another thing.”26 If we understand him correctly, his point is
just that when asked what a sentence like “Tormenting the cat is wrong” is really
about, one should simply say that it is about tormenting the cat being wrong. To
say anything further would be to risk running afoul of the open question argu-
ment and/or queerness worries. Taking such a conservative approach to the
semantics of ethical sentences allows us to capture all the virtues of traditional
propositional-compositional semantics while remaining totally silent on the
appropriate metaphysics for ethical claims.27 And it is this silence that allows us
to accommodate ethical antirealism.
Perhaps a further thing to say in response to this objection would be to ques-
tion how coherent it ever was to ask what sentences in a given area of discourse
are about. In particular we wonder what sense can be made of the notion of
25. Richard forthcoming provides reasons for expecting semantics to yield assignments of
meaning that go beyond “innocent” propositions.
26. Blackburn here adapts a famous line from Joseph Butler via G. E. Moore (1993,
29): “Everything is what it is, and not another thing.”
27. This is an instance of the sort of problem Price (2004) refers to as “placement problems.”
See also Price 2013, chap. 1,sec. 4, for a helpful discussion of the origin of these problems
and how to diffuse them.
“sentences in a given area of discourse,” especially when it is noted that such

sentences can typically be embedded rather easily into force-stripping contexts,
some of which will be logically complex sentences with “mixed” parts—that
is, parts of which belong in the supposed domain and parts that do not. For
instance, “If he φ-ed, then he ought to be punished” seems like a sentence
belonging to the domain of ethics, but what about “If he ought to be punished,
then he φ-ed”? And what about sentences like “It’s either morally wrong to φ
or it isn’t”? Such possibilities and the ease with which they are entertained may
tell against the very notion of a given domain. And if sufficient sense cannot be
made of this notion, then it would appear as if all that remains are particular
meanings of particular sentences to be determined lexically and composition-
ally—an outcome with which we are perfectly comfortable.
Furthermore, our semantic conservatism allows us to remain neutral on the
nature of the attitude expressed when making ethical claims. Any number of
attitudes can do as long as it can be argued that the relevant attitude is suitably
linked to motivation (a matter left to moral psychology). Again, we see it as an
advantage of our view that it is not forced to settle in the semantics of ethical
discourse what attitudes agents express when employing ethical vocabulary.28
Some may wonder, however: even if ethical neo-expressivism does not owe
us an account of the states expressed when making ethical claims, does it not
owe us an account of the connection between these states and ethical claims
understood as products? After all, in the case of avowals the connection seems
pretty straightforward: the sentence typically explicitly mentions the very state
that is a-expressed. So what accounts for the connection between a sentence
like “Tormenting the cat is wrong” and, say, disapproval if the latter does not
figure into the semantic content of the former?
Here again, there are two things to say. First, it is a mistake to think that
the connection between mental states and avowals is simply a matter of the
sentences typically explicitly mentioning the relevant states. Consider two sen-
tences from our earlier discussion of a-expression:
(1) “It’s so great to see you!”

(2) “I’m so glad to see you!”
As we explained before, these sentences can be used as vehicles for the

a-expression of one and the same mental state—being glad or feeling joy.
However, if we were to explain the connection between the sentence and the
mental state in (2) in terms of the sentence explicitly mentioning the relevant
28. This problem has been called “the specification problem” for expressivism. See Björnsson
and McPherson forthcoming; Köhler 2013.
state, we would apparently have to tell some other kind of story about how the
same connection is forged in (1). Instead, we think this connection is forged in
the same way for both (1) and (2) as well as for other vehicles for the a-expression
of being glad, such as a hug. We prefer to account for this connection by mak-
ing reference to the conditions underlying competence with the use of expres-
sive vehicles in performing relevant types of expressive acts. In the case of both
(1) and (2), for instance, competence with the sentence requires a speaker to
know that they are fit vehicles with which to a-express being glad, feeling joy,
and the like. The same goes for acts such as hugging. (With a little imagina-
tion, the reader could envision various cases of misplaced, poorly timed, or
otherwise inappropriate hugs and ask what sorts of conclusions we would draw
about the hugger’s competence.)29 Explicitly mentioning the relevant state is
neither necessary nor sufficient for establishing, mastering, or preserving the
expressive link to it in an expressive vehicle.30
Second, going back to the question of how ethical claims in particular come
to be linked to certain motivational states, we think it instructive to consider
the case of slurs and other “linguistic expressives,” since these too are terms
(or phrases) with a clear connection to certain mental states (though, unlike
avowals, the connection is not explicit). One can a-express anger toward John
by saying “I’m angry with John!” or by saying “John is such an a-hole!” Here
again, one of the sentences “s”-names the relevant attitude, but both are obvi-
ously effective (and common) as expressive vehicles for one’s anger. And this is
because competence with the term “a-hole” requires implicit knowledge that
it is a fit vehicle with which to a-express anger (or related sentiment). To this
extent, we agree with hybrid expressivists who think that, for example, pejo-
ratives and slurs nicely illustrate how terms and sentences can be bound up
with certain noncognitive attitudes, despite bearing no explicit connection to
the relevant states.31 Where we part ways, however, is in our denial that the
noncognitive attitudes somehow figure into the semantic contents of expressive
terms or of the sentences containing them. Though there may be much to be
29. See Bar-On 2004, 320ff., 419ff., for relevant discussion of what she calls “expressive fail-
ures,” which can arise even in the case of inadvertent expressive behaviors, such as yelps or
grimaces.
30. That it is not sufficient can be clearly seen from the fact that mental states are regularly
named in sentences that are not used in acts of (a-)expressing them.
31. One finds references to slurs throughout the literature on hybrid expressivism, as many
think they nicely embody the basic point of hybrid theories. David Copp (2009, 169–170),
for instance, writes that ethical terms “are similar to pejorative terms in that their use can
both ascribe a property and express a relevant conative attitude.” See also Boisvert 2008; Hay
2013; Schroeder 2013.
said about the state of mind characteristically (a-)expressed by (proper) uses of

“a-hole” (and though there may be interesting things to say about what makes
it the case that someone is an a-hole), “John is an a-hole!” can still be said to
s-express the proposition that, well, John is an a-hole. (And it is in virtue of this
that the sentence can be embedded in a variety of contexts, partake in logical
inferences, etc.)32
Re (b): According to internalists, one who makes a claim such as “Charity
is good” competently cannot be completely indifferent to charity, whereas one
who makes the grammatically similar claim “Charity is common” can. There
is little consensus, even among self-avowed internalists, about the character of
the “cannot be” here, but we would still like to outline briefly how our view can
accommodate an internal connection. No doubt there will be internalists who
deem the connection insufficiently tight. But it is far from clear that any tighter
connection is defensible.33
Earlier we mentioned the neo-expressivist explanation of the epistemic
asymmetry between avowals and evidential reports: someone who is avowing
feeling annoyed—as opposed to reporting annoyance in consequence of, say,
therapy—is taken to engage in an act of direct expression of annoyance (and
not simply expressing the belief that she or he is annoyed). Though it is not
conceptually impossible for someone to make a mental self-ascription and not
be in the self-ascribed state, it is a propriety condition on avowing that one is in
the self-ascribed mental state. The neo-expressivist explanation of the motiva-
tional asymmetry between ethical claims and descriptive claims can be seen as
analogous. Someone who is making an ethical claim (as opposed to producing
a descriptive report of some state of affairs) is a-expressing the relevant moti-
vational attitude—the very attitude whose presence would explain why she or
he is suitably motivated. Even if it is not conceptually impossible for someone
to make an ethical claim without having the relevant attitude, having the atti-
tude can still be a propriety condition on making (genuinely) ethical claims. So
someone who makes the claim while lacking the attitude violates a propriety
condition on acts of ethical claim making. And adequate mastery of ethical dis-
course and ethical concepts requires grasping this propriety condition.34 This
32. Note that in the case of linguistic expressives it may be plausible to suggest that, e.g.,“John
is such an a-hole” is more or less synonymous with “What an a-hole!” (applied to John). Yet
the former can—but the latter cannot—be embedded in a conditional, or serve as a premise
in an instance of modus ponens, and so on.
33. See Björnsson and Francén Olinder 2013 and Fletcher unpublished for critiques of stron-
ger forms of internalism. See Björklund et al. 2012 and Fletcher 2014 for helpful discussion
of the many different forms and strengths of internalism and its dialectical role in metaethi-
cal debate.
34. Again, thinking of slurs and other linguistic expressives is instructive here.
arguably provides resources for capturing a fairly strong internal connection

between ethical claims and action as well as providing a more nuanced array
of diagnoses of different ways the connection between making an (apparently)
ethical claim and motivation can be broken.35
Propriety conditions lay down norms for what counts as doing things prop-
erly. For example, a propriety condition on making promises may be that one
has an intention to do as the promise says. This means that if one has made a
promise without having the relevant intention, one has failed to properly issue
a promise. Note that improperly making a promise is not the same as not mak-
ing a promise at all. An actor on stage who says “I promise” is only making a
pretend promise; a parrot or a not-yet-competent speaker who utters the words
is making no promise at all rather than pretending to. Does this mean that it
is conceptually impossible for one to say “I promise” meaningfully without hav-
ing the relevant intention? Of course not, since it is possible to say “I promise”
insincerely. But notice that we are willing to credit the insincere speaker with
understanding of and semantic competence with the relevant term despite her
failing to have the requisite intention. What licenses this? It is the fact that we
have some independent grounds for crediting her with this receptive and pro-
ductive competence. (Indeed such competence must feature in the explanation
of her insincere performance.) But now suppose we have a speaker who is not
insincere; she says “I promise” but fails to have the relevant intention—perhaps
due to some temporary or chronic psychological feebleness. Still, we may have
reason to credit her with competence. For example, she may exhibit full aware-
ness of the propriety conditions on the practice of promising; she just fails to
meet them. Of such a speaker, we would want to say that she has made a prom-
ise, albeit an improper one.
In the ethical case we are similarly inclined to say that it ought to be con-
ceptually possible to have a speaker (or thinker) who has achieved competence
with ethical vocabulary but who makes an ethical claim while still failing to have
the relevant attitude. This, however, does not commit us to saying that one can
achieve competence with the ethical vocabulary without grasping the propriety
condition mentioned earlier. Moreover, suppose we come across a whole com-
munity whose speakers appear to use ethical vocabulary but who never have
the appropriate attitudes. Given semantic conservatism, we should allow that
the sentences used by members of this community may have the same propo-
sitional contents as our ethical sentences. But this does not settle the question
whether they are actually making ethical claims (ever). If their discourse is gov-
erned by entirely different propriety conditions, then it is not ethical discourse.
35. For further discussion, see Bar-On and Chrisman 2009, 143–150.
Compare: If in some twin community there is no connection whatsoever

between the use of a type of sentence and trying to amuse/being amused, that
tells against interpreting the use of that type of sentence as “telling a joke.” Still,
this would not be reflected in the translation of individual jokes. What makes an
utterance an act of making a joke (involvement in a practice governed by certain
norms) need not be read off the meaning of individual sentences used as jokes.
Similarly, we want to say, what makes a claim an ethical claim need not be read
off the meaning of individual sentences used as ethical claims.
An analogy may be helpful here. Suppose I am playing a game of chess, and
I move the rook diagonally. We might say either that I have made an improper
chess move or that I have not made a chess move at all. There is good rea-
son to say the former. To “make a chess move” is just to make a move while
engaged in a certain sort of practice, which is itself governed by rules. This is
the difference between you moving the rook diagonally while playing a game
of chess and say, a one-year-old child grabbing a rook and moving it elsewhere
on the board (even if perfectly in line with the rules). The child is not playing
chess; you are. Similarly, the psychopath, when he confesses that he behaved
wrongly—despite not feeling guilty, regretting his actions, and so forth—is still
making an ethical claim, though he makes his claim improperly. Genuinely
making ethical claims is to be contrasted with things like using moral/evalu-
ative language while acting. Regardless of whether the actor on stage has the
relevant attitudes, she is not making a genuinely ethical claim—she is just pre-
tending to engage in ethical discourse. Similarly for using moral/evaluative
language sarcastically, playfully, and so forth. For example, saying sarcasti-
cally “Good job, Ryan!” or saying playfully “You’re so bad.” These are not cases
of people using the language improperly but rather cases of not making any
genuine ethical claims at all. By contrast, it does not seem plausible to think
that Ted Bundy was pretending to make ethical claims, nor was he using the
language playfully or sarcastically. As far as he was concerned, he was just as
engaged in the practice of moral evaluation as we are when we morally judge
his behavior—and so we would perhaps affirm that he is making an ethical
claim. But on the assumption that Bundy felt no remorse (etc.) at all, he was
nonetheless doing so improperly.36
36. Note that Bundy’s case could be assimilated to the “deviant community” case if there
were reasons to think he simply has not cottoned on to the propriety conditions of ethical
discourse. It depends on the case.
C O N C L U S I O N : NE O - E X PR E SSI V I SM
A N D H Y B R I D E X PR E SSI V I SM ?
Protoexpressivism flounders on the challenge of developing a systematic

account of the meaning of arbitrary sentences out of the essentially bifurcated
thought that ethical sentences express noncognitive attitudes while nonethical
sentences express propositions. The response to this challenge from mainline
expressivists of the 1960s through the 1990s was to suggest that the meaning of
all sentences is to be given in terms of the attitudes they express; it is just that
descriptive and ethical sentences express importantly different kinds of atti-
tudes: beliefs versus desires (or other motivational states). As we mentioned
at the outset, more recent “hybrid theorists” have suggested that this approach
still faces a number of difficulties stemming from the semantic continuities
between descriptive and ethical sentences, and these might be overcome by
suggesting that although the meaning of all sentences is to be given in terms of
the attitudes they express, ethical sentences express a distinctive hybrid attitude
constituted by both belief-like and desire-like elements. Roughly speaking, the
belief-like attitude is supposed to capture the semantic continuities between
ethical sentences and descriptive sentences, while the desire-like attitude is sup-
posed to retain the original benefits of mainline expressivism.
In light of this it is worth asking, is the neo-expressivist view we have been
discussing here a “hybrid” view? It might seem that way, since on the face of
it we maintain that when someone makes an ethical claim (in speech or in
thought) two different things get expressed: a proposition and the speaker’s
or the thinker’s motivational attitude. But as should by now be clear, this is
misleading, since on our view the two things are expressed in two different
senses—the proposition is s-expressed by the relevant sentence (or thought)
token, whereas the motivational attitude is a-expressed by the individual mak-
ing the claim. Importantly, in sharp distinction from other hybrid views, we
do not maintain that the “two things” constitute two parts of the meaning of
ethical sentences. Nor do we locate the attitude expression in a conventionally
implicated or conversationally inferable proposition that self-ascribes the atti-
tude, as do implicature views.37 Thus rather than pursuing a nonconservative
semantics for ethical sentences and then hybridizing the mental state said to be
semantically expressed by ethical sentences, ethical neo-expressivism proposes
that we use the more basic and less controversial distinction between what sen-
tences express (i.e., propositions) and what people express (i.e., mental states)
to capture the internalist intuitions but in a way that allows us to remain neutral
37. See Copp 2001, 2009; Finlay 2004, 2005. Bar-On and Chrisman 2009, 150–158, compares
and contrasts ethical neo-expressivism with these implicature views.
on the metaphysical issue of realism as well as on the moral psychological issue

of cognitivism.
The notion of a-expression seems more general and more basic than any
of the following: (i) semantically expressing a proposition; (ii) semantically
expressing a mental state; (iii) stating, implicating, or otherwise conveying the
proposition that one is in a particular mental state. This is why we think it is
the appropriate place to look to explain the supposedly distinctive way ethical
claims express attitudes. Moreover, we take the notion of s-expression to estab-
lish a framework for thinking about semantic content within the constraints
of compositionality and the semantic competence of speakers. However, it
is a framework that we take to be attractively neutral on metaphysical and
moral-psychological issues, which is why we think it is preferable for anyone
seeking to accommodate ethical antirealism and to capture motivational inter-
nalism. This is why neo-expressivism offers a nice way to retain one of the core
insights of traditional expressivism (that acts of making an ethical claim “have
the job” of expressing attitudes distinctively connected to motivation) while
avoiding one of the main pitfalls of many later versions of expressivism (result-
ing from abandoning propositionalist explanations of the semantic content of
ethical sentences). We close by responding to three potential worries.
The first worry is that neo-expressivism does not meet desiderata (a) and
(b) from the second section (avoiding spooky facts and accommodating the
internal connection to motivation) in the semantics of ethical sentences and it
does not meet them in the pragmatics of ethical discourse, so it is unclear what
part of the view of ethical language does allow us to meet them. Our answer to
this is that these desiderata should be met in the theory of ethical claim making,
including the propriety conditions on such acts. We are suggesting that there
is a particular kind of act (both in public speech and in private thought) whose
propriety is linked to having certain motivational dispositions, and this link is
what explains whatever distinctive internal connection there is between ethical
claims and motivation.
The second worry can be presented as a critical question from hybrid
expressivists like Michael Ridge who think our view is not so different from
theirs: “Even if you do not want to endorse it as your ‘semantics’ of ethical
sentences, are you not also committed to hybridizing not only the expres-
sion relation but also the mental state expressed (you will say “a-expressed”)
by ethical claims? To capture their psychological continuity with other
claims, you will have to say that ethical claims a-express beliefs, but to cap-
ture the difference in their motivational potentials, you will have to say
that they also a-express motivational attitudes.” Our response is that ethical
neo-expressivism itself is neutral on the psychology of motivation and on the
extent to which ethical thoughts are psychologically continuous with other

kinds of thoughts. It is open to us to say ethical claims a-express beliefs with
ethical content, which as a matter of their psychological nature have a dis-
tinctive motivational capacity compared to other sorts of belief, but we can
also say that, in light of the Humean distinction between cognitive and cona-
tive attitudes in the psychology of motivation, acts of making ethical claims
a-express both. Although ethical neo-expressivism was originally framed
against the background of this Humean distinction, which is a key tenet of
many forms of expressivism, the view is also compatible with anti-Humean
accounts of motivation.
A final worry concerns ethical claims made in thought rather than in speech.
Many find mysterious the idea that inner thoughts are subject to propriety
conditions about what mental state one should be in—for how would those
propriety conditions get enforced, where would they come from, and what evi-
dence could there be of their existence? In response we would stress that our
view is that it is part of what it is to make an ethical claim, whether in speech
or in thought, that one has done so inappropriately if one is not in the relevant
motivational state. This “comes from” the point and function of a practice of
ethical claim making. Although the rules of that practice may be most easily
enforced when it is public, we see no reason that these rules could not be inter-
nalized and be thought to govern “silent” acts of ethical claim making. We take
this to be a mundane phenomenon illustrated by other cases of expressing in
thought, such as making promises to oneself (or to God), avowing, or using
various expressives. Thus all in all we think understanding the sense in which
ethical claims express motivational states along our proposed neo-expressivist
lines should allow expressivists to meet some of the main desiderata that origi-
nally motivated expressivists while avoiding the main difficulties besetting
traditional expressivism and its heirs and while maintaining neutrality on key
debatable issues.
References
Ayer, A. 1946. Language, Truth, and Logic. 2nd ed. London: Gollancz.

