Documente Academic
Documente Profesional
Documente Cultură
Drawing Morals
Essays in Ethical Theory
Thomas Hurka
Commonsense Consequentialism
Wherein Morality Meets Rationality
Douglas W. Portmore
In Praise of Desire
Nomy Arpaly and Timothy Schroeder
Confusion of Tongues
A Theory of Normative Language
Stephen Finlay
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
Index 295
CONTRIBUTORS
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:
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
(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.
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.
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
(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.
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
Reason is, and ought only to be, the slave of the passions.
—David Hume, A Treatise of Human Nature
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
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:
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
6 H aving I t B oth Ways
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.
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.
How to Insult a Philosopher 9
R AT I ON A L I T Y AS A SUCCE SS NO TI ON
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.
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.
12 H aving I t B oth Ways
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.
How to Insult a Philosopher 13
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,
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
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.
16 H aving I t B oth Ways
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
How to Insult a Philosopher 17
12. For a similar view in the moral case, see Strandberg 2012.
18 H aving I t B oth Ways
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.
How to Insult a Philosopher 19
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
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.
20 H aving I t B oth Ways
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.
———. 2014. Impassioned Belief. Oxford: Oxford University Press.
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.
How to Insult a Philosopher 21
Expressivism, Nondeclaratives,
and Success-Conditional
Semantics
DANIEL R. BOISVERT n
INT R OD U C T I O N
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:
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
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.
26 H aving I t B oth Ways
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.
9. See Geach 1958, 1960, 1965. See also Frege 1997; Ross 1939, chap. 2.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 27
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).
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
28 H aving I t B oth Ways
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
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.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 29
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
30 H aving I t B oth Ways
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.
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!”
32 H aving I t B oth Ways
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.
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.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 33
Likewise for exclamatives and optatives. Taking “Yes!” as our example, an ade-
quate success-conditional meaning theory would allow permit this derivation:
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.
34 H aving I t B oth Ways
(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.
(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.
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.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 35
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:
(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.
36 H aving I t B oth Ways
(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.
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.
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):
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.
38 H aving I t B oth Ways
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.
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
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.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 39
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:
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
40 H aving I t B oth Ways
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:
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:
Expressivism, Nondeclaratives, and Success-Conditional Semantics 41
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:
But not every instance of this form would preserve success. For example:
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:
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.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 43
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:
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).
44 H aving I t B oth Ways
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”:
(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
Expressivism, Nondeclaratives, and Success-Conditional Semantics 45
(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.
(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.
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.
46 H aving I t B oth Ways
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):
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:
According to the hybrid view that adopts SCS, this argument would be
rendered as:
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:
29. Perhaps another way to understand the inconsistency is in terms of the impossibility of
joint sincerity:
30. I am assuming of course that F-ness is not disapproval of insulting others, that is, I am
assuming that relativism is false.
Expressivism, Nondeclaratives, and Success-Conditional Semantics 49
CONCLUSION
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.
50 H aving I t B oth Ways
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.
52 H aving I t B oth Ways
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
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.
54 H aving I t B oth Ways
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.
Can a Hybrid Theory Have It Both Ways? 55
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).
56 H aving I t B oth Ways
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.
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.
Can a Hybrid Theory Have It Both Ways? 59
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
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
62 H aving I t B oth Ways
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.
Can a Hybrid Theory Have It Both Ways? 63
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).
64 H aving I t B oth Ways
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.
16. I am grateful to Ridge and Bach for helpful discussion (see Bach 2014).
Can a Hybrid Theory Have It Both Ways? 65
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.
66 H aving I t B oth Ways
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.
Can a Hybrid Theory Have It Both Ways? 67
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
68 H aving I t B oth Ways
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
Can a Hybrid Theory Have It Both Ways? 69
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):
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.
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
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.
74 H aving I t B oth Ways
RYA N H AY n
I N T R OD U C TI O N
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
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
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.
78 H aving I t B oth Ways
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.
Attitudinal Requirements for Moral Thought and Language 79
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.”
80 H aving I t B oth Ways
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
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:
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
Attitudinal Requirements for Moral Thought and Language 83
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:
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.
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.
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.
Attitudinal Requirements for Moral Thought and Language 85
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:
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?
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
86 H aving I t B oth Ways
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) 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
Attitudinal Requirements for Moral Thought and Language 87
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.
88 H aving I t B oth Ways
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
90 H aving I t B oth Ways
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).
Attitudinal Requirements for Moral Thought and Language 91
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!”
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.
Attitudinal Requirements for Moral Thought and Language 93
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.
94 H aving I t B oth Ways
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
96 H aving I t B oth Ways
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
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
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
Diachronic Hybrid Moral Realism 99
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
100 H aving I t B oth Ways
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.