Bar-On, D. 1993. “Indeterminacy of Translation: Theory and Practice.” Philosophy and
Phenomenological Research 53: 781–810.
———. 1995. “ ‘Meaning’ Reconstructed: Grice and the Naturalizing of Semantics.”
Pacific Philosophical Quarterly 76:83–116.
———. 2004. Speaking My Mind: Expression and Self-Knowledge. Oxford: Clarendon.
———. 2012. “Expression, Truth, and Reality: Some Variations on Themes from Wright.”
In Mind, Meaning, and Knowledge: Themes from the Philosophy of Crispin Wright, ed.
A. Coliva, 162–194. Oxford: Oxford University Press.
———. Forthcoming. “Expression: Acts, Products, and Meaning.” In Minimalism,

Pragmatism, Expressivism: Essays on Language and Metaphysics, ed. S. Gross,
M. Williams, and N. Tebben. Oxford: Oxford University Press.
Björklund, F., G. Björnsson, J. Eriksson, R. Francén Olinder, and C. Strandberg.
2012.“Recent Work on Motivational Internalism.” Analysis 72:124–137.
Björnsson, G., and R. Francén Olinder. 2013.“Internalists Beware—We Might All Be
Amoralists!” Australasian Journal of Philosophy 91:1–14.
Björnsson, G., and T. McPherson. Forthcoming. “Moral Attitudes for
Non-Cognitivists: Solving the Specification Problem.” Mind.
Blackburn, S. 1984. Spreading the Word: Groundings in the Philosophy of Language.
———. 1998. Ruling Passions: A Theory of Practical Reasoning. Oxford: Clarendon.
Boisvert, D. 2008.“Expressive-Assertivism.” Pacific Philosophical Quarterly 89:169–203.
Carnap, R. 1935. Philosophy and Logical Syntax. London: Kegan Paul, Trench, Trubner.
Chrisman, M. 2012. “On the Meaning of ‘Ought.’ ” In Oxford Studies in Metaethics, vol.
7, ed. R. Shafer-Landau, 304–332. Oxford: Oxford University Press.
———. 2014. “Attitudinal Expressivism and Logical Pragmatism in Metaethics.”
In Pragmatism, Law, and Language, ed. G. Hubbs and D. Lind, 117–135.
New York: Routledge.
Copp, D. 2001.“Realist-Expressivism: A Neglected Option for Moral Realism.” Social
Philosophy and Policy 18:1–43.
Davis, W. 2003. Meaning, Expression, and Thought. Cambridge: Cambridge University
Press.
Finlay, S. 2004. “The Conversational Practicality of Value Judgement.” The Journal of
Ethics 8: 205–223.
———. 2005. “Value and Implicature.” Philosophers’ Imprint 5: 1–20.
Fletcher, G. 2014. “Moral Judgment Internalism: Its Scope, Status and Significance.”
Accessed April 22, 2014, https://edinburgh.academia.edu/GuyFletcher.
Geach, P. 1965. “Assertion.” Philosophical Review 74:449–465.
Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Green, M. 2007. Self-Expression. Oxford: Oxford University Press.
Hare, R. 1952. The Language of Morals. Oxford: Oxford University Press.
———. 1963. Freedom and Reason. Oxford: Oxford University Press.
Language.” European Journal of Philosophy 21: 450–474.
Jackson, F. 2001.“What Is Expressivism?” Philosophical Books 42:10–17.
Kalderon, M. 2005. Moral Fictionalism. Oxford: Clarendon.
Köhler, S. 2013. “Do Expressivists Have an Attitude Problem?” Ethics 123:479–507.
Mackie, J. 1977. Ethics: Inventing Right and Wrong. London: Penguin.
Moore, G. 1993. Principia Ethica. Rev. ed. Cambridge: Cambridge University Press.
Pettit, D. 2010. “The Semantics of ‘Good.’ ” Accessed April 9, 2014, http://www.aca
demia.edu/1045862/The_Semantics_of_Good.
Price, H. 1994. “Semantic Minimalism and the Frege Point.” In Foundations of Speech
Act Theory: Philosophical and Linguistic Perspectives, ed. S. Tsohatzidis, 132–155.
London: Routledge.
———. 2004. “Naturalism without Representationalism.” In Naturalism in Question,
ed. D. Macarthur and M. de Caro, 71–88. Cambridge, MA: Harvard University
Press.
———. 2013. Expressivism, Pragmatism, and Representationalism. Cambridge: Cambridge
University Press.
Richard, M. Forthcoming. “What Would an Expressivist Semantics Be?” In Minimalism,
Pragmatism, Expressivism: Essays on Language and Metaphysics, ed. S. Gross, M.
Williams, and N. Tebben. Oxford: Oxford University Press.
Ridge, M. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116: 302–336.
———. 2007. “Ecumenical Expressivism: The Best of Both Worlds?” In Oxford Studies
in Metaethics, vol. 2, ed. R. Shafer-Landau, 51–76. Oxford: Oxford University
Press.
Schroeder, M. 2008. Being For: Evaluating the Semantic Program of Expressivism.
———. 2013. “Tempered Expressivism.” In Oxford Studies in Metaethics, vol. 8, ed. R.
Shafer-Landau, 283–314. Oxford: Oxford University Press.
Sellars, W. 1969. “Language as Thought and as Communication.” Philosophy and
Soames, S. 2010. What Is Meaning? Princeton, NJ: Princeton University Press.
Stevenson, C. 1937. “The Emotive Meaning of Ethical Terms.” Mind 46:14–31.
———. 1944. Ethics and Language. New Haven, CT: Yale University Press.
11
Why Go Hybrid?
A Cognitivist Alternative to Hybrid Theories
of Normative Judgment
LAURA SCHROETER AND FRANÇOIS SCHROETER n
Why go for a hybrid model of normative judgment? Why not be a pure cogni-
tivist instead and claim that the semantic function of normative predicates is
just to attribute a property? A standard objection to pure cognitivism is that it
cannot adequately explain the distinctive action-guiding role of moral judg-
ments and of normative judgments in general. On a cognitivist approach, we
are told, the link between normative judgment and motivation will be purely
external and contingent.1
In this chapter we outline a new cognitivist account of normative judg-
ment. The account is grounded in a general model of concepts, according to
which concepts are individuated by something like anaphoric relations among
thoughts. Although the model applies across the board to concepts about any
topic, we believe it can help explain why the link between normative judgment
and motivation is not purely external and contingent on a pure cognitivist
account of the semantic values of normative judgments.
Recent debates about hybridism and the alleged deficiencies of pure cog-
nitivism tend to focus on the level of linguistic communication. Hybridists
1. Expressivists have long argued that cognitivist accounts of normative judgments cannot
adequately explain their distinctive action-guiding and motivational role. More recently they
have been joined by a growing contingent of hybrid theorists who contend that a defensible
account of normative judgment should construe simple normative assertions as expressing
both a belief and a desire-like attitude. See, for instance, Barker 2002; Boisvert 2008; Copp
2001, 2008; Ridge 2006, 2007. See also Schroeder 2009 for an overview and a discussion of
hybrid approaches to moral and normative judgments.
Why Go Hybrid? 249
appeal to conventional or conversational implicature, for instance, to explain

the link between normative judgment and motivation. On such accounts the
internal link between normative judgment and normative motivation is forged
by linguistic conventions or by norms for felicitous communication. In con-
trast, our approach focuses on the level of thought. The internal link between
normative judgment and motivation is forged by normative concepts deployed
in those beliefs.
We will start with conceptual structure (first section). We will then show that
a proper appreciation of the relational nature of conceptual structure leads to a
new account of concept identity (second section). This will put us in a position
to show how cognitivists can explain why a link between normative judgment
and motivation is not purely external or contingent (third section).
C O N C E P T U A L STR UCTUR E
There is a fundamental distinction at the level of thought between representing

the same and representing as the same. If you see the same abstract painting on
two separate occasions, you may register exactly the same visual details. Even
so, you may not see the painting as familiar on the second occasion—you do
not recognize it as the same.2
When you redeploy a concept in a train of reasoning, however, your
thoughts automatically present themselves as pertaining to the same topic.
You may read about Jackson Pollock’s Blue Poles, for instance, and then won-
der what it looks like, who owns it, and what it is worth. From your perspec-
tive as a thinker, it seems obvious and incontrovertible that there is just one
topic in question. We will call this the appearance of de jure sameness of
topic.3 Not all thoughts that you take to pertain to the same topic exhibit
this distinctive epistemic appearance. You may learn to your surprise, for
instance, that Pollock’s Blue Poles is identical to Pollock’s Number 11, 1952. In
this explicit identity judgment, sameness of topic seems neither obvious nor
incontrovertible.
2. This distinction between thinking about something twice (perhaps even using the same
representation) vs. implicitly recognizing the content as the same has been highlighted by
a number of theorists: see, e.g., Fine 2007; Kaplan 1990; Millikan 1994, 2000, chap. 9; Perry
1980.
3. The notion of apparent de jure sameness of topic was first introduced in Schroeter 2007,
2008. For further discussion of its epistemic role and its importance to concept individua-
tion, see Schroeter 2012. For a related notion of de jure coreference, see Pinillos 2011. Like
many others, Ángel Pinillos assumes that the appearances associated with de jure corefer-
ence are infallible, an assumption we reject (see Schroeter 2007).
Gottlob Frege (1892) took the appearance of de jure sameness to be crucial

to individuating the meanings of coreferential expressions. Like many others,
we believe it is also a reflection of the underlying representational structure of
thought.
Ordinary human learning and reasoning about particular topics would be
impossible without some basic ways of organizing our cognitive states by topic.
Your ability to learn about caffe lattes, for example, to amass a body of atti-
tudes and dispositions pertaining to that topic, to form new beliefs about it, to
retrieve relevant information via memory, and to stay on topic in reasoning—
all these basic cognitive abilities depend on stable organizational structures. In
particular they require something like a mental filing system: a system for bind-
ing and segregating attitudes by topic in such a way that you are immediately
disposed to take bound attitudes as obviously and incontrovertibly pertaining
to the same topic.4
The metaphor of a mental file folder helps characterize the functional orga-
nization of the mechanisms, whatever they are, that underwrite the relation
of apparent de jure sameness. In your office filing cabinet, file folders serve to
group together a changing body of documents as pertaining to a single topic
and to distinguish that group from other bundles of documents. When you pick
up your “Mike Ridge” file, for instance, you are immediately disposed to treat
all the bound documents as pertaining to the same topic, Ridge. Notice that this
disposition is based solely on the filing system, not on your independent scru-
tiny of the items bound together. Indeed it would defeat the whole purpose of a
filing system if you had to know the contents of the documents before deciding
whether to treat them as pertaining to the same topic.
The central idea behind the mental file metaphor is that conceptually artic-
ulated attitudes are organized in a similar way. There are stable mechanisms
that bind together and segregate your changing bodies of attitudes in such a
way that you are immediately disposed to treat bound attitudes as pertaining
de jure to the same topic. Your various “Mike Ridge” beliefs, for instance, will
strike you as obviously and incontrovertibly pertaining to a single individual.
And that body of beliefs is segregated from your “Mike Tyson” beliefs, which
do not seem obviously and incontrovertibly to pertain to the same person as
4. Early proponents of the view that cognition of individuals involves mental files or “dos-
siers” of information about a topic include Evans 1973; Grice 1969; Strawson 1974, chap.
2. The file metaphor has been developed in various ways by subsequent theorists; see, e.g.,
Bach 1987; Dickie 2010; Forbes 1989; Jeshion 2002; Lawlor 2001; Perry 2001; Recanati 1993,
2012; Schroeter 2007, 2008. Related notions have been developed in linguistics to deal with
discourse anaphora (Heim 1983, 1988; Karttunen 1974) and in cognitive psychology to
explain binding and recalling information about individuals (Anderson 1977; Kahneman
and Treisman 1984).
Why Go Hybrid? 251
your “Ridge” beliefs. Moreover, you do not engage in any comparison of your
various “Ridge” thoughts prior to treating them as pertaining to the same indi-
vidual: your memories and beliefs simply present themselves directly to your
conscious attention as pertaining de jure to the same person. The mental file
metaphor nicely captures the immediacy of this apparent sameness relation: it
is not based on a prior assessment and matching of particular attitudes.
We have suggested elsewhere that a mental filing system is required for con-
ceptually articulated thought.5 To deploy a particular concept, on this view,
a thought must be linked to other thoughts in such a way as to generate the
appearance of de jure sameness (either occurrently or dispositionally). To
count as deploying a concept of caffe latte, for instance, it is not enough for
your current representational state to be linked to the world in certain char-
acteristic ways—say, by being causally provoked by lattes in your vicinity and
causing latte-oriented actions. Having a particular concept requires a capacity
for keeping track of the relevant topic in thought: to count as having a concept
of caffe latte, your thoughts must involve some stable cognitive structure
that allows you to entertain a wide variety of thoughts that you take to pertain
de jure to the same topic.
It should be obvious that normative beliefs have this kind of conceptual
structure: to judge that an action is morally wrong, you must have a capacity
to keep track of the topic of moral wrongness. Any account of normative belief
that does not explain this capacity to keep track of de jure sameness of norma-
tive topic is a nonstarter. Jesse Prinz (2007, 94), for instance, has recently sug-
gested that “to believe something is wrong in a non-deferential way is to have
a sentiment of disapprobation toward it.” Read literally, this seems to be a cat-
egory mistake. A sentiment does not bind your various moral attitudes together
in such a way that they all strike you as pertaining de jure to the same topic.6
Any theorist who proposes to reduce normative judgments to mere sentiments
or attitudes, like approval, endorsement, planning, or what have you, must
explain how the attitude in question can underwrite our capacity to keep track
of sameness of normative topic among our thoughts. Expressivists often seem
to endorse such reductions. According to Allan Gibbard, for instance, to think
something rational is to accept norms that permit it (Gibbard 1990, 46), and
5. Schroeter 2013; Schroeter and Schroeter 2014.

6. Our argument here has nothing to do with the question of whether emotions “track” or
“represent” features of the world. According to Prinz, emotions like fear represent anthropo-
centric properties like danger. Although we are skeptical about this mind/world representa-
tional claim, our focus in this section is exclusively on a mind/mind relation: the appearance
of de jure sameness of topic that links propositional attitudes.
to think an action is the thing to do is to decide to do it (Gibbard 2003, 7–8).

Prima facie at least, this reduction seems to run into the same type of difficulty
as Prinz’s: different decisions or states of norm-acceptance are not bound in
such a way that they strike the subject as pertaining de jure to the same topic.7
Cognitivists in the metaethical literature have tended to highlight the repre-
sentational (mind-to-world) function of concepts: concepts constitute ways of
picking out objects, kinds, or properties in thought. But your concept of moral
wrongness does not merely serve to demarcate the property of being wrong, it
also serves to cross-index your various moral wrongness attitudes as pertain-
ing de jure to the same topic. This mind-to-mind aspect of concepts—the way
they mark apparent sameness of topic among mental states—has been insuffi-
ciently emphasized in the metaethical literature. But it is crucial to understand-
ing conceptually articulated thought.
One way of appreciating the importance of this mind-to-mind aspect of con-
cepts is to focus on logical relations, like validity, consistency, and contradic-
tion. Logical relations depend on more than reference alone. An inference of
the following form, for instance, is not logically valid:
1. Phosphorus rises in the morning.

2. So Hesperus rises in the morning.
Of course the truth of the premise metaphysically guarantees the truth of

the conclusion—since the two claims in fact attribute the very same prop-
erty to the very same object. But logical validity requires more than mere de
facto sameness of topic: it requires de jure sameness of topic (and the subjec-
tive appearance of de jure sameness). This inference is not valid, because it
seems possible, from the perspective of the thinker, that Hesperus may not
be Phosphorus.
Focusing on mind-to-mind relations of apparent de jure sameness, we think,
supports an alternative to the standard, broadly Fregean account of concepts. In
the next section we present this alternative account.
7. This objection is directed specifically against Gibbard’s core formulation of his proposal.
It is not intended to vitiate all versions of Gibbardian expressivism much less to extend to all
expressivist theories. Forming two plans to perform distinct actions does not eo ipso dispose
you to treat those plans as obviously and incontrovertibly pertaining to a common topic.
Accounting for the conceptual structure of normative judgments will require additional
resources than the ones Gibbard’s equation of normative judgment with plans suggests. The
binding model of concepts we propose below would be one type of additional resource the
expressivist could appeal to in order to account for conceptual structure. (On the compat-
ibility of our binding model of normative concepts and expressivism, see Schroeter and
Schroeter 2014)
Why Go Hybrid? 253
MAT C H I N G V E R SUS B I NDI NG
There are two aspects to an account of concepts. First, we need a theory of

concept identity. What makes it the case that two of your token attitudes—say a
belief and a desire—deploy the very same concept? And what makes it the case
that two different speakers deploy the very same concept in thought? Second,
we need an account of how the reference—or more generally the semantic
value—of a thought deploying a particular concept is determined. What makes
it the case that your latte thoughts refer to lattes, whereas your moral right-
ness thoughts are about moral rightness?8
According to the traditional “matching” account of concepts based on the
Fregean notion of sense (Frege 1892), a single aspect of understanding plays
both explanatory roles. First, concept identity: two token thoughts deploy the
very same concept just in case each token is associated with the very same
reference-fixing criterion—a set of cognitive dispositions that put the subject
in a position to conclusively identify the reference. Second, reference determi-
nation: the reference of a token thought is determined by the reference-fixing
criterion associated with that token (together with context). On this tradi-
tional account, concepts are finer-grained than reference, since different cri-
teria can pick out the same object, kind, or property. And this is the desired
result: although coreferential, your “Hesperus” and “Phosphorus” thoughts
deploy different concepts. Moreover, the account promises to explain why dif-
ferent thoughts deploying the same concept strike the subject as obviously and
incontrovertibly coreferential. Ultimately, you cannot imagine a case where
Hesperus is not Hesperus, on this account, because there is a precise match in
the reference-fixing criteria associated with the two tokens.9
We favor a “binding” account of concepts that disaggregates these two
explanatory tasks—concept identity and reference determination. First, what
makes it the case that two token thoughts deploy the same concept, we suggest,
8. Until now we have been neutral about the nature of the semantic value of particular con-
cepts. The topic picked out by a concept might be representational (e.g., an object, kind,
property, or truth function) or it might be purely expressive (e.g., a specific attitude, moti-
vation, or inferential disposition). But it will simplify matters to couch our discussion of
concepts in purely representational terms, to speak of concepts’ reference and coreference
relations. This is the traditional approach to concepts, and nothing in our account of concept
identity or of the determination of semantic values hangs on the precise nature of the seman-
tic values assigned. We will assume that the reference of a predicate is a property (rather than
a function or an extension), since this coheres with a commonsense understanding of the
representational purport of predicates. But nothing we say here depends on this choice.
9. We believe the explanation of the appearance of de jure sameness proposed by the match-
ing model is inadequate, but we will not press this point here. For details, see Schroeter 2012.
is a relational fact about them: the fact that they are bound together in a partic-
ular way. This binding relation does not depend on any match in the properties
of the bound tokens considered independently. Second, the reference of a token
thought is determined not by facts about that token considered in isolation but
rather by facts about the whole group of thoughts to which that token is bound.
Thus on the binding account, there is no need for each token deployment of an
element of thought to be associated with the very same reference-fixing crite-
rion to guarantee coreference.10
The core relation essential to concept identity, we suggest, is a lot like
anaphora in public language. Consider the following exchange:
a: “Julia Gillard was an efficient leader.”
b: “Sorry, but who is she?”
In this exchange, it seems obvious and incontrovertible to both parties that
there is just one topic in question—Gillard. But clearly the two have very dif-
ferent substantive understanding of the topic in question. Presumably A has a
rich understanding of the characteristics of the Australian politician picked out
by the name “Gillard,” whereas B may know nothing more than that Gillard
is the person (if any) A has just referred to. It seems clear that in this case the
appearance of de jure sameness is not based on a prior match in substantive
understanding: what is crucial is simply that A’s and B’s utterances (as well as
the associated thoughts) are linked together in an appropriate way. The prere-
flective linguistic mechanisms of discourse anaphora establish direct cognitive
binding between the two token attitudes so that both parties will take those
thoughts to pertain de jure to the same topic.
Being bound by apparent de jure sameness relations is, we suggest, the basic
ingredient for concept identity. In the previous section we posited a mental
10. To get an intuitive feel for the binding account of concept identity, consider an analogy.
Theories of personal identity seek to explain what makes it the case that different person
stages are stages of the same person. Why does that five-year-old child in 1940 count as
the same person as this seventy-year-old woman in 2005? Traditional accounts of personal
identity seek to specify some fact about each person stage that is necessary and sufficient for
being the same person: e.g., each stage is associated with the same immaterial soul, or each
has the very same personality traits. Relational accounts of personal identity, in contrast, seek
to explain personal identity in terms of a distinctive kind of connection among the person
stages: e.g., the two stages are linked by a distinctive kind of causal relation, or they are linked
by chains of apparent memories. Derek Parfit (1971) was an influential proponent of the
relational approach to explaining what is crucial to personal identity. In a similar spirit we
are suggesting that a direct binding relation is crucial to concept identity. On the basic idea
that Fregean cognitive significance phenomena can be explained on the basis of relationally
individuated linguistic expressions types, see Fiengo and May 1994, 2006; Hawthorne and
Lepore 2011; Kaplan 1990.
Why Go Hybrid? 255
filing system that binds and segregates different elements of thought in such a
way that they seem to pertain de jure to the same topic. Your different thoughts
about lattes share the same concept, we suggest, largely in virtue of the fact that
they are bound together by these basic cognitive mechanisms: they are bound
by a single mental file.11 As with discourse anaphora, being bound together in
the relevant way does not depend on any specific match in the independent
properties of the token states so bound.
The advantage of this binding approach is that it does not require a precise
match in topic-fixing criteria. As many theorists have pointed out, isolating a
determinate pattern of understanding that is necessary and sufficient for con-
ceptual competence has proven elusive—not just for names and natural kind
concepts but even for relatively superficial concepts like latte. For virtually
any particular assumption, disposition, or motivation you currently associate
with the topic, it seems you could give up that aspect of understanding without
thereby compromising the appearance of de jure sameness. Moreover, there
are good theoretical reasons why an account of concept identity should allow
for this sort of flexibility in substantive understanding. If a concept can remain
stable through open-ended inquiry and debate, its identity cannot depend on
a specific pattern of understanding. It is a virtue of a direct binding account of
concept identity that it does not require a precise match in substantive under-
standing for sameness of concept. Like a relational account of personal identity,
a relational account of concept identity allows for much more variation in asso-
ciated understanding than any traditional matching account.
This is not to say that anything goes. A further condition on concept identity,
in addition to the apparent de jure sameness relation, is that the bound thoughts
really do represent the very same topic. And this means that the substantive
understanding cannot diverge so radically as to preclude univocal semantic
interpretation. To get an intuitive feel for this condition, let us reconsider the
discourse anaphora case.
a: “Julia Gillard was an efficient leader.”
b: “Sorry, who is she?”
c: “I went to school with her—she was fair dinkum then.”
d: “I know her too—she’s a famous cardiologist at the Mayo clinic.”
11. Particular mental files are interconnected networks of attitudes bound together by sub-
personal cognitive mechanisms in such a way as to seem de jure coreferential. Individual
mental files are simply networks of bound states. It is worth emphasizing a misleading aspect
of the file folder metaphor: whereas manila file folders exist independently of the items they
bind together, the same is not true of mental files. There is nothing more to file identity than
the particular network of attitudes bound together in the relevant way.
All four utterances are directly linked by linguistic mechanisms that gen-
erate the appearance of de jure sameness of topic. This automatic binding
relation helps explain how the first three speakers nonaccidentally corefer
despite significant variation in their substantive understanding of the topic
in question. But the fourth speaker is different: although her thought stands
in the same anaphoric relations, the substantive understanding and history
directly associated with her token thought precludes coreference with the
other three. Thus the binding relations established by discourse anaphora can
be defeated: subjects’ substantive understanding may diverge so radically from
others’ as to preclude a univocal interpretation despite the initial appearance
of de jure sameness.
Similarly, a theory of concept identity is beholden to two distinct factors: one
factor captures the subjective perspective on de jure sameness of topic, and the
other captures the objective facts about sameness of semantic value. To a first
approximation, we suggest that two token elements of thought involve the same
concept just in case they satisfy two conditions:
1. Connectedness: The token elements of thought are stably bound

together by basic cognitive mechanisms that present them as
pertaining de jure to the same topic.
2. Congruence: The substantive understanding directly associated with
the tokens does not diverge so radically as to preclude a univocal
interpretation.
In the intrapersonal case, the relevant epistemic connections are generated

by the subject’s mental filing system: particular mental files bind a diverse bun-
dle of attitudes and dispositions together in such a way that the subject treats
them as obviously and incontrovertibly pertaining to the same topic. Mental
files of course cannot be shared by different individuals. But as the anaphora
example brings out, the appearance of de jure sameness among speakers can be
generated through public language. In the interpersonal case, we suggest, the
relevant connections among elements of thought are established through cog-
nitive mechanisms stably linking mental files to basic lexical expressions like
names, adjectives, and predicates.12 Such interpersonal binding mechanisms
are less reliable than intrapersonal mechanisms—there is more chance that you
and I fail to corefer when we take our use of the name “Aristotle” to pertain de
jure to the same individual than when I take my various Aristotle thoughts
to pertain de jure to the same individual. The reason for this, we suggest, is that
12. For a helpful articulation of this position, see Cumming 2013a, 2013b.
Why Go Hybrid? 257
the chance that congruence will fail is greater between individuals than within one
individual.13
This relational account of concept identity opens the way for a new, more plau-
sible approach to reference determination. On the traditional Fregean approach,
token elements of thought must earn their semantic values on their own, indepen-
dently of how they are linked to other tokens. So the basic unit of interpretation
is an individual token considered in isolation from other tokens. The semantic
value is determined by the pattern of understanding associated with that token—
its reference-fixing criterion.14 On this Fregean model reference-determination is
explanatorily prior to apparent de jure sameness relations: two tokens appear de
jure coreferential because they are independently associated with the very same
reference-fixing criterion. Let us call this a “token-based” approach.
The binding model reverses this order of explanation. On this model the
basic units of interpretation are not token elements of thought considered in
isolation but rather the entire bundle of thought elements that the subject sees
as pertaining de jure to the same topic. The norms for semantic interpretation
take the apparent de jure sameness at face value: they look for a univocal ref-
erential assignment for the bound states considered as a corporate body. So on
the binding model, apparent de jure sameness is explanatorily prior to reference
determination: a token element of thought has a particular reference in virtue
of how it is bound together with other tokens by de jure sameness relations to
form a default unit of interpretation. Call this a “tradition based” approach.15
13. But congruence can also fail in the intrapersonal case, leading to unreliable appearances
of de jure sameness of topic. See Schroeter 2007 for discussion.
14. Neo-Fregeans allow that the criteria associated with a token may fix the reference indi-
rectly, as a function of empirical facts about the actual context (Chalmers 2002; Evans 1982;
Jackson 1998; Lewis 1981; McDowell 1984). In that case de jure coreference requires both a
match in the criteria and a match in the relevant features of the external context.
15. In taking facts about the representational tradition as a whole to affect the reference
of an individual’s token uses of concepts, our binding model resembles “social deference”
accounts of conceptual competence (for the locus classicus, see Putnam 1972). However,
unlike such accounts, the binding model does not require individuals to single out some
class of “experts” whose current understanding fixes the reference of your term. A simple
deference model makes two implausible assumptions: (i) that you have some determinate
criterion for identifying who counts as an expert; (ii) that you are disposed to treat the opin-
ions of the designated experts as the ultimate court of appeal in settling which things fall
into the extension of your concepts. Neither assumption seems to fit with normal epistemic
standards. We believe that the binding model fits much more neatly with our best epistemic
standards: it is natural to take the representational traditions to which you belong as rel-
evant input into interpretation of your own concepts precisely because you take yourself to
de jure corefer with them. But you need not have any fail-safe criteria for identifying just
who belongs to the relevant representational tradition except by tracing out the networks of
This “tradition-based” approach fits well with our epistemic practices.

Consider your token desire to emulate Aristotle. What aspects of your substan-
tive understanding of this token are relevant to determining who, if anyone,
this desire is about? Presumably your understanding of Aristotle includes a
wide range of standing attitudes: for example, your belief that Aristotle was the
last great philosopher of antiquity, that he wrote the Nicomachean Ethics, that
he was the student of Plato, that you learned about him in your first philoso-
phy class, that he was named “Aristotle,” and so on. All of these attitudes are
bound by your mental filing system and thus presented as pertaining de jure
to the same topic. According to the tradition-based approach, the entire body
of bound attitudes forms a basic unit of interpretation: in deciding who exactly
counts as Aristotle, the interpreter is seeking to find the referent of all of these
attitudes simultaneously—and any of the bound attitudes could potentially be
relevant to answering that question.
Now consider what happens when your understanding of Aristotle changes
over time—perhaps quite radically. On the token-based approach, aspects of
understanding that are relevant for interpretation must be replicated in all
deployments of your Aristotle concept. So any change in your understanding
of Aristotle will lead to a change of concept unless that aspect of understanding
is entirely irrelevant to interpretation. The tradition-based approach does not
have this consequence. Since your past and present thoughts are bound by rela-
tions of apparent de jure sameness, those past thoughts are still relevant to the
interpretation of your current beliefs—even if you currently reject them. Just as
in the case of an anaphoric connection, the mere fact that the thoughts stand in
this apparent de jure coreference relation makes it plausible to treat them as a
single unit for purposes of semantic interpretation.
The same point applies at the interpersonal linguistic level. When you hear
other people use the name “Aristotle” in the right context, you will automati-
cally hear their claims as pertaining de jure to the same topic that you associate
with that name. As in the case of anaphora, your capacity to track lexical struc-
ture gives rise to the appearance of de jure sameness of topic linking different
individuals’ thoughts. In both cases, when interpreting the semantic content of
your current token thought, it makes sense to take into account other people’s
thoughts to which it is bound by apparent de jure sameness. In our little dialog
about Gillard, for instance, B’s thought cannot be assigned a semantic value
without taking into account A’s thought. Similarly, when we seek to interpret
your desire to emulate Aristotle, we should take into account the chains of
apparent de jure sameness that link your “Aristotle” thoughts to those of others
apparent de jure coreference, and there need not be anyone in particular whom you treat as
the final court of appeal in settling the reference of your term.
Why Go Hybrid? 259
in your linguistic community. The default unit of interpretation is not your

token element of thought considered in isolation but a complex network of
appropriately connected thoughts bound together via individuals’ mental files
and public language so as to form a single representational tradition.16
To sum up, the binding model offers a natural account of concept identity and
of reference determination that nicely reflects commonsense semantic and epis-
temic commitments. First, the binding model can easily explain how concepts
could remain stable through open-ended inquiry and substantial variations in
understanding. This flexibility in understanding has proven difficult to fully vin-
dicate on a matching model of concepts, since such accounts take concept identity
to depend on the reproduction of particular canonical patterns of understand-
ing. On the binding model, in contrast, sameness of concept is largely a matter of
being linked up in the appropriate way. So there is much more scope for variation
in substantive understanding associated with token occurrences of the concept.
Second, it is natural to take the representational traditions in which we participate
to be relevant to determining the reference of an individual’s thoughts. The bind-
ing account can smoothly accommodate this assumption: since the default units
of interpretation are defined by relations of apparent de jure sameness, the under-
standing relevant to interpretation can be socially and temporally distributed—it
need not be directly associated with each token occurrence of the concept.
T H E I N T E R N AL LI N K TO M O TI VATI O N
Let us return to the objection we started with, that pure cognitivism about nor-
mative judgments cannot explain the distinctive action-guiding role of these
judgments. A binding model of concepts, we will argue, is better equipped than
a traditional matching model to respond to this challenge.
An important advantage of a binding model of concepts, we have suggested,
is its flexibility in accommodating variability in competent subjects’ substantive
understanding. Variability is a feature of all concepts, but it seems particularly
acute in the case of normative concepts. We have argued elsewhere that the
binding model provides new resources for vindicating traditional versions of
cognitivism, according to which all conceptually competent subjects pick out
the very same property despite variations in their substantive understanding
(Schroeter and Schroeter 2009b, 2014). Here we will argue that this flexibility
also helps deflate the worry that cognitivism cannot explain the action-guiding
role of normative judgments.
16. For further motivation and defense of our tradition-based approach to the determination
of semantic values, see Schroeter 2012 and Schroeter and Schroeter 2014.
If the function of normative judgments is simply to attribute a property, why

is the link between normative judgment and motivation not purely external and
contingent? And why does not our justification for assuming normative judg-
ment and motivation normally go hand in hand just based on brute empirical
observation? Just as the connection between being American and liking peanut
butter is justified by statistical generalization, so the connection between judg-
ing an action right and being motivated to act accordingly would be justified by
statistical generalization.
The most straightforward way to vindicate an internal link between norma-
tive judgment and motivation would be to impose a strict motivational inter-
nalist (SMI) constraint as a precondition for competence with the standard
normative concepts we share:
(SMI) A subject does not count as competent with our shared normative
concepts unless she or he is appropriately motivated to act in accordance
with her or his normative judgments.
Such a constraint on competence is consistent with a pure cognitivist seman-

tics: the semantic role of normative concepts may simply be to attribute a spe-
cific property to an action. However, SMI is implausible, we think, for much the
same reasons that strict constraints on descriptive understanding are implau-
sible. According to common sense, someone can continue to think about nor-
mative questions even if, because of cynicism or depression, she or he loses any
motivational disposition to act accordingly. But the ultimate reason why SMI
should be rejected is not simply that it clashes with commonsense intuitions
about cases. The deeper and more important reason to reject SMI is a theoreti-
cal one that is grounded in the role concepts play in grounding stability of topic
in thought and talk. Strict preconditions on competence with normative con-
cepts—whether motivational or cognitive—are implausible precisely because
they arbitrarily restrict open-ended inquiry and debate. SMI would place arbi-
trary and epistemically vicious restrictions on the potential for direct rational
engagement through variation in substantive motivational states (Schroeter
and Schroeter 2014).17
However, our aim here is not to argue against strict motivation internalism.
Our point is that on the binding model, the fact that motivation is not strictly
17. A central role of concepts, we have noted, is to keep track of a topic despite variations
in understanding. If someone does not share our concept of morally right, for instance,
then she or he is not coordinating on the same topic with us when she or he uses the term
“morally right.” The effect of imposing strict preconditions for conceptual competence is that
anyone who fails to conform to those preconditions is eo ipso excluded from rational debate
about a given topic with the rest of us. But this is a significant threat to virtuous epistemic
Why Go Hybrid? 261
necessary for conceptual competence does not entail that it is semantically irrel-
evant. On the contrary, the binding model allows motivation to play a signifi-
cant semantic role at the communal level in fixing reference and determining
competence. In particular we will argue for the following three theses.
1. Reference fixing: The motivational role played by normative concepts

within the shared representational tradition helps determine which
properties are picked out as the reference.
2. Competence: Being motivated to act in accordance with one’s
normative judgments is a constitutive element of normal competence
with normative concepts.
3. Prediction: This aspect of normal competence justifies a default
assumption connecting normative judgment with motivation.
If we are right, then the binding model can vindicate an internal connection
between normative judgment and motivation—while allowing that the motiva-
tion itself or belief in the general motivational role are not strictly necessary for
competence with normative concepts.
Reference Fixing
Since the basic units for interpretation on the binding model are shared represen-
tational traditions, the reference must be fixed by facts about that tradition as a
whole. Elsewhere we have defended a general account of reference determination
for the binding model of concepts (Schroeter and Schroeter 2009b). Our sugges-
tion is that reference is determined by holistic rationalizing interpretation of the
entire representational tradition associated with a given term. Roughly the idea
is that the semantic value of a token use of an expression is determined by the
upshot of broad reflective equilibrium when fully informed about the diachron-
ically and socially extended representational tradition to which it is linked. The
correct semantic interpretation is the one that captures what is most important
practices. The cost of ruling out knaves and moral skeptics as competent with our concept
morally right, for instance, is that we forfeit the possibility of learning from them about
the real nature of the imperfectly understood topic we have been talking about under the
heading “morally right.” And we also forfeit the possibility of their learning from us. Direct
epistemic engagement and rational persuasion is ruled out by fiat. For a related point about
why restrictive competence conditions are inappropriate for linguistic meanings, see Finlay
2005, 18. For the view that normative concepts in particular have the function of facilitating
direct epistemic engagement, see Gibbard 2003, 260. A central role of normative concepts is
to allow us to “put our heads together” in thinking about how to live.
to justifying and sustaining a representational tradition. The key claim for pres-
ent purposes is this: the action-guiding role prototypically played by normative
concepts like morally right is crucial to sustaining and justifying our continued
interest in our representational tradition. We are interested in morally right actions
in part because people commonly cite judgments about moral matters in deciding
what to do and they justify their opinions about which actions are moral with an
eye to that action-guiding role. Given its centrality to our representational prac-
tice, this action-guiding role will have a significant impact on the correct interpre-
tation of moral and other normative concepts. In particular the action-guiding
role will affect precisely which property is assigned as the reference.
To get a sense of how an assessment of which features are most important to
a representational tradition can shape our ultimate verdicts about the reference
of the corresponding concepts, let us consider a different kind of case. Theorists
disagree about how to interpret racial concepts like black or white. Do they
pick out any determinate categories at all? If so, are those categories demarcated
by biological characteristics like lineages or by social characteristics like being
a member of a certain historically oppressed group?18 This debate hinges, we
suggest, on one’s assessment of the primary interests that justify and sustain the
corresponding representational traditions. Those who favor a biological inter-
pretation—or who argue for eliminativism on biological grounds—privilege
the theoretical analogies between racial categorizations and other biological
claims. Those who favor a social construction model, in contrast, point to the
role that racial categorizations play in a systematic oppression of racially desig-
nated individuals. This oppression, they argue, is the real reason racial catego-
ries have been of interest to us, and it explains why our judgments about who
counts as a member of a given race are shaped by social rather than biological
criteria. Our aim here is not to settle this debate but only to highlight the fact
that determining the primary point or interest of a representational tradition
plays a crucial role in grounding specific verdicts about reference.
In the case of normative concepts, determining the point of a representa-
tional tradition also plays a key role in grounding verdicts about the precise
nature of the topic picked out. Debunking accounts of morality, for instance,
privilege the role of moral judgments in imposing the interests of one group
on others. The concept of morally right, a debunker might argue, picks out a
class of actions demarcated by social conventions that promote the interests of
the dominant social elite through false consciousness. In that case there is a case
to be made that morally right picks out the property of serving the interests
18. For instance, K. Anthony Appiah (1996) and Naomi Zack (2002) argue for eliminativism
about standard racial concepts; Philip Kitcher (2007) supports a refined biological view; and
Sally Haslanger (2000) and Ron Mallon (2006) argue in favor of social constructivist accounts.
Why Go Hybrid? 263
of the elite. One way to resist this type of simple debunking interpretation is
to argue that the role of moral categorization in mutually acceptable justifica-
tion is in fact more important to justifying and sustaining our representational
tradition with moral concepts than social oppression. As a consequence the
property of moral rightness must reflect ideal standards of interpersonal justifi-
ability rather than conventions of social domination.
Similarly, the central action-guiding role of normative categorization affects
which property (if any) can be assigned as reference. Contrast normative
concepts like right with evaluative concepts like good or admirable. It is
uncontroversial that categorizing actions as right normally plays a direct role
in monitoring, assessing, and guiding agents’ choices—this action-guiding
role is central to sustaining our communal tradition in categorizing actions as
right. In contrast, categorizing something as good does not play the same sys-
tematic role as providing a direct guide to action. This difference in the nor-
mal action-guiding role of normative and evaluative concepts has an impact
on their semantic interpretation. In the case of good, it makes sense to attri-
bute a gradable property as the reference: goodness is often a matter of degree,
and there is no theoretical pressure for a sharp cutoff point. But a gradable
property is not an adequate interpretation of the concept of right: if rightness
were a matter of degree, it could not play the characteristic role of providing
a standard for choosing one action over other available options. The central
action-guiding role of normative concepts thus helps ground the attribution of
a categorical property as the reference rather than a gradable property. Similarly,
the action-guiding role of normative concepts will tend to favor interpretations
that are at least partly in line with human capacities. In the case of an evalua-
tive concept like good, it may make sense to attribute a property that wholly
outruns human capacities to bring about good states of affairs. But it is not clear
that the same could be said of normative concepts. Given the importance of
guiding and assessing choice to our practice with normative concepts, the prop-
erty picked out by concepts like right or ought should not be totally detached
from human capacities. Arguably, there is some sense in which “ought” implies
“can”: right actions must be ones that it is possible for the agent to perform.19
If we are right that this sort of rationalizing interpretation is what fixes refer-
ence, then the correct referential assignment for our shared representational
19. What if members of our community suddenly lose any disposition to act in accordance
with their judgments about which actions count as “morally right”? Will the reference of our
concepts shift? If we are right that the action-guiding role associated with “morally right” in
the linguistic community at large helps determine the reference of the associated concept,
then it is plausible that reference may eventually shift after these action-guiding disposi-
tions disappear. But this shift will not happen overnight: our interest in coordinating with
past usage will preserve the original reference for a certain period until the new patterns
traditions will be affected by the central action-guiding role normative concepts

play within our linguistic community. Taking action guidingness into account
in reference determination is compatible with a pure cognitivist semantics for
normative concepts, since no motivational profile is assigned as part of the
semantic value: the semantic value is just an ordinary property (together with
conceptual structure). Our point is that insofar as action guidingness helps fix
reference, it is not semantically irrelevant. One cannot explain the semantic
properties of our representational traditions with normative concepts without
appealing to the central action-guiding role they play within our community
as a whole. It would be a mistake, then, to suppose that the relation between
normative concepts and motivation is purely external and contingent.
Competence
On a binding model what is crucial to conceptual competence is the ability to

keep track of sameness of topic through direct relations of apparent de jure
sameness. On this approach the only strictly necessary condition for individu-
als to count as competent with a shared concept is their being appropriately
causally connected via de jure sameness relations to the shared representational
tradition. As we noted earlier, SMI is not required for competence with norma-
tive concepts. Despite deviant motivational dispositions, knaves and sociopaths
may count as competent with our concept morally right (and thus capable of
making genuine moral judgments) in virtue of being appropriately linked up to
our shared representational tradition.
It does not follow, however, that motivation is simply irrelevant to compe-
tence with such normative concepts. On our binding model motivation can
affect competence in two distinct ways, via the connectedness or the congru-
ence constraints on competence.
of use begin to dominate. Eventually, however, our continued use of the term may come
to be primarily justified and sustained by some new set of practices that ground a distinct
referential assignment. If such new practices become dominant, the reference of the repre-
sentational tradition will shift—and the shift in reference grounds a change in concept. The
case here is analogous to Gareth Evans’s (1973) Madagascar case and Tyler Burge’s (1988)
“slow-switching” case, in which subjects unwittingly start using a term in new ways that shift
the reference over time. For more details about how concepts are individuated in cases like
these on our binding model, see Schroeter 2008. It is worth emphasizing, however, that a
shift in reference is not a forgone conclusion if the community’s general motivational dispo-
sitions change. For instance, the dominant communal interests associated with the practice
may be to ensure a stable reference to the property originally picked out by earlier genera-
tions of morally motivated individuals. Perhaps something like this actually happens in the
case of outmoded value terms, like “chaste.”
Why Go Hybrid? 265
First, consider connectedness. A number of metaethicists have suggested

that knaves’ and sociopaths’ competence with moral concepts is “parasitic” on
that of normal users of those concepts: morally unmotivated individuals count
as competent with normative concepts only insofar as others in their commu-
nity (or the subject’s past selves) are disposed to act in accordance with their
moral judgments.20 Clearly, such a dependence relation requires some theo-
retical explanation: why are the mental states of this group relevant to char-
acterizing the intrinsic nature of mental states entertained by that individual?
Simply being abnormal with respect to an arbitrary population does not give
us any reason to posit a dependence relation. Your liking peanut butter and not
vegemite is normal with respect to the class of Americans and abnormal with
respect to the class of Australians. Yet these normality or abnormality facts are
obviously irrelevant to characterizing the intrinsic nature of your preference.
Given that the particular mental states that we are focusing on here—nor-
mative judgments—are conceptually articulated, the explanation of the alleged
dependence relation must be grounded in an account of the relevant concepts.
Our binding model of concepts provides a clear theoretical explanation of this
dependence. First, concepts are relationally individuated on this model: indi-
viduals count as competent with our normative concepts only insofar as they
are appropriately connected to our shared representational traditions. Second,
these representational traditions constitute the default unit of semantic inter-
pretation. Third, the prototypical action-guiding role of normative concepts
within the community as a whole, we have suggested, is crucial to fixing their
reference. So there is a sense in which this action-guiding role within the tra-
dition as a whole is indirectly constitutive of anyone’s competence with our
concept morally right: it is only because knaves and sociopaths are linked to
our tradition in which prototypical users are appropriately motivated that they
possess the particular concepts they do.21
Second, consider the congruence constraint. On the binding model, the
congruence constraint requires that the subject’s practice with a concept
20. Bedke 2009, 191–195; Blackburn 1998, 59–68; Dreier 1990, 9–14; Tresan 2006, 149–152;
van Roojen 2010.
21. Notice that the claim here is not merely that all competent individuals are connected to
a past tradition in which this motivational role was widespread. A similar statistical connec-
tion holds between our concept marriage and its past application exclusively to opposite-sex
partners. But this widespread historical categorizing practice does not suffice to show exclu-
sive application to opposite-sex partners is constitutive to our concept marriage: arguably
this pattern could change without disrupting referential and conceptual stability. The crucial
test for a constitutive relation is not statistical but whether the pattern plays a role in fixing
the reference.
must not diverge so radically from the rest of the group so as to preclude a
common interpretation. In effect congruence functions like a rough “fam-
ily resemblance” condition on competence: although no particular aspect of
conceptual practice is strictly required, there must nonetheless be some broad
overlap between the individual’s practice and the “center of gravity” within the
group as a whole.
In the case of normative concepts, one of the most central and obvious facts
about the communal practice is that judging an action as right normally plays an
action-guiding role for the judger. The connection between classifying an action
as right for the agent to do and intending to perform that action is something we
all learn at our parents’ knees with more or less vivid force. And given normal
human psychology, most people eventually internalize this link within their moti-
vational systems. Conforming to this characteristic action-guiding role goes a long
way toward ensuring that an individual’s understanding of a normative term like
“right” is congruent with that of the group as a whole.
Thus having appropriate motivational dispositions can play a genuinely consti-
tutive role in securing competence with normative concepts. Such motivational
dispositions, we suggest, are near the top of a list of “family resemblance” traits
that help satisfy the congruence constraint on competence with shared normative
concepts. So a motivation to act in accordance with one’s normative judgments
can help make it the case that one counts as conceptually competent.
The prototypical action-guiding dispositions, we believe, are neither necessary
nor sufficient for congruence with shared normative concepts. Against neces-
sity: If the knave’s substantive understanding overlaps in enough other ways with
the “center of gravity” of the group—for example, if he has detailed theoretical
beliefs that are congruent in important ways with those of others in the group—he
can successfully participate in the shared representational tradition. Against suf-
ficiency: If a subject’s sole criterion for counting an action as “right” was that it
lengthened the agent’s shadow, we would not treat her or him as de jure coreferring
with the rest of us—even if the subject was motivated to act in accordance with
those judgments. To count as competent with our shared normative concepts, the
subject’s beliefs about which actions are right and why they are so must be intel-
ligible to others in the group.22 The general lesson is that a binding model allows
for many different ways of satisfying the congruence constraint. Nevertheless,
being appropriately motivated is partly constitutive of normal conceptual compe-
tence: by conforming to the action-guiding role individuals can partially fulfill the
22. This point is widely accepted in the case of moral concepts, but it is controversial in the
case of thin normative concepts like “all-told right” or “is a reason.” We will not argue for it
here. For an elaboration, see Schroeter and Schroeter 2009a.
Why Go Hybrid? 267
requirement that their dispositions be sufficiently in line with that of others so as

to warrant a common interpretation.23
In sum, the standard action-guiding role associated with our shared norma-
tive concepts can play a constitutive role, indirectly or directly, in individuals’
competence with those concepts. The action-guiding role is indirectly constitu-
tive of everyone’s competence with these concepts on the binding model, since
it helps fix the reference of the shared representational tradition and each indi-
vidual participant in the tradition inherits this semantic value from the group.
The standard action-guiding role is also directly constitutive of competence
with these concepts for normal members of the group, since this is part of how
most individuals satisfy the congruence constraint on conceptual competence
with our normative concepts. So unlike matching models of concepts, our
binding model can explain why the action-guiding role of normative concepts
is partly constitutive of conceptual competence without imposing any strict
motivational preconditions on conceptual competence.
Prediction
Finally, let us turn to the question of our epistemic justification for expecting
that normative judgment and motivation typically go hand in hand. The worry
for pure cognitivism, you will recall, is that the link between normative judg-
ment and motivation would have to be based on pure statistical generalization.
This worry might be warranted if conceptual competence were entirely inde-
pendent of motivational dispositions. But as we have seen, the binding model
allows for an internal connection between competence and motivation. If moti-
vational dispositions are partly constitutive of normal conceptual competence,
then this fact helps justify a default assumption connecting normative judgment
with motivation.
As we have seen, the action-guiding role played by normative concepts con-
stitutes a distinctive “center of gravity” within our representational tradition,
so internalizing this action-guiding role helps individuals meet the congruence
requirement. Sometimes the center of gravity of a representational tradition
23. Again, it is possible for the community as a whole to lose this motivational disposi-
tion. As we noted above (footnote 19), this can result in a shift of reference. The loss of the
motivational dispositions within the community would also affect how individuals meet the
congruence constraint on competence with communal concepts: it may take some time, but
eventually individuals will no longer be able to qualify as congruent with the shared repre-
sentational tradition in virtue of motivational dispositions after those dispositions have died
out in the rest of the group.
may be opaque to participants: according to debunking accounts of race, gen-

der, or morality, for instance, we are largely unaware of the social oppressive
roles that are most important to sustaining and justifying the relevant represen-
tational traditions. But the link between normative concepts and motivation is
obvious. As a group we consciously strive to foster the relevant motivational
dispositions among users of normative concepts. So the expectation that nor-
mative judgment is usually associated with motivation does not require any
special theoretical insight: it is a product of the normal process of acquiring
competence with normative concepts. The very same processes that tend to
inculcate a motivational disposition in subjects also tend to inculcate a belief
connecting normative judgment and motivation.
On the binding model this belief about the normal action-guiding role is not
strictly required for conceptual competence: some competent subjects may lack
the relevant expectations due to cynicism or oversight. But when present, this
belief contributes to one’s congruence with the “center of gravity” of normative
concepts. In subjects who lack the motivational disposition itself, say socio-
paths or Martian anthropologists, this belief may play a particularly important
role in establishing congruence with our communal representational tradition.
The upshot, then, is that normal conceptual competence involves both a moti-
vational disposition to act in accordance with one’s normative judgments and a
corresponding belief that competent subjects normally have this disposition.24
It is constitutive of normal conceptual competence to expect people to be moti-
vated by their normative judgments. In short, normal competence involves a
default assumption linking normative judgment and motivation, and normal
competence also ensures that this default assumption is reliable.
So the binding model has the resources to explain why this motivational gen-
eralization is not justified in the same way as the generalization that Americans
tend to like peanut butter. There is no constitutive connection between com-
petence with the concepts of liking peanut butter and being American: accept-
ing the peanut butter generalization plays no role in justifying and sustaining
24. It is worth emphasizing that this is not a version of motivational internalism. Internalism,
as we understand it, places constraints on the motivational states of any conceptually com-
petent subject: e.g., all subjects must be motivated to act in accordance with their norma-
tive judgments, or all subjects must be normally motivated to act in accordance with their
normative judgments. The binding model need not place any motivational requirement on
all competent subjects. Our suggestion in the text is that it is a readily accessible a posteriori
truth that most competent subjects have some motivation to act in accordance with their
normative judgments. This generalization, moreover, is a contingent one: changes in normal
human psychology might alter the dispositions of a community without eo ipso altering the
concept expressed by the relevant terms. (Over time, however, a shift in psychological dispo-
sitions may well ground a shift in the concept expressed.)
Why Go Hybrid? 269
the relevant representational traditions. So this generalization does not come

for free simply in virtue of normal forms of conceptual competence. We must
earn the right to the peanut butter generalization through statistical reasoning
involving the two concepts. In contrast, the motivational generalization does
come for free with normal competence with normative concepts. Once we have
acquired normal competence, we rely on statistical generalization to identify
exceptions to the generalization.
CONCLUSION
Let us take stock. We have argued that a pure cognitivist can explain the phe-
nomena that are standardly cited in favor of hybrid theories. On the binding
model of concepts, being motivated to act in accordance with one’s normative
judgments is not a strict necessary condition for conceptual competence. But
it does not follow that motivation is semantically irrelevant. A binding theorist
can plausibly argue that the relevant motivational dispositions play an impor-
tant role both in constituting normal conceptual competence with our shared
normative concepts and in fixing the reference of these concepts. Moreover,
the expectation that people are normally motivated to act in accordance with
their normative judgments is typically justified on the basis of normal con-
ceptual competence alone. So the binding model of concepts can reconcile
a pure cognitivist account of normative judgment with the fact that the rela-
tion between normative judgment and motivation is not purely external and
contingent. On the binding account, moreover, the link between normative
judgment and motivation is ultimately explained at the level of thought, not
at the level of language. In contrast with hybrid theories, the link is not simply
established by substantive conventions for the use of normative terms, nor is it
explained by pragmatic rules governing particular conversational contexts. It
is anchored in the nature of our shared normative concepts.25
References
Anderson, John. 1977. “Memory for Information about Individuals.” Memory and
Cognition 5:430–442.
Appiah, K. Anthony. 1996. “Race, Culture, Identity: Misunderstood Connections.” In
Color Consciousness, ed. K. Anthony Appiah and Amy Gutmann, 30–105. Princeton,
NJ: Princeton University Press.
25. We would like to thank audiences at Monash University and at the Edinburgh Conference
on Hybrid Expressivism for very helpful comments on earlier versions of this paper. Special
thanks to the conference organizers, Guy Fletcher and Mike Ridge, and to Josh May.
Bach, Kent. 1987. Thought and Reference. Oxford: Oxford University Press.

Analysis 60:268–279.
Bedke, M. S. 2009. “Moral Judgment Purposivism: Saving Internalism from Amoralism.”
Philosophical Studies 144:189–209.
Blackburn, Simon. 1998. Ruling Passions. Oxford: Oxford University Press.
89:169–203.
Burge, Tyler. 1988. “Individualism and Self-Knowledge.” Journal of Philosophy
85:649–665.
Chalmers, David. 2002. “On Sense and Intension.” Philosophical Perspectives
16:135–182.
Copp, David. 2001. “Realist-Expressivism: A Neglected Option for Moral Realism.”
Social Philosophy and Policy 18:1–43.
Cumming, Sam. 2013a. “From Coordination to Content.” Philosophers’ Imprint 13:1–17.
———. 2013b. “Creatures of Darkness.” Analytic Philosophy 54: 379–400.
Dickie, Imogen. 2010. “We Are Acquainted with Ordinary Things.” In New Essays on
Singular Thought, ed. Robin Jeshion, 213–245. Oxford: Oxford University Press.
Dreier, James. 1990. “Internalism and Speaker Relativism.” Ethics 101:6–26.
Evans, Gareth. 1973. “The Causal Theory of Names.” Proceedings of the Aristotelian
Society, supp., 47:187–208.
———. 1982. The Varieties of Reference. New York: Oxford University Press.
Fiengo, Robert, and Robert May. 1994. Indices and Identity. Cambridge, MA: MIT Press.
———. 2006. De Lingua Belief. Cambridge, MA: MIT Press.
Fine, Kit. 2007. Semantic Relationalism. Oxford: Blackwell.
Finlay, Stephen. 2005. “Value and Implicature.” Philosophers’ Imprint 5:1–20.
Forbes, Graeme. 1989. “Cognitive Architecture and the Semantics of Belief.” Midwest
Studies in Philosophy 14:84–100.
Frege, Gottlob. 1997. “On Sinn and Bedeutung (1892).” In The Frege Reader, ed. Michael
Beaney, 151–171. Oxford: Blackwell.
Gibbard, Allan. 1990. Wise Choices, Apt Feelings: A Theory of Normative Judgment.
Cambridge, MA: Harvard University Press.
Gibbard, Allan. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Grice, H. P. 1969. “Vacuous Names.” In Words and Objections, ed. David Davidson and
Jaakko Hintikka, 118–145. Dordrecht, Netherlands: Reidel.
Haslanger, Sally. 2000. “Gender and Race: (What) Are They? (What) Do We Want Them
to Be?” Noûs 34:31–55.
Hawthorne, John, and Ernest Lepore. 2011. “On Words.” Journal of Philosophy
108:447–485.
Heim, Irene. 1983. “File Change Semantics and the Familiarity Theory of Indefinites.”
In Meaning, Use, and Interpretation of Language, ed. Rainer Bäuerle, Christoph
Schwarze, and Arnim von Stechow, 164–189. Berlin: de Gruyter.
———. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York: Garland.
Why Go Hybrid? 271
Jackson, Frank. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis.

Jeshion, Robin. 2002. “Acquaintanceless De Re Belief.” In Meaning and Truth:
Investigations in Philosophical Semantics, ed. Joseph K. Campbell, Michael O’Rourke,
and David Shier. Topics in Contemporary Philosophy, 53–78. New York: Seven
Bridges.
Kahneman, David, and Anne Treisman. 1984. “Changing Views of Attention and
Automaticity.” In Varieties of Attention, ed. Raja Prasuraman and D. R. Davies, 29–61.
New York: Academic.
Kaplan, David. 1990. “Words.” Proceedings of the Aristotelian Society, supp., 64:93–119.
Karttunen, Lauri. 1974. “Discourse Referents.” Syntax and Semantics 7:363–385.
Kitcher, Philip. 2007. “Does ‘Race’ Have a Future?” Philosophy and Public Affairs
35:293–317.
Lawlor, Krista. 2001. New Thoughts about Old Things: Cognitive Policies as the Ground of
Singular Concepts. New York: Garland.
Lewis, David. 1980. “Index, Context, and Content.” In Philosophy and Grammar, ed.
Seig Kanger and Sven Öhman. 79–100. Dordrecht, Netherlands: Reidel.
Mallon, Ron. 2006. “Race: Normative, Not Metaphysical or Semantic.” Ethics
116:525–551.
McDowell, John. 1984. “De Re Senses.” Philosophical Quarterly 34: 283–294.
Millikan, Ruth. 1994. “On Unclear and Indistinct Ideas.” Philosophical Perspectives
8:75–100.
———. 2000. On Clear and Confused Ideas. Cambridge: Cambridge University Press.
Parfit, Derek. 1971. “Personal Identity.” Philosophical Review 80:3–27.
Perry, John. 1980. “A Problem about Continued Belief.” Pacific Philosophical Quarterly
61:317–332.
———. 2001. Reference and Reflexivity. Palo Alto, CA: Center for the Study of Language
and Information.
Pinillos, N. Ángel, 2011. “Coreference and Meaning.” Philosophical Studies 154:301–324.
Prinz, Jesse. 2007. The Emotional Construction of Morals. Oxford: Oxford University
Press.
Putnam, Hilary. 1972. “The Meaning of ‘Meaning.’ ” Minnesota Studies in the Philosophy
of Science 7:131–193.
Recanati, François. 1993. Direct Reference: From Language to Thought. Oxford: Blackwell.
———. 2012. Mental Files. Oxford: Oxford University Press.
Ridge, Michael. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116:302–336.
———. 2007. “Ecumenical Expressivism: The Best of Both Worlds?” In Oxford Studies in
Metaethics, vol. 2, ed. Russ Schafer-Landau, 51–76. Oxford: Oxford University Press.
Schroeder, Mark. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119:257–309.
Schroeter, Laura. 2007. “The Illusion of Transparency.” Australasian Journal of
Philosophy 85:1–22.
———. 2008. “Why Be an Anti-Individualist?” Philosophy and Phenomenological
Research 77:105–141.
———. 2012. “Bootstrapping Our Way to Samesaying.” Synthese 189:177–197.
———.2013. “Are Concepts Creatures of Darkness? Critical Discussion of Robert
Stalnaker’s Our Knowledge of the Internal World.” Analytic Philosophy 54:277–292.
Schroeter, Laura, and François Schroeter. 2009a. “Reasons as Rightmakers.” Philosophical

Explorations 12:279–296.
———. 2009b. “A Third Way in Metaethics.” Noûs 43:1–30.
———. 2014. “Normative Concepts: A Connectedness Model.” https://www.academia.
edu/916243/Normative_Concepts_a_connectedness_model_--_Revised_version.
Strawson, Peter. 1974. Subject and Predicate in Logic and Grammar. London: Methuen.
Tresan, Jon. 2006. “De Dicto Internalist Cognitivism.” Noûs 40:143–165.
van Roojen, Mark. 2010. “Moral Rationalism and Rational Amoralism.” Ethics
120:495–525.
Zack, Naomi. 2002. Philosophy of Science and Race. New York: Routledge.
12
The Truth in Hybrid Semantics

MARK SCHROEDER n
It is a familiar idea that the meanings of some words cannot be fully captured
through a compositional characterization of truth conditions. So far as its
semantic contribution to truth-conditional compounds goes, “but” is the same
as “and,” but “but” means something more than “and,” having a conventionally
associated secondary content of some sort. It is intuitive to many that this same
characterization is satisfied by racial and ethnic slurs, among a wide range of
other words in natural language. H. P. Grice (1989) called this sort of meaning
“conventional implicature.” With Grice as their forebear and the example of
racial and ethnic slurs as a model, a growing number of contemporary theorists1
have proposed that distinctively moral vocabulary is special precisely because
its meaningfulness exceeds core truth-conditional content. Such theorists offer
hybrid accounts of moral language, on which a further, often noncognitivist
component is appended to a core, cognitivist account.
In my view hybrid theories of this kind are primarily of interest not because
they are a way of reviving the nondescriptivist or irrealist aspirations of tra-
ditional forms of noncognitivism but instead because they offer a promising
tool for naturalist or reductive realist theories. If moral facts or properties are
simply natural properties but moral words are loaded ways of speaking about
those properties and moral thoughts are loaded ways of thinking about them,
as the hybrid model suggests, then the hybrid naturalist can offer an account
of what is distinctive and important about moral language and thought, despite
the place of its subject matter in the natural world. At any rate, that is the
1. See in particular Barker 2000; Boisvert 2008; Copp 2001, 2009, this volume; Hay 2011;
Strandberg 2011.
background motivation for the kind of hybrid view in which I will be interested
in this chapter.2
As I argued in an earlier paper,3 one of the most important issues facing
hybrid metaethical theories concerns the interaction of moral vocabulary and
attitude verbs. To attain any distinctive virtues not sharable by a similar purely
cognitivist theory, I argued, the hybrid theorist must hold that the second-
ary content of moral terms, although it “projects” through truth-conditional
connectives, contributes to the primary, truth-conditional content of attitude
verbs—in particular “believes.” In contrast, I argued, the secondary content of
racial slurs seems to project through attitude verbs and does not seem to con-
tribute to their truth-conditional content. So if moral vocabulary fits the hybrid
theorist’s needs, I argued, racial slurs are not really such a close analogy after all.
This chapter takes up where that earlier paper left off by presenting a simple,
flexible model for secondary contents. This model shows clearly that there is
nothing fundamentally strange about secondary contents affecting the core,
“truth-conditional” content of attitude verbs, and I will argue that there are
clear examples of this in natural language—though racial slurs are not among
them. Then I will take up the all-important question of how we should expect
the word “true” to work in a language that follows this flexible model. This
will allow me to restate the worry about whether moral vocabulary could really
work as the hybrid theorist needs in a sharper way.
T W O PAT T E R N S O F PR O J E CTI O N
As already noted, prominent examples of words whose conventional mean-

ings appear to encode more than we think of as figuring into their core,
“truth-conditional” contents are numerous. This is often said, for example, not
only about “but” and “even” but about racial and ethnic slurs, of which I will take
“cheesehead” to be a mild exemplar for purposes of this chapter.4 These words
share several important patterns. First, when speakers use simple sentences
involving these words, they communicate more than one thing. A speaker who
says “Shaq is huge but agile” communicates both that Shaq is huge and agile and
2. Compare Boisvert 2007; Copp 2001; and Hay 2011 and contrast Ridge 2006, 2007. David
Copp’s motivations and commitments differ from those of Daniel Boisvert and Ryan Hay
in an important way that I will discuss in the fourth section. Jon Tresan (2006, 2009) also
sketches a view with many of the same features as that discussed here.
3. “Hybrid Expressivism: Virtues and Vices,” published in Ethics (2009) and reprinted in the
2009 Philosophers’ Annual.
4. See http://en.wikipedia.org/wiki/Cheesehead.
The Truth in Hybrid Semantics 275
also that she sees a contrast between being huge and being agile.5 Similarly, a
speaker who says “the neighborhood is filling up with cheeseheads” communi-
cates both that the neighborhood is filling up with people from Wisconsin and
also that she does not think very well of people from Wisconsin.
The reason we do not take these second aspects of meaning to be “truth con-
ditional,” however, is that they seem to make a different contribution to so-called
truth-conditional connectives like negation and disjunction. If a speaker says
“Shaq is not huge but agile,” we read her as communicating not that either Shaq
is not huge, or he is not agile, or she does not see a contrast between the two
but rather that he is not both huge and agile and that she does see a contrast
between the two. Similarly, if she says “the neighborhood is not filling up with
cheeseheads,” we do not take her to have communicated that either the neigh-
borhood is not filling up with people from Wisconsin or she does think well of
people from Wisconsin but rather that the neighborhood is not filling up with
people from Wisconsin and that she does not think very well of people from
Wisconsin. So if negation operates on truth conditions, then the extra semantic
contribution of “but” and “cheesehead” does not contribute to truth conditions.
Similar points apply to other constructions: someone who says “either Shaq
is huge but agile or my eyes deceive me” is not committed to the claim that
Shaq is both huge and agile, but she does communicate that she sees a contrast
between being huge and being agile. Similarly, someone who says “either the
neighborhood is filling up with cheeseheads, or I still can’t tell the difference
between Wisconsin and Minnesota accents” is not committed to the claim
that the neighborhood is filling up with people from Wisconsin, but she does
communicate that she does not think very well of people from Wisconsin.
In general the two intuitive parts of the meanings of “but” and “cheesehead”
make distinct contributions to complex sentences formed by “truth-conditional”
contexts like negation and disjunction. One of these parts appears to be affected
by the connectives in the familiar truth-conditional way, but the other appears
to merely be “passed up” to the more complex sentence. When a piece of infor-
mation is inherited by a complex sentence directly from one of its parts, we
may say, following the literature on presupposition, that it projects. What we
have observed so far is that “but” and “cheesehead” appear to be conventionally
associated with information that projects through constructions like negation
and disjunction. It is natural to conjecture that this may be true of all complex
sentences in which these terms figure—that the “extra” semantic content asso-
ciated with “but” and “cheesehead” will project through all ways of embedding
them in more complex constructions. But that conjecture would be incorrect.
5. This example comes from Bach 1999.

Though it has some promise for the case of slurs, it is clearly incorrect for “but.”
The main source of the difficulty arises from attitude verbs.
To see this, first consider “Jill believes that the neighborhood is filling up
with cheeseheads.” Arguably, someone who thinks poorly of people from
Wisconsin can use this sentence to report Jill’s belief even if she recognizes
that Jill is soft on people from Wisconsin, but in saying so, she communi-
cates her own poor regard for people from Wisconsin. If that is right, then
the extra content of “cheesehead” can at least sometimes project through atti-
tude verbs like “believes” (though contrast “said,” which arguably behaves dif-
ferently). However, in contrast, consider “Marv believes that Shaq is huge but
agile.” Someone who says this does not communicate that she sees any contrast
between Shaq’s being huge and his being agile, and she speaks falsely unless
Marv sees such a contrast.6 So the extra semantic content of “but” does not
project through “believes”; on the contrary, it directly contributes to the truth
conditions of “believes” reports.
This contrast is important, because as we will see in what follows it is impor-
tant for the purposes of the hybrid theorist in whom I will be interested that
moral words like “wrong” pattern with “but” rather than with “cheesehead” in
this respect. This is because it is an important part of the prospect for the hybrid
view to tell us something distinctive about the practical import of moral lan-
guage and thought that sentences like “Karen believes that stealing is wrong”
are true only if Karen has an appropriately related noncognitive attitude.
Though slurs like “cheesehead” contrast with “but” in their projection behav-
ior, it is important to distinguish both from presuppositions proper. Typically,
when a sentence Q has a presupposition P, that presupposition will not be car-
ried by the conditional sentence if P, then Q. For example, “the king of France
has a mustache” presupposes that France has a king, but “if France has a king,
then the king of France has a mustache” does not. In other words, condition-
als are filters for presuppositions, because whether the presupposition of the
consequent passes to the whole sentence depends on what figures in the ante-
cedent. By contrast, the additional content of “Mark is a good driver but from
Wisconsin” seems to be that most people from Wisconsin are not good drivers,
but “if most people from Wisconsin are not good drivers, then Mark is a good
driver but from Wisconsin” does not seem to be free of the extra commitment
carried by its consequent. And I cannot see how to construct a sentence of the
form “if P, then Mark is a cheesehead” that does not implicate the speaker in
disdain for people from Wisconsin. So it is best, I think, to think of each of
these as a different phenomenon from that of presupposition worth investigat-
ing in its own right.
6. Compare again Bach 1999.

B I G H Y P OT H E SI S SE M AN TI CS
In “Hybrid Expressivism: Virtues and Vices” (Schroeder 2009) I argued that a

promising hybrid theory needs the secondary content of moral words to inter-
act with attitude verbs like “believes” in the way that I have just argued that
“but” does. I called this the “Big Hypothesis” and argued that the case of slurs
should not give us any optimism that it is correct, since slurs do not in fact work
in this way. But it is important not to misunderstand the force of this point. The
fact that slurs do not work in this way does not show that moral terms can-
not—it just shows that insofar as they do, slurs are only an imperfect analogy
for moral terms.7 Indeed it is very easy to describe how a language could work
in which there are both terms like “but” whose secondary contents interact with
attitude verbs and terms like “cheesehead” whose secondary contents do not.
All that we need to do is to allow for two different kinds of secondary contents.
In this section I will set out a simple, flexible framework for thinking about
secondary contents. What makes this framework flexible is that it is specifi-
cally formulated to allow a distinction between secondary contents that project
through attitude verbs (as for “cheesehead”) and those that instead contribute to
their truth conditions (as for “but”). So to make this distinction, each sentence
will be assigned to a triple of semantic values: a core content, a primary set, and
a secondary set. The primary set and secondary set are each to be thought of as
sets of “extra” content possessed by the sentence, and for the purposes of this
chapter, I will assume that their contents are states of mind—mental properties.
However, for simplicity I will take the core content to be a conventional propo-
sition. It will be easy to adapt the framework offered here to change either of
these two assumptions; indeed this point will be important later.
So our semantics is going to work by assigning every sentence of the lan-
guage to a triple—a core content, a primary set, and a secondary set. Let us say
that [S]‌0a denotes the core content associated with “S” relative to assignment a,
that [S]1a denotes the primary set associated with “S” relative to assignment a,
and that [S]2a denotes the secondary set associated with “S” relative to assign-
ment a. So we need our semantics to assign each well-formed formula “S” to
the triple <[S]0a,[S]1a,[S]2a>. Because only [S]0a represents “core” content, it is the
only thing we want our “truth-conditional” connectives to substantively affect.
With respect to such connectives, the primary and secondary sets will behave
in the same way. But attitude verbs will treat them differently, allowing the pri-
mary set but not the secondary set to contribute to the core content of attitude
ascriptions. Those are the ideas encapsulated in the following rules:
7. Hay (2011) argues persuasively that a more promising analogy for moral terms are what
he calls “general pejoratives,” like “jerk” and “asshole.”
Connectives:
[A and B]0a = TRUE iff [A]‌0a = TRUE and [B]0a = TRUE

[A and B]1a = [A]‌1a ∪ [B]1a
[A and B]2a = [A]‌2a ∪ [B]2a
[~A]0a = TRUE iff [A]‌0a ≠ TRUE
[~A]1a = [A]‌1a
[~A]2a = [A]‌2a
[∀x(A)]0a = TRUE iff for every assignment b that differs from a at most in
what it assigns to x,
[A]‌0b = TRUE
[∀x(A)]1a = [A]‌1a.
[∀x(A)]2a = [A]‌2a.
Attitude verbs:
[x believes that A]0a = TRUE iff o believes [A]‌0a and for all P∈ [A]1a, o
instantiates P, where a assigns o to x.
[x believes that A]1a = ∅, for all a.
[x believes that A]2a = [A]‌2a.
[x wonders whether A]0a = TRUE iff o wonders whether [A]‌0a and for all
P∈ [A]1a, o instantiates P, where a assigns o to x.
[x wonders whether A]1a = ∅, for all a.
[x wonders whether A]2a = [A]‌2a.
These rules distinguish between primary and secondary contents in precisely

the way anticipated. So all it takes to see how these principles work in action
is to introduce some vocabulary that contributes to the primary and secondary
contents.
Predicates:
[x is from Wisconsin]0a = TRUE iff o is from Wisconsin, where a assigns x

to o
[x is from Wisconsin]1a = ∅, for all a
[x is from Wisconsin]2a = ∅, for all a
[x can drive in traffic]0a = TRUE iff o can drive in traffic, where a assigns x
to o
[x can drive in traffic]1a = ∅, for all a
[x can drive in traffic]2a = ∅, for all a
[x is a cheesehead]0a = TRUE iff o is from Wisconsin, where a assigns x to o.

[x is a cheesehead]1a = ∅, for all a.
[x is a cheesehead]2a = {DISDAIN(people from Wisconsin)}, for all a.
New Connective:
[A but B]0a = TRUE iff [A]‌0a = TRUE and [B]0a = TRUE

[A but B]1a = [A]‌1a ∪ [B]1a ∪ {bf(A contrasts with B)}
[A but B]2a = [A]‌2a ∪ [B]2a
On this picture the extra contents associated with “but” and with “cheeseh-
ead” behave differently with respect to attitude verbs. Someone who says “Mark
believes that Schroeder is a cheesehead” communicates her own disdain for
people from Wisconsin rather than telling us about Mark’s disdain, but some-
one who says “Mark believes that Schroeder is from Wisconsin, but Schroeder
can drive in traffic” communicates that Mark, rather than the speaker, sees a
contrast between being from Wisconsin and being able to drive in traffic. And
finally, someone who says “Caroline believes that Schroeder is a cheesehead but
can drive in traffic” communicates that he, rather than Caroline, has disdain
for people from Wisconsin but that Caroline, rather than he, sees a contrast
between being from Wisconsin and being able to drive in traffic.
We can think of the framework described here as a minimal model for marking
the distinction between terms that behave in these two different ways. The model
is highly flexible both in that it leaves open that attitude verbs other than “believes”
may treat the primary and secondary contents differently and that it allows in
principle for words that affect both the primary and the secondary contents. It is
also easy to use it to implement other interesting views. For example, Ryan Hay’s
view about the semantics of “jerk” can be implemented (roughly) as follows:8
Jerk:
[x is a jerk]0a = TRUE iff o is exhibits sufficient disregard for others, where

a assigns x to o.
[x is a jerk]1a = {RESENTMENT(people who exhibit sufficient disregard
for others)}, for all a.
[x is a jerk]2a = ∅, for all a.
8. Officially, Hay (2011) rejects the idea that there is any straightforward paraphrase of the
core content of “jerk,” and the formulation here is more concrete about the associated atti-
tude than Hay is, but this is essentially the shape of his picture.
It is easy to use our compositional principles to derive the conclusion that some-
one who says “Caroline believes that not everyone who believes no cheeseheads
can drive in traffic is a jerk” communicates that Caroline resents people with
sufficient disregard for others without expressing this attitude herself but does
express her own disdain for people from Wisconsin.
LO G I C A N D I NFE R E N CE
The foregoing framework makes it easy to observe two important lessons about
hybrid metaethical theories: first, that for purposes of understanding logic and
inference, it does not matter whether the “extra” content of moral vocabulary
is primary or secondary content but, second, that for purposes of attaining
the distinctive advantages of hybrid theories over ordinary cognitivism, it is
important that moral terms contribute to primary content. I will take these two
lessons one by one, focusing on logic and inference in this section and on the
distinctive advantages over cognitivism in the next.
The problem of accounting for logic and inference can be broken down into
two parts: those concerned with inconsistency, and those concerned with infer-
ence.9 Inconsistency is the easier of these two issues, because it can be fully
accounted for in terms of core contents without the need to worry about pri-
mary or secondary contents. Our compositional rules guarantee that any two
sentences are logically inconsistent just in case no interpretation could make
their core contents both true. But inference is potentially harder to explain. In
general logically valid arguments should commit a thinker who endorses their
premises to endorsing their conclusions. But since some sentences are associ-
ated with primary and secondary attitudes, endorsing the conclusion of a valid
argument often involves having such further states of mind in addition to the
belief in the core content. For example, endorsing “P and Q” should not commit
a thinker to endorsing “P but Q.” It follows that understanding how endorsing
the premises of a valid argument can commit a thinker to endorsing their con-
clusion requires paying attention to primary and secondary sets.
Fortunately, both primary and secondary sets have exactly the feature that is
required to solve this problem. I will first illustrate with two examples that we
have already introduced and then generalize. First, consider a simple modus
ponens argument involving “cheesehead”:
P1 Mark cannot drive in traffic.

P2 If Mark cannot drive in traffic, then Mark is a cheesehead.
9. For further discussion, see Schroeder 2009; 2010, chap. 6.

C1 Mark is a cheesehead.
Endorsing the premises of this argument should transparently commit some-

one to endorsing its conclusion (though of course this commitment can be given
up by giving up one of the premises). The conclusion has the core content that
Mark is from Wisconsin and has the secondary set consisting of disdain for people
from Wisconsin. Endorsing the conclusion therefore requires both believing this
primary content and also having disdain for people from Wisconsin. And so a
thinker is committed to endorsing the conclusion only if she is committed to both
this belief and this disdain. There is no puzzle where the commitment to this belief
comes from, because endorsing the premises of the argument requires the thinker
to believe their core contents, and it is a familiar observation that having beliefs
commits you to believing their consequences. But similarly, there is no puzzle
where the commitment to the disdain comes from, because it is also part of the
secondary set for premise P2. That means that if the thinker endorses the premises
of the argument, then she already has disdain for people from Wisconsin.
The exact same explanation goes for a modus ponens argument involving a
nonvacuous primary set (our examples could be more interesting if our lan-
guage were richer):
P3 Mark is from Wisconsin.

P4 If Mark is from Wisconsin, then Mark is a jerk.
C2 Mark is a jerk.
C2 is associated with a primary set including resentment of people who have

sufficient disregard for others, and so to endorse it a thinker must be in that
state of resentment in addition to believing its core content. But this state of
resentment is also part of the primary content of the second premise. And so a
thinker cannot endorse both of the premises without being in this state. So any
thinker who endorses both premises is already in all of the states that are in the
primary set for the conclusion. But believing the core contents of the premises
commits a thinker to believing the core content of the conclusion. So a thinker
who endorses the premises of this argument is committed full stop to endorsing
its conclusion (though again, she can give this commitment up by giving up one
of the premises).10
In general, productive, relevant validities of predicate logic are generally
going to have this property—that states of mind that appear in either the pri-
mary or the secondary set for the conclusion also appear in the primary or
10. Compare Boisvert 2008; Schroeder 2009.

secondary set for one of the premises. This principle only fails when the conclu-
sion introduces new terminology that does not appear in one of the premises.
So, for example, endorsing the premises of the following argument does not
commit you to endorsing its conclusion:
P5 Mark cannot drive in traffic.

C3 Either Mark cannot drive in traffic, or he is a cheesehead.
Though endorsing the premise of this argument commits you to believing

the core content of its conclusion, it does not commit you to endorsing the
conclusion full stop, because to endorse the conclusion full stop you must have
disdain for people from Wisconsin, but nothing about accepting the premise
commits you to this disdain. But this is the right prediction. If “cheesehead” is
a slur, thinkers should not be committed to endorsing sentences involving it,
even by vacuous inferences like disjunction introduction. So I conclude that the
framework described here accommodates just the right amount of inference.
The same will go for moral terms if they follow either the model of “jerk” and
affect the primary set or the model of “cheesehead” and affect the secondary
set—or even if they affect both. So the distinction between primary and sec-
ondary sets is not relevant for accounting for logic and inference—at least not
the inferences that we want from ordinary predicate logic.11
A D VA N TA GE S O V E R C O G NI TI V I SM
In contrast, the difference between primary and secondary contents is crucial

for making good on what I take to be the most significant prospects for a hybrid
metaethical theory to have important advantages over ordinary cognitivist
views. This is because, though it follows from both the view that moral language
contributes to the primary set and the view that moral language contributes to
the secondary set that moral language is special, only on the former view does
it follow that moral thought is special. And a hybrid theory that makes sense of
the specialness of moral language without making sense of the specialness of
moral thought will be very limited in what advantages it can claim.
Consider, for example, the case of David Copp’s realist expressivism.12
According to Copp, moral language is special because moral sentences con-
ventionally simplicate that the speaker has a corresponding attitude. But Copp
11. I will turn to a different pattern of inference involving the word “true” in the fifth section.
12. Copp 2001, 2009, this volume.
denies that having this attitude is required for having moral beliefs. So accord-
ing to Copp, you can have moral beliefs without having any attitude that would
motivate you to act as a result. It is not surprising that Copp should defend
such a view, because he has long advocated judgment externalism about moral
judgments. But this leads to the question as to what Copp gains by hybridizing
his view that he could not have within an ordinary cognitivist view. According
to Copp, what he gains is an explanation of why there is an illusion of judgment
internalism on the grounds that when speakers overtly assert moral sentences,
we expect them to be motivated, because those sentences conventionally sim-
plicate that the speaker has the attitude required for her to be motivated.13
But this does not strike me as a particularly compelling explanation of why
moral judgment internalism would seem to be true. Judgment internalism, after
all, says not only that speakers who assert moral claims can be expected to be
motivated but that people who genuinely believe moral claims will be motivated.
It may help to compare an example that may plausibly be treated as affecting
secondary content, the formal/informal distinction for the second-person pro-
noun. Complicating the simple language that we have been describing enough
to formally incorporate the formality distinction for second-personal pronouns
would take us further astray than the resulting payoff would justify, but the
idea is that second-personal pronouns like vous/tu, Sie/du, and the now-dead
“you/thou” each make a contribution to a sentence’s secondary set, intimating
an expectation either of familiarity or of distance on the part of the speaker
toward her addressee. It makes sense to incorporate this into the secondary set
rather than the primary set, because when a speaker says, “Julia glaubt, daβ du
Deutsch sprichst,” he communicates not that Julia expects familiarity with his
addressee but that he does.14
The formal/informal distinction makes for a useful test of Copp’s hypoth-
esis of how there could be an illusion of judgment internalism, because we do
expect someone who asserts “Du sprichst Deutch” to expect familiarity with
her audience, but this would not lead us in any way to suspect, as we might put
it, that jemand, der glaubt, daβ du Deutch sprichst would expect to be familiar
with you. Nor would we describe someone who did suspect this as succumbing
to a natural but misleading intuition. Since there is no illusion in the formal/
informal case, that leads me to doubt how there could be a similar illusion in
the moral case.
13. “Second, I will contend that it can explain the intuitions that lead people to accept moral
judgment internalism (or judgment motivational internalism) in a way that is compatible
with externalism” (Copp this volume).
14. For a wide collection of examples of words and constructions that clearly work along the
model of secondary sets, including discussion of the formal/informal distinction, see Potts 2005.
I conclude that hybridization offers much deeper prospects to explain some-

thing fundamentally distinctive about morality only if it promises to tell us some-
thing not only about moral language but about moral thought. And it is only on
the view that moral language affects primary content that this is so. The reason for
this is simple. We can only succeed in saying that there is something special about
moral belief, using the sentence “there is something special about moral belief,” if
the meaning of the word “belief ” allows us to say this. So whatever we say about
moral belief needs to be constrained by the semantics for “belief ” or “believes.”15
The distinction between primary and secondary sets makes all the difference
for whether the distinctive extra feature of moral language affects the content
of claims of the form “x believes that stealing is wrong.” So it is what allows us
to coherently use the word “believes” to say that something more is required to
believe that stealing is wrong than to believe its core content. Here is a simple
example of such a view:
Wrong, first pass:
[x is wrong]0a = TRUE iff o fails to maximize happiness, where a assigns x

to o.
[x is wrong]1a = {DESIRE(not to do things that fail to maximize
happiness)}, for all a.
[x is wrong]2a = ∅, for all a.
On this simple picture “Jim believes that stealing fails to maximize happiness”
can be true even if “Jim believes that stealing is wrong” is not. So “Jim believes
that stealing fails to maximize happiness, but Jim does not believe that stealing is
wrong” can be true. So, semantically descending, we can agree that Jim believes
that stealing fails to maximize happiness but does not believe that stealing is
wrong. Hence on this view we can make sense of the idea that there is more to
moral belief than belief in some ordinary, natural fact.16
On the particular view that I have just described, the “more” that there is to
moral belief is exactly what would be needed for it to turn out that judgment
internalism is true, even if ordinary beliefs never motivate by themselves but only,
as the Humean theory of motivation claims, in connection with a desire. This is
15. Observe that when Copp (this volume) tries to extend his view to argue that there is
something distinctive of moral thought, he is forced to use quotation and talk about what is
required to “think that Brenda is a ‘pom.’ ”
16. And by this we mean not only that there is more to the belief that something is “wrong”
than belief in some natural fact, as Copp (this volume) can allow, but that there is more to
the belief that it is wrong than belief in some natural fact. This further claim is not available
to Copp, since he rejects the Big Hypothesis.
because it tells us that “Jim believes that stealing is wrong” is not true unless Jim
believes that stealing fails to maximize happiness and desires not to do things
that fail to maximize happiness. Hence, semantically descending again, it tells us
that Jim does not believe that stealing is wrong unless Jim has both this belief
and this desire. But the combination of this belief with this desire is exactly what
we would expect to motivate Jim not to steal. Hence placing the extra content of
“wrong” into the primary set is exactly what we would need to validate judgment
internalism.
Yet a hybrid theory could also preserve the prediction that there is more to
moral belief than any ordinary descriptive belief without endorsing any strong
form of judgment internalism.17 To do so we just need to change what we put in
the primary set for “wrong” sentences. Here is one such alternative:
Wrong, second pass:
[x is wrong]0a = TRUE iff o fails to maximize happiness, where a assigns x

to o.
[x is wrong]1a = {DISAPPROVAL(things that fail to maximize happiness)},
for all a.
[x is wrong]2a = ∅, for all a.
If it is possible to believe that stealing is F and to disapprove of F things with-

out being motivated not to steal, then this version of the hybrid theory will not
validate judgment internalism about “wrong” beliefs. But it will maintain the
feature—preserved by all views on which “wrong” contributes to the primary
set—that there is more to moral beliefs than simply beliefs in natural, descrip-
tive facts. This is, I think, a very intelligible advantage to be interested in over
nonhybrid naturalist or reductive views.
As we have just seen, hybrid theory that holds that “wrong” contributes to
the primary set can but does not need to validate the truth of judgment inter-
nalism about “wrong.” This version of judgment internalism is what Jon Tresan
(2006) calls de dicto internalism, according to which necessarily someone who
believes that stealing is wrong will be motivated not to steal. Tresan contrasts
de dicto internalism with de re internalism, according to which the belief that
stealing is wrong is such that it necessarily motivates someone not to steal.
Tresan advocates a view on which de dicto internalism is true but de re inter-
nalism is false. On Tresan’s view the belief that stealing is wrong is just an ordi-
nary descriptive belief that could exist in the absence of any motivation, but it
17. Compare Hay this volume.

does not count as a moral belief or as the belief that stealing is wrong unless
that motivation is present.
The hybrid theory in I am interested in this chapter is consistent with but does
not entail Tresan’s de dicto internalist cognitivism. Whether we develop the hybrid
view in a way that leads to Tresan’s view or in a way that also endorses de re inter-
nalism depends on what we take phrases like “the belief that stealing is wrong”
to refer to. Tresan’s suggestion that these phrases refer to the ordinary descriptive
belief that stealing fails to maximize happiness but do so by a description that is not
satisfied unless a further motivation is present is consistent with everything that
I have said so far. On this interpretation the hybrid theory offers an explanation of
some of the puzzling features of Tresan’s view, which he himself acknowledges.18
T R U T H I N H Y B R I D SE M AN TI CS
So far in this chapter I have given a simple model for how a “Big Hypothesis”
semantic framework might work—one that allows for a distinction between
two different ways the conventional meaning of a word can exceed the contri-
bution that it makes to the so-called truth-conditional contexts of negation,
conjunction, and disjunction. This distinction, which I have called the distinc-
tion between the “primary” and “secondary” sets associated with a sentence,
allows us to distinguish between words whose extra conventionally associ-
ated character contributes to what it takes to have the associated beliefs and
those that do not. Arguably, “but” and “cheesehead” fall on opposite sides of
this distinction in ordinary English, and I have argued that the advantages of
adopting a hybrid theory that holds that “wrong” patterns with “cheesehead”
are illusory. The most that we can get from such a hybrid theory is that there
is something distinctive of moral assertions but not that there is anything—or
even an illusion of there being anything—distinctive about the thought that
something is wrong. To hold that there is something distinctive of the thought
that something is wrong, the hybrid theorist must hold that “wrong” patterns
with “but” so that it can affect the semantics of thought words like “thought”
and “believes.” To say this is just to semantically ascend from the claim that
there is something distinctive of beliefs and thoughts about what is wrong.
I now turn to the question of how the word “true” fits into our semantic
picture. In particular how should “true” interact with the primary and sec-
ondary sets associated with sentences? Pictures like the one described in
18. In fact, however, I think it is more natural for the hybrid theorist to identify the belief
that stealing is wrong with the conjunctive state that figures in the primary content of “x
believes that stealing is wrong.” And that is the way that I interpret the hybrid view in
Schroeder 2013. The resulting view shares many of the advantages carefully documented by
the last few sections are often described as ones on which meaning exceeds
truth-conditional content. On this view only core content is truth conditional,
which suggests that “true” and “false” should only report on core content. This
suggests, at a first pass, the following simple picture for “true”:
[it is true that P]0a = TRUE iff [P]‌0a = TRUE

[it is true that P]1a = ∅
[it is true that P]2a = ∅
This picture yields the strange conclusion that you can assert “it is true that
Mark is a cheesehead” without implicating yourself in disdain for people from
Wisconsin or assert “it is true that Mark is from Wisconsin but can drive in traf-
fic” without implicating yourself in any contrast between being from Wisconsin
and being able to drive in traffic. But both of these seem wrong.
But perhaps the idea that only core content is “truth conditional” should not
be identified with the idea that “true” does nothing other than to report on core
content. Perhaps it should instead be identified with the idea that the only thing
“true” reports on is core content. In other words, perhaps in addition to reporting
on core content sentences involving “true” may also, at least in some cases, have
nonempty primary or secondary sets, because “true” does not block the projec-
tion of primary or secondary sets. Here is what such a view would look like:
[it is true that P]0a = TRUE iff [P]‌0a = TRUE

[it is true that P]1a = [P]‌1a
[it is true that P]2a = [P]‌2a
This view makes perfect sense of why you would assert “it is true that Mark is
a cheesehead” only if you would also be willing to assert “Mark is a cheesehead”
and similarly for “it is true that Mark is from Wisconsin but can drive in traffic.”19
Tresan (2006), including maintaining that representational and motivational states are mod-
ally separable, but parts from his view in denying that “belief,” in natural language, picks out
the modally separable state that is exhausted by its representational character. This allows the
view to maintain, contra Tresan, that the belief that stealing is wrong cannot exist without
being the belief that stealing is wrong. However, Tresan’s view has at least a couple of advan-
tages. Because he is not committed to the background condition on a belief counting as a
moral belief being a purely psychological condition, he can avail himself of a wider range of
such background conditions than can plausibly be identified as part of the belief itself. For
example, following suggestions by Philippa Foot (1978a) and James Dreier (1990) in Tresan
2006, he explores the possibility of social conditions. He also suggests that de dicto internal-
ist cognitivism can mix and match any cognitivist and expressivist views.
19. Contra Bach 1999, this view preserves the thesis that conventional implicatures like that
of “but” do not contribute to truth conditions while honoring the idea that when we say that
Moreover, in company with our clause for “believes,” it makes sense of why mini-
mally reflective agents will believe that it is true that Mark is from Wisconsin but
can drive in traffic just in case they believe that Mark is from Wisconsin but can
drive in traffic. These seem like plausible predictions.
These predictions are plausible enough for sentences of the form “it is true
that P.” But most interesting sentences containing the word “true” do not have
this form. After all, if “it is true that P” and “P” are equivalent, then we can get
by just as well using only the latter as using the former. So interesting sentences
involving “true” say things like “what Caroline believes is true” and “everything
that cheesehead said is true.” We can use such sentences and accept them even
if we do not know what Caroline believes or everything that the person being
referred to as a “cheesehead” said.
Paying attention to these sentences helps us zero in on which words in “it
is true that Mark is a cheesehead” are responsible for the contents of its sec-
ondary set. This is because “what Caroline believes is true” does not implicate
the speaker in disdain for people from Wisconsin, even if it happens to be
the case that Caroline would express her belief using the sentence “Mark is
a cheesehead.” For the speaker of “what Caroline believes is true” might not
even know what Caroline believes about Mark. In contrast, “everything that
cheesehead said is true” does implicate its speaker in disdain for people from
Wisconsin—even if the person being referred to as a “cheesehead” does not
have any beliefs they would express by using the word “cheesehead.” So in
both the case of “it is true that Mark is a cheesehead” and that of “everything
that cheesehead said is true,” the secondary set contains disdain for people
from Wisconsin, but in neither case is this contributed by the word “true”
any more than it is contributed by the word “or” in “either Mark is a cheeseh-
ead, or he is not from Wisconsin.”20 Instead, it is contributed by the word
“cheesehead” and merely projected upward through the word “true.” So this
tells us that secondary sets project upward through “true,” just as they proj-
ect through “truth-conditional” connectives and attitude verbs like “believes.”
This should not be a surprise.21
it is true that Shaq is huge but agile we are also committing to the claim that it is true that
there is a contrast between being huge and being agile. But on this view this is because the
extra conventionally associated content projects upward through “true” rather than because
it contributes to the core content of “true” sentences.
20. Contrast Michael Ridge (2009), on whose view all uses of “true” express the very same
attitude that is expressed by normative words like “wrong” and “reason.”
21. It would be easy to amend our semantics to show how this works and to allow for the
construction of sentences like P7 and P9 by replacing our operator treatment of “believes”
with a clause that treats “believes” as expressing a relation between agents and the semantic
values for sentences. I will not carry out this exercise here.
The fact that “what Caroline believes is true” does not implicate its speaker
in disdain for people from Wisconsin has important consequences for the fol-
lowing argument:
P6 What Caroline believes is true.

P7 What Caroline believes is that Mark is a cheesehead.
C4 Mark is a cheesehead.
This argument is not only transparently valid but also clearly has the property
that accepting the premises commits you, on pain of giving one of them up, to
accepting the conclusion. But accepting the conclusion requires having disdain
for people from Wisconsin. What ensures that someone who accepts the prem-
ises will have or be committed to having such disdain? I have just been arguing
that “what Caroline believes is true” does not implicate the speaker in such
disdain. So it does not come from accepting the first premise. But fortunately,
our semantic clause for “believes” tells us that “Caroline believes that Mark is
a cheesehead” does implicate the speaker in such disdain. So sentences with
secondary sets validate the inference schema “what Caroline believes is true”;
“what Caroline believes is that P”: “P,” and they do so precisely because second-
ary sets project through attitude verbs.
If this is right, then it should make us worried about the idea that moral
words are associated with primary sets rather than secondary sets. For the fol-
lowing argument, like the previous one, is not only transparently valid but
clearly has the property that accepting the premises commits you, on pain of
giving one of them up, to accepting the conclusion:

P9 What Caroline believes is that stealing is wrong.
C5 Stealing is wrong.
But as we have seen, you can accept a sentence like P8 without having any
particular attitude and in particular without disapproving of things that fail to
maximize happiness. And on the assumption that “stealing is wrong” has a non-
empty primary set rather than a nonempty secondary set, it follows that you can
also accept P9 without having any particular attitude and in particular without
disapproving of things that fail to maximize happiness. But to accept C5 you
need to have some further attitude. On our second-pass gloss at a primary-set
hybrid view of “wrong,” it is that you must disapprove of things that fail to
maximize happiness. And so you are committed to accepting C5 by accepting
P8 and P9 only if accepting them commits you to this state of disapproval. But
now we have no explanation—or at least none on a par for how things worked
for the productive validities of classical logic or for the argument from P6 and
P7 to C4—of why this would be so. This looks like a problem for the view that
“wrong” contributes to the primary set rather than to the secondary set.
You might be suspicious that something has gone wrong in the setup of this
problem. For all along in this chapter I have worked with a working example of
a word that I have been suggesting does contribute to the primary rather than
the secondary set: “but.” So should not the foregoing considerations lead us to
predict that the following argument should not have the property that accept-
ing the premises commits you, on pain of giving one of them up, to accepting
its conclusion?

P11 What Caroline believes is that Mark is from Wisconsin but can drive
in traffic.
C6 Mark is from Wisconsin but can drive in traffic.
If my foregoing remarks are on the right track, then you can accept P10 without
believing that there is a contrast between being from Wisconsin and being able to
drive in traffic. And if “but” really does contribute to the primary but not the sec-
ondary set, as I have argued, then you can accept P11 without believing that there
is a contrast between being from Wisconsin and being able to drive in traffic. So
should not there be a puzzle about how accepting P10 and P11 commits you to
accepting C7, which does require you to believe that there is a contrast between
being from Wisconsin and being able to drive in traffic? But this argument does
seem to commit someone who accepts its premises to accepting its conclusion.
There is an important difference, however, between the case of “but” and the
case of “wrong,” at least as the hybrid theorist conceives of it. And that is because
the primary set for “but” includes a belief. So when you accept P11, you believe
not only that Caroline believes that Mark is from Wisconsin and can drive in
traffic but also that Caroline believes that there is a contrast between being from
Wisconsin and being able to drive in traffic. So if you also accept P10 and think
that what Caroline believes is true, then you are committed to thinking that it
is true that there is a contrast between being from Wisconsin and being able to
drive in traffic and hence that there is such a contrast. So you really are commit-
ted to the primary set for C7 in virtue of accepting both P10 and P11—and the
reason you are so committed is that the primary set of P11 is just another belief.
In contrast, on the hybrid view the primary set for “wrong” is something other
than another run-of-the-mill belief. For if it were just another run-of-the-mill
descriptive belief, then moral belief would not turn out, after all, to be something
over and above ordinary, run-of-the-mill descriptive belief. On this view both
the argument from P6 and P7 to C4 and the argument from P10 and P11 to C6
commit someone who accepts their premises to accepting their conclusion, but
they do so for different reasons—and neither of these reasons extends to the
argument from P8 and P9 to C5 on the hybrid view under consideration. So the
case of “but” should not make us optimistic; on the contrary, it strongly suggests
that the only things that can figure in primary sets are further beliefs.
S E I Z I N G B O TH HO R N S?
In the fourth part I argued that to be able to say that moral belief is something
over and above ordinary descriptive belief the hybrid theorist must say that
moral words contribute to primary sets. But in the fifth part I argued that to
make sense of the way simple arguments involving “true” preserve commit-
ment the hybrid theorist must not say that moral words contribute only to pri-
mary sets. But we can now put these two conclusions together to argue that an
adequate hybrid theory should say that moral words must contribute to both
primary and secondary sets.
The argument is simple. By the considerations from the fourth part we can
say that there is something distinctive about moral thought only if the primary
set for moral sentences such as “stealing is wrong” is nonempty. This is what
allows us, speaking the language, to coherently say such things as that it is
possible to have any set of nonmoral beliefs without believing that stealing is
wrong. So without loss of generality, let us call the contents of the primary set
for “stealing is wrong” A. It follows that to be committed to accepting the con-
clusion of the following argument you must be committed to having attitude A:

P9 What Caroline believes is that stealing is wrong.
C5 Stealing is wrong.
But P8 does not have A in either its primary or its secondary set. So absent
some different, nonhybrid style explanation of how accepting the premises com-
mits you to having A, the only way for accepting the premises to guarantee that
you have A is if A is in either the primary set or the secondary set for P9. But by
our semantic clause for “believes,” this is true only if A is in the secondary set
for “stealing is wrong.” Consequently, “stealing is wrong” must have a nonempty
primary set, and everything in its primary set must also be in its secondary set.
It could certainly be that moral words work like this. But I find the argument
that leads to this conclusion troubling. For it strikes me that it is no coincidence
that arguments of the pattern of P6 and P7 to C4, P8 and P9 to C5, and P10
and P11 to C6 all have the property that accepting the premises commits you to
accepting their conclusions. I suspect that we will find that all arguments with
this form will have this “inference-licensing” property. And that means that one
of two things must be true. Either primary sets never contain attitudes other
than ordinary descriptive beliefs, or else when they do their nonbelief mem-
bers must always also be in the secondary set. But I find the latter of these two
possibilities very hard to credit. We know from the example of “but” that there
are some words that contribute to the primary set but not the secondary set. So
what possible mechanism could prevent words from doing this when what they
contribute to the primary set is something other than a belief?
In contrast, the hypothesis that primary sets only include beliefs is much
more plausible. On this view there is no need to explain what ensures that non-
belief members of primary sets also need to be members of secondary sets,
even though in general a state of mind can be in a primary set without being in
a secondary set. In fact on this view it is plausible to treat the contents of pri-
mary sets as propositions rather than as mental states and to alter the clause for
“believes” so that it requires belief in each of the members of the primary set as
well as in the core content.
WRA P P I N G U P
In this chapter I have shown that far from being difficult to imagine how moral
words could obey the principle that I have called the Big Hypothesis, it is easy to
describe systems on which this is so and indeed that there are plausible models
for this in natural language. However, I have tried to put pressure on whether
this is the right picture for moral language by looking in greater detail at how
to think about how the word “true” works. This pressure is not perfectly sharp.
It can be resisted by moving to hybrid views like Copp’s realist expressivism, on
which the advantages over corresponding cognitivist theories are even subtler.
It can be resisted by insisting that moral words really do contribute to both pri-
mary and secondary sets. And for all that I have said here, it may be resistible by
adopting yet a different view of how “true” interacts with primary and second-
ary sets—one that gives up the idea that “core” content counts as characterizing
“truth conditions” on any natural sense. Still, despite these caveats I do think
that it should make us cautious about drawing quick conclusions about how
much help the hybrid maneuver may be to realist naturalism in metaethics.22
22. Special thanks to David Copp, Daniel Boisvert, Mike Ridge, Guy Fletcher, Julia Staffel,
Ryan Hay, Caleb Perl, Eric Brown, and Jake Ross.
References
Bach, Kent. 1999. “The Myth of Conventional Implicature.” Linguistics and Philosophy
22 (4): 327–366.
Analysis 60 (3): 268–279.
Boisvert, Daniel. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly 89
(2): 169–203.
Copp, David. 2001. “Realist Expressivism: A Neglected Option for Moral Realism.”
Social Philosophy and Policy 18 (2): 1–43.
Dreier, James. 1990. “Internalism and Speaker Relativism.” Ethics 100 (1): 6–26.
Foot, Philippa. 1978a. “Approval and Disapproval.” In Virtues and Vices, 189–208.
Berkeley: University of California Press.
Grice, H. P. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University
Press.
Language.” European Journal of Philosophy 21 (3): 450–474.
Potts, Christopher. 2005. The Logic of Conventional Implicatures. New York: Oxford
University Press.
Ridge, Michael. 2006. “Ecumenical Expressivism: Finessing Frege.” Ethics 116
(2): 302–336.
———. 2007. “Ecumenical Expressivism: The Best of Both Worlds?” In Oxford Studies in
———. 2009. “The Truth in Ecumenical Expressivism.” In Reasons for Action, ed. David
Sobel and Stephen Wall, 219–242. Cambridge: Cambridge University Press.
Schroeder, Mark. 2009. “Hybrid Expressivism: Virtues and Vices.” Ethics 119
(2): 257–309.
———. 2010. Noncognitivism in Ethics. New York: Routledge.
———. 2013. “Tempered Expressivism.” In Oxford Studies in Metaethics, vol. 8, ed. Russ
Shafer-Landau, 283–314. Oxford: Oxford University Press.
Strandberg, Caj. 2011. “The Pragmatics of Moral Motivation.” Journal of Ethics 15
(4): 341–369.
Tresan, Jon. 2006. “De Dicto Internalist Cognitivism.” Noȗs 40 (1): 143–165.
———. 2009. “Metaethical Internalism: Another Neglected Distinction.” Journal of
Ethics 13 (1): 51–72.
INDEX
Absolutist semantics, 124 Bach, Kent, 55n10, 201–202

Acceptance-inference-licensing thesis, 65, Barker, Stephen, xiv, 89, 166, 174, 199
66, 68 Bar-On, Dorit, xiv–xv, 90, 90n16, 192,
Actual preference, 146n29 216n12, 223, 233
Advice, 132–133, 135, 138 Belief-inference-licensing thesis, 65–66,
A-expression (action 67, 68, 71
expression), 162n24, 225, 231–235, Big Hypothesis, x, 52, 61, 69–71, 72, 292
243–244 Björnsson, Gunnar, 125
Amoralist case, 91, 92, 107, 216, 219 Blackburn, Simon, 23, 38n22, 41, 237
Antirealism, 96, 100, 120, 230, 235–242, Boisvert, Daniel R., xii, 22, 75, 80, 80n9,
244 89, 97, 98, 154, 155, 155n13, 224n4
Antisubjectivism, 213 Brand-name genericization, 102
Apparent de jure sameness, 249n3, Brandom, Robert, 207–208
253n9, 255 Brink, David, 107
Appiah, K. Anthony, 262n18 Burges, Tyler, 264n19
Approval, 87–88, 163
Aptness-conditional semantics, 24n6 Calculability, 174, 179, 190–191
Assertability semantics, 158 Cancelability, 174, 179, 186–190
Assertion, 32, 57, 205–211 Canonical causes, 218
Assertion type, 209 Carnap, Rudolf, 223n1
Atomic sentences, 33, 34 Categorical property, 263
Attitude Chains of reasoning, 32n17
attribution sentences, 40–41 Chrisman, Matthew, xiv–xv, 90, 90n16,
attribution verbs, 45n28, 274, 277, 192, 216n12, 223
278, 288 Cognitive expressivism, 200
cancelability of expressions Cognitivism, x, xv, 248–272
of, 187–188 competence, 264–267
conditional content and, 137–140 conceptual structure, 249–252
disagreement and, 167 internal link to motivation, 259–262
neutral content, 227, 227n10 matching vs. binding, 253–259
Aversion, 55, 59, 62, 67, 68, 69–70 prediction, 267–269
Avowals, 234 reference fixing, 261–264
Ayer, A. J., 23–24, 234–235 Cognitivist-realist approach, 161
296 Index
Color predicates, 214 Conversational implicature, 175–179

Competence, 239, 257n15, 260, 261, calculability of, 179
264–267 cancelability of, 179
Concept identity, 253–254, 254n10, 255, context requirements, 192
256–257, 259 Grice’s characterization of, 193
Conceptual competence, 257n15 indeterminacy of, 179
Conditional declaratives, 28n10, nondetachability of, 179
137–138, 140–145 Particular Conversational
Conditional imperatives, 27n10, 35n20 Implicature, 177, 181n14
Congruence of concept identity, 256, pragmatics of, 200
265–266 reinforceability of, 179, 181
Conjunctive antecedents, 24, 44, 45 Conversational success, 31, 31n15,
Connectedness of concept identity, 256, 36, 131
265, 265n21 Cooperative principle, 132, 133, 176, 190
Connectives, 278 Copp, David, xii, 51, 75, 79n7, 90, 97, 98,
Consequents of conditionals, 24 174, 184, 185, 188, 205n4, 239n31,
Conspicuousness, 117 274n2, 282–283, 284n15, 292
Content-individuation, 95, 104–106 Counterfactuals, 140n20
Content necessity, 104–106
Content relation, 82 Davidson, Donald, 38
Content sufficiency, 104–106 Davis, Wayne A., 154n10, 159–160, 161,
Contextual disambiguation, 101n5 163, 176
Contextualism, 100, 168, 168n32 Declaratives
Contribution requirement, 26–27, 37, 38 conditional, 28n10, 137–138, 140–145
Conventional implicature, xiv, 199–222 expressivism and, 31, 32
assertion and, 205–211 moral, 29, 41, 44, 48
attitude expression and, 184 neo-expressivism and, 225, 227
belief states and, 219–221 De dicto internalism, 285–286
calculability of, 179 Default assumptions, 267
cancelability of, 179 Defensive stance, 208–209
Grice’s conception of, 175–179 Deflationary approach, 161
indeterminacy of, 179 Déjà Vu analyses, 130
intersubjective dimension, 201–205, Deliberation, 135
211–215 Demythologizing antirealists, 97
mental states defended and Dennett, Daniel, 13
expressed, 210 Dependence, 265
motivational externalism Descriptive meaning, 155, 162–165
and, 215–219 Descriptive moral semantics, 97–98
multipropositional sentences, 201–204 Descriptive relativism, 119–120, 119n30
nondetachability of, 179 Descriptivistic analyses of moral
pragmatic presupposition, 205 belief, 52
pure expressivism and, 210–211 Design stance, 13
reinforceability of, 179, 181 Desire, 82–86
subjective probability and, 216–219 Diachronic hybrid moral realism
type-generality and, 89 (DHR), xiii, 95–123
Conventional nonverbal expressions, 231 challenges for, 100–104
Index 297
content-individuation and, 104–106 ERT (end-relational semantics), 126,

functionalism and, 106–113 145–147
humanly special properties, 113–120 Evaluative beliefs, 221
proposal, 97–104 Evaluative meaning, 150–151, 163, 164,
Direct assertives, 32, 44 165, 167–168
Direct directives, 32 Evans, Gareth, 264n19
Direct illocutionary acts, 28, 29–31, Exclamatives, 25, 31, 33, 41
45 Expressive acts, 233
Direct illocutionary force semantic Expressive adjectives, 44, 45
theory, 23–24 Expressive-assertivism, 42–48, 89
Direct interest, 132–134 “Expressive-Assertivism” (Boisvert), 80,
Disagreement, xiii, 124–148 224n4
attitudes with conditional Expressive verbal utterances, 231
content, 137–140 Expressivism, 22–50. See also Pure
conditionals with attitudinal expressivism
content, 140–145 cognitive, 200
information relativity and, 125–129, ecumenical, 6, 12, 89
132–137 Frege-Geach problem for, 26–48
pragmatics of, 129–132, 145–147 meaning and, 153–159
Disapproval, 87–88, 87n15, 155, 163 minimal assertability
Dis-endorsement, 59–60 expressivism, 76n1
Disjunctions, 24, 275 moral, 51
Disjunctive theory, 143 neo-expressivism. See
Dispositional attitude, 88 Neo-expressivism
Doubly hypothetical preferences, 143 noncognitivist, 51
Dowell, Janice, 127n5 protoexpressivism, 242
Dreier, James, 78, 92, 107 quasi-expressivism, 125, 145–146
realist-expressivism. See
Ecumenical cognitivism, 15–20 Realist-expressivism
Ecumenical expressivism, 6, 12, 89 specification problem for, 238n28
“Ecumenical Expressivism: Finessing success-conditional semantics (SCS)
Frege” (Ridge), 80 and, xii, 24–25, 24n6, 38–42
“Ecumenical Expressivism” Expressivist cognitivists, 168n30
(Ridge), 224n4 Extended descriptions, 54n7
Eliminativism, 262n18 Extensionality principle, 80n10
Endorsement, 59–60 Externalism, 59, 91, 215–219
End-relational semantics (ERT), 126,
145–147 Figurative use of language, 102
End-relational theory, 125 Finlay, Stephen, xiii, 90, 124, 174, 184,
Epistemic asymmetry, 233, 234 187, 191n32, 216n12
Epistemic dependence Fletcher, Guy, ix, xiv, 173
assumption, 111n16, 214n9 Foot, Philippa, 107–108
Epithets. See Pejoratives; Slurs Frege, Gottlob, 53, 79n6
Eriksson, John, xiii–xiv, 149 Frege-Geach problem, x, 26–49, 149, 215
Error-theoretical normative Frugality, 95, 96
theories, 117n26 Fulfillment-conditional semantics, 24n6
298 Index
“Fully coherent anorexic” example, 11 Humean intuition, 78, 79n6, 81, 244–245
Functionalism, 95–96, 100, 101n4, 104 “Hybrid Expressivism: Virtues and Vices”
Functional normativity, 16 (Schroeder), 277
Fundamental disagreements, 124 Hybrid semantics, xv, 273–293
Big Hypothesis and, 277–280
Geach, P. T., 28, 79n6 cognitivism vs., 282–286
Generality principle, 80, 80n9, 82–86 logic and inference in, 280–282
Generalized conversational implicature projection patterns, 274–276
(GCI) truth in, 286–291
calculability of, 191 Hypothetical Imperative, 10, 10n7
determinacy of, 180 Hypothetical preferences, 140, 142
ecumenical cognitivism and, 17,
18, 20 Ideationalist conception of
Grice’s conception of, 177, 177n9 meaning, 223–224, 224n5
moral attitude expression and, 174 Identity, 254n10. See also Concept
nondetachability of, 182–183 identity
problems for, 192–196 Illocutionary acts, 30–31, 206. See also
Gert, Bernard, 12n8 Direct illocutionary acts
Gert, Joshua, 12n8 Imperative logic, 27n10
Gibbard, Allan, 11, 23, 38n22, 41, 52, Imperatives
56, 71, 137, 139, 140n19, 154, 158, conditional, 27n10, 35n20
159n19, 164, 251, 252n7 expressivism and, 25, 31, 32, 41
Global nonfactualism, 228n11 Hypothetical Imperative, 10, 10n7
Glucksberg, Sam, 102n7 Impermissibility, 9
Gradable property, 263 Implicative acts, 205
Grice, H. Paul, x, 17, 132, 175–179, 180, Implicative ecumenical cognitivism, 3–4,
193, 273 19–20
Guaranteed Conative State, 62, 63 Implicative type, 209
Guaranteed Failure option, 62, 63 Implicature, xiv, 173–198. See also
Guilt, 104 Conventional implicature;
Conversational implicature
Hare, R. M., 100n3, 108, 149, 150–151, calculability, 174, 179, 190–191
151n3, 223n1 cancelability, 174, 179, 186–190
Haslanger, Sally, 262n18 criteria for, 174
Hay, Ryan, xii–xiii, 19n13, 54n8, 75, 186, Gricean, 175–179
277n7, 279, 279n8 indeterminacy, 174, 179, 180–181
Highsmith, Patricia, 9 nondetachability, 174, 179, 182–186
Hill, Thomas, 10 normative, 16–17
Hindsight Agreement, 136–137, 138 quality of, 176
Holistic rationalizing quantity of, 176
interpretation, 261–264 reinforceability, 174, 179, 181–182
Honorifics, 45 relevance of, 176, 177
Humanly special wrongish property type-generality and, 89
(HSW), 115–120 Impossibility of joint success, 34–35,
Hume, David, 3, 77, 78n4 47n29
Index 299
Inconsistent Belief, 125–129, 131, 135, Linguistic expressives, 239

140, 145–146 Logical connectives, 33, 75, 252,
Inconsistent Preference, 134, 135, 280–282
138–139 Logically contradictory sentences, 36
Inconspicuousness, 116–117, 119 Logical Preservation requirement, 26–
Indeterminacy, 120n31, 174, 179, 28, 34, 37, 39–40, 46–47
180–181
Indirect illocutionary acts, 30 MacFarlane, John, 131
Indirect interest, 133, 134–137 Mallon, Ron, 262n18
Inference-licensing property, x, 52, Manner of implicature, 176
60–68, 292 Meaning, xiii–xiv, 149–170
Information relativity descriptive, 155, 162–165
in direct interest contexts, 132–134 evaluative, 150–151, 163, 164, 165,
in indirect interest contexts, 134–137 167–168
normative disagreement and, 125– expressivism and, 153–159
129, 132–137 Hare on, 150–151
Instrumentalist analysis, 13, 143 hybrid expressivism and, 162–165
Intentional stance, 13 ideationalist conception of, 223–224
Internalism. See also Moral judgment logic and, 165–169
internalism moral inference, 151–153
cognitivist approach, 259–262 moral language and, 165–169
de dicto internalism, 285–286 template for, 159–162
motivational, 78, 90, 90n16, 98, 99, truth-conditional theory of, 22–23n2,
107, 268n24 23
neo-expressivism and, 234, 235–242 Mental filing system metaphor, 250,
Interrogatives, 31, 32 250n4, 255n11
Intersubjective dimension of Mentalist semantic theories, 23
implicature, 201–205, 211–215 Metaethical naturalism, 55
Intuitive sense of disagreement, 128 Metamoralizing, 99, 101, 103, 104
Inverted commas proposal, 184 Metanormative theory, defined, ix
Irrationality, 3–4, 11–12, 14, 18 Metonymic target, 102
Minimal assertability expressivism, 76n1
Judgment motivational internalism, 56, Monosemous realism, 114
60, 283, 283n13 Moral declaratives, 29, 41, 44, 48
Justification, 209 Moral duty, 104
Moral expressivism, 51
Kant, Immanuel, 7–8, 10 Moral inference, 151–153
Kitcher, Philip, 262n18 Moral judgment externalism, 59
Kolodny, Niko, 16, 131 Moral judgment internalism, 56, 59, 69,
70, 195, 283n13
The Language of Morals (Hare), 149, Moral language, 76
150 Moral naturalism, 59
Lenman, James, 107 Moral nihilists, 84
Lepore, Ernest, 24n6 Moral predicates, 54, 54n8, 79
Less Informed analyses, 129–132 The Moral Problem (Smith), 77
300 Index
Moral realism, 60 Parfit, Derek, 254n10

Moral utterances, 175 Particular Conversational
Motivation Implicature, 177, 181n14
attitudes, 244 Pejoratives, x, 19n13, 53–54, 53n5, 70,
cognitivist approach, 249, 259–262 155n13, 239n31, 277n7. See also
neo-expressivism and, 244–245 Slurs
noncognitive type-generality Permissibility, 9
and, 86–93 Personal identity, 254n10
prediction and, 268 Physical stance, 13
Motivational asymmetry, 234 Pinillos, Ángel, 249n3
Motivational externalism, 215–219 Placement problems, 237n27
Motivational internalism, 78, 90, 90n16, Polysemy, 99, 99n2, 100, 113
98, 99, 107, 268n24 Postmortem analyses, 129–132, 135,
Mythologizing realists, 97 138, 143
Potts, Christopher, 178
Natural expressions, 231 Practicality, 95, 98
Negations, 143, 144, 275 Practical rationality, 3–4, 5, 67
Neo-expressivism, xiv–xv, 223–247 Pragmatic presupposition, 199–222
a-expression vs. Pragmatics, 125, 129–132, 145–147
s-expression, 231–235 Predicate logic, 281–282
antirealism and, 235–242 Predicates, 214, 278–279
internalism and, 235–242 Prediction, 261, 267–269
semantic conservatism, 226–231 Prescriptivism, 100
Neo-Fregean thought, 257n14 Preservation of success conditions, 37
Neutral counterpart proposition, 79 Price, Huw, 228n11, 237n27
News sensitivity, 127 Prinz, Jesse, 251, 251n6
Noncognitive attitude, x, 80 Proattitude, 17
Noncognitive type-generality, xii–xiii, Probabilities, 202, 210, 216–219
75–94 Production correctness, 202
generality and desire, 82–86 Projectivism, 100
motivation and embedding, 76–82, Propositionality, 82, 220, 230
86–93 Propositions, 225, 226n8
Noncognitivist expressivism, 51 Propriety conditions, 240–241, 242n36
Noncoincidental ambiguity, 99n2 Protoexpressivism, 242
Nondeclaratives, 23, 24, 29, 48 Pseudoadvice, 135
Nondefensive stance, 209 Psychological continuity, 244
Nondetachability, 174, 179, 182–186 Pure expressivism
Nonfactualism, 228n11 conventional implicature
Nonidentity, 101 and, 210–211
Normative implicature, 16–17 hybrid expressivism vs., 221–222
Normative-theoretic success-conditional semantics
open-mindedness, 117 for, 38–42
Objectivity, 95, 96, 98–99 Quality of implicature, 176

Open Question challenge, 59, 60 Quantifiers, 24
Optative sentences, 31, 33 Quantity of implicature, 176
Index 301
Quasi-disagreement, 134n15 153, 154, 155, 157, 157n17, 224n4,

Quasi-expressivism, 125, 145–146 288n20
Robinson, Denis, 134n15
Racial concepts, 262, 262n18 Robust Inconsistent Hypothetical
Racial slurs. See Slurs Preference (RIHP), 142–143
Rationality, xi–xii, 3–21 Role-individuation, 107
capacity sense, 3, 4–9, 14 Role necessity, 107
ecumenical cognitivism, 15–19 Role-sensitive interpretations, 103n8,
function of, 14 107–108, 110, 112
normative disagreement and, 143 Role sufficiency, 107
normative judgment and, 6
success sense, 3, 9–15 Sadock, Jerrold M., 181
theoretical, 11 Same-content account, 158
Rational pressure, 66, 67 Same State Proposal, 63–64, 68, 71, 72
Realist-expressivism, xii, 51–74 Saul, J. M., 193n38
Big Hypothesis and, 69–71 Scare quotes, 184
inference-licensing property Schroeder, Mark, xv, 28, 52, 55, 60–64,
and, 60–68 68–71, 75, 76n1, 80, 81, 158–159,
motivation and, 90 186, 273, 277
promise of, 56–60, 71–72 Schroeter, François, xv, 248
proposal for, 52–56 Schroeter, Laura, xv, 248
Schroeder's argument on, 60–64 SCS. See Success-conditional semantics
Realist naturalism, 292 Searle, John, 79n6
Redundancy, 182n16 Sellars, Wilfrid, 231
Reference determination, 253–254, Semantics
257, 259 absolutist, 124
Reference fixing, 261–264 aptness-conditional, 24n6
Reference shifts, 267n23 assertability, 158
Reinforceability, 174, 179, 181–182 conservatism in, 226–231, 236, 241
Relational properties, 124, 131–132 continuity in, 233
Relativism descriptive moral, 97–98
descriptive, 119–120, 119n30 fulfillment-conditional, 24n6
implicature and, 213–214 mentalist semantic theories, 23
information relativity, 125–129, monism, 105
132–137 presuppositions, 201
intersubjective dimension and, 204n3 semantic minimalism, 228n11
meaning and, 165 value, 253n8
nondeclaratives and, 25 Sentential moods, 31–38
success-conditional semantics and, 41 S-expression (semantic
truth, 128, 129 expression), 162n24, 225, 231–235
Relevance of implicature, 176, 177 Shame, 104
Requirement notion of rational agent Sias, James, xiv–xv, 223
behavior, 9 Simple Pragmatic Story (SPS), 195
Richard, Mark, 131 Simplicature, 55n10
Ridge, Michael, ix, xi–xii, 3, 58n13, Slurs, x, 19, 45, 221–222, 239, 239n31,
62n14, 65n17, 75, 79, 79n7, 80, 89, 274, 276, 277. See also Pejoratives
302 Index
SMI (strict motivational internalist) The Talented Mr. Ripley (Highsmith), 9

constraint, 260, 264 Temporal relation of speaker, 131
Smith, Michael, 6n4, 7, 77 Testimony-based utterances, 218
Soames, Scott, 161 Theoretical rationality, 11
Social deference, 257n15 Thomson, Judith Jarvis, 127n3
Social intelligence, 103 Thought-tokens, 161, 232n18
Speaker orientation, 131, 185 Token-based approach, 257,
Specification problem, 238n28 257n15, 258
Speech-act type attitudes, 220 Token-for-type metonymy, 102–103,
SPS (Simple Pragmatic Story), 195 102n7
Stalnaker, Robert, 205 Tradition-based approach, 257–258
Stanhope, Philip Dormer, 4 Treatise of Human Nature (Hume), 77
States of mind, 231 Tresan, Jon, xiii, 95, 285, 287n18
Stevenson, Charles L., 134, 142n21, 157, Truth-conditional connectives, 274,
157n17, 223n1 287, 288
Stocker, Michael, 107 Truth-conditional semantics, xi, 23, 161,
Strandberg, Caj, 174, 190, 191, 192n33, 226–231, 235, 273
192n36, 196 Truth-conditional theory of
Strict motivational internalist (SMI) meaning, 22–23n2, 23
constraint, 260, 264 Truth relativism, 128, 129
Subjective probabilities, 202, 210, Truth theory, 22, 286–291
216–219 Twin-wrong judgments, 112–113,
Subjectivism, 166, 168, 168n32 112n16, 121
Success-conditional semantics (SCS) Two Gun Roulette case, 126
direct illocutionary acts and, 29–31
expressivism and, xii, 24–25, 24n6, Utilitarianism, 117n25, 118n28
38–42
Van Roojen, Mark, 28
Frege-Geach problem for, 26–49
Vranas, Peter, 27n10, 35n20
hybrid expressivism and, 42–48
sentential moods and, 31–38 Wise Choices, Apt Feelings
Svavarsdóttir, Sigrún, 195 (Gibbard), 52
Synchronic hybrid realism, 99,
105, 106 Zack, Naomi, 262n18

(Oxford Moral Theory) Guy Fletcher, Michael Ridge - Having It Both Ways - Hybrid Theories and Modern Metaethics (2014, Oxford University Press)

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

(Oxford Moral Theory) Guy Fletcher, Michael Ridge - Having It Both Ways - Hybrid Theories and Modern Metaethics (2014, Oxford University Press)

Încărcat de

Drepturi de autor:

Formate disponibile

Having It Both Ways

OXFORD MORAL THEORY

Against Absolute Goodness

The Lewd, the Rude and the Nasty

The Virtues of Happiness

Having It Both Ways

EDITED BY GUY FLETCHER

PART ONE Optimism about Hybrid Theories

PART TWO Pessimism about Hybrid Theories

Stephen Barker is Associate Professor and Reader in Philosophy at the

(i) Moral (/aesthetic/prudential/epistemological) claims express both

The new millennium has seen a blossoming of interest in hybrid theories

Hybrid theories themselves can come in both cognitivist and noncognitiv-

cognitivism because of the role of a certain kind of generalized conversa-

(2) Daniel R. Boisvert, “Expressivism, Nondeclaratives, and Success-

(4) Ryan Hay, “Attitudinal Requirements for Moral Thought and

(6) Stephen Finlay, “The Pragmatics of Normative Disagreement.” In

(7) John Eriksson, “Hybrid Expressivism: How to Think about Meaning.”

(8) Guy Fletcher, “Moral Utterances, Attitude Expression, and Implicature.”

(11) Laura Schroeter and François Schroeter, “Why Go Hybrid?

(12) Mark Schroeder, “Truth in Hybrid Semantics.” In previous work

Optimism about Hybrid

How to Insult a Philosopher

Let there be no crude suspense: one of the best ways to insult a philosopher

1. See Ridge 2006, forthcoming.

We must not suppose that, because a man is a rational animal, he will,

(1) Rational agents as such can set and abandon ends.

3. See especially Ridge2014 but also Ridge 2006.

This is no easy task, particularly for antireductive and nonnaturalist realists.

Consider a more down-to-earth illustration of the central point. Tom Ripley

Recall my opening methodological maxim: “rational” in the capacity sense and

A rational agent’s action is rationally permissible just in case it is compat-

The idea that these notions (permissibility, impermissibility, and require-

7. Interestingly, the hypothetical imperative is itself a categorical imperative—it applies to us

More importantly, in reaching a reflective equilibrium, sometimes some intu-

9. Compare Davidson 1980.

10. I defend this at length in Ridge 2014.

H. Paul Grice (1975) called a “generalized conversational implicature” or GCI.12

In this chapter I have defended a form of what I call implicative ecumenical

on offer is broadly instrumentalist. Unlike its more Kantian rivals, it preserves

Smith, M. 1994. The Moral Problem. Oxford: Blackwell.

Theories of meaning explain how we understand the complex expressions of a lan-

(3) Kaya serttir means in Turkish that rocks are hard.

structured composites) and their semantically significant subsentential components (e.g.,

illocutionary force” semantic theory, which explains our understanding of

(4) “If you ate all of your dinner, terrific!”

Furthermore, nondeclaratives are embeddable into a wide variety of linguis-

4. See Ayer 1936, chap. 6.

E X P R E S S I V I S M AND THE FR E G E - G E ACH P ROBLEM

The Frege-Geach problem is the perennial problem for expressivism.9 What is

(6) The light is green.

Sentences (7)–(11) embed (6). We understand (7)–(11) by understanding their

is the Contribution requirement. Contribution: an adequate meaning theory for

(12) The light is green.

(16) If the light is green, go.

(19) Insulting others is wrong.

us what they contribute to our understanding of more complex sentences into

(22) Thank you.

Still, if expressivism holds that moral declaratives work in most important

Direct Illocution, Sentential Mood, and SCS

such as “Thank you!” Illocutionary acts can be direct or indirect. An indirect

A direct illocutionary act, then, can be said to be “conversationally” suc-

SCS for the Sentential Moods