Diachronic Hybrid Moral Realism 101
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
102 H aving I t B oth Ways
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:
Diachronic Hybrid Moral Realism 103
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).
104 H aving I t B oth Ways
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).
9. Bering 2011; Chudek and Henrich 2011; Chudek et al. 2013; Kitcher 2006; O’Gormana
et al. 2008.
Diachronic Hybrid Moral Realism 105
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.
106 H aving I t B oth Ways
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.
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).
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
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
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
Diachronic Hybrid Moral Realism 111
so
so,
so,
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
112 H aving I t B oth Ways
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.
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.)
21. I am here relying on the second assumption about WR mentioned at the end of the first
section.
Diachronic Hybrid Moral Realism 115
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).
22. F has to be instantiated, since we are looking for a candidate for the realist property W.
116 H aving I t B oth Ways
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.
Diachronic Hybrid Moral Realism 117
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
118 H aving I t B oth Ways
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.
Diachronic Hybrid Moral Realism 119
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.
120 H aving I t B oth Ways
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.
Diachronic Hybrid Moral Realism 121
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.
122 H aving I t B oth Ways
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
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
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.
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,
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
The Pragmatics of Normative Disagreement 127
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.
128 H aving I t B oth Ways
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.
The Pragmatics of Normative Disagreement 129
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:
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)
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.
130 H aving I t B oth Ways
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.
The Pragmatics of Normative Disagreement 131
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:
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:
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.
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.
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.
134 H aving I t B oth Ways
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
testable predictions. While the theoretical space here is large, we will not need
to look far beyond the simplest, disjunctive version:
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:
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
The Pragmatics of Normative Disagreement 137
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.
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:
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.
The Pragmatics of Normative Disagreement 139
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.
140 H aving I t B oth Ways
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
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.
The Pragmatics of Normative Disagreement 141
directly asserting the suggested information has the same effect. Suppose the
conversation had gone like this:
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
142 H aving I t B oth Ways
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.
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.
The Pragmatics of Normative Disagreement 143
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
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.
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
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.
The Pragmatics of Normative Disagreement 147
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.
Oxford: Oxford University Press.
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.
New York: Oxford University Press.
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.
148 H aving I t B oth Ways
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.”
Philosophical Quarterly 50 (196): 337–352.
Unwin, Nicholas. 2001. “Norms and Negation: A Problem for Gibbard’s Logic.”
Philosophical Quarterly 51 (202): 60–75.
7
Hybrid Expressivism
How to Think about Meaning
JOHN ERIKSSON n
I N T R OD U C TI O N
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.
150 H aving I t B oth Ways
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
MO R A L I N F ER E N CE
(1) Stealing is wrong.
(2) If stealing is wrong, then murder is wrong.
(3) Murder is wrong.
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.
152 H aving I t B oth Ways
(1d): Lying 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,
E X P R E S S I V I SM AN D M E ANI NG
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.
154 H aving I t B oth Ways
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.
Hybrid Expressivism 155
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:
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.
156 H aving I t B oth Ways
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?
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.
Hybrid Expressivism 157
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.
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).
Hybrid Expressivism 159
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?
22. Below I will sometimes drop talk of “expression” and merely say that meaning consists
in states of mind.
Hybrid Expressivism 161
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
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
24. See also Dorit Bar-On and Matthew Chrisman’s (2009) distinction between s-expression
(semantic expression) and a-expression (action expression).
Hybrid Expressivism 163
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.
164 H aving I t B oth Ways
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.
Hybrid Expressivism 165
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.
166 H aving I t B oth Ways
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:
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
168 H aving I t B oth Ways
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.
Hybrid Expressivism 169
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.
Bar-On, D., and M. Chrisman. 2009. “Ethical Neo-Expressivism.” In Oxford Studies in
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.
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.
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.
170 H aving I t B oth Ways
I N T R OD U C TI O 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:
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.
174 H aving I t B oth Ways
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
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
176 H aving I t B oth Ways
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.
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:
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.
Moral Utterances, Attitude Expression, and Implicature 177
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:
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.
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.
Moral Utterances, Attitude Expression, and Implicature 179
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:
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.
180 H aving I t B oth Ways
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:
T E S T 2 : R E I N FO R C E AB I LI TY
Conventional Implicature
“Chelsea has signed that kraut Ballack.”
# “Chelsea has signed that kraut Ballack. I have derogatory attitudes toward
German people.”15
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.
182 H aving I t B oth Ways
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.
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.
Moral Utterances, Attitude Expression, and Implicature 183
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:
This conveys that O has negative desire-like attitudes toward torturing terror sus-
pects. The same attitude expression occurs with
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.
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.
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:
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.
Moral Utterances, Attitude Expression, and Implicature 187
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.)
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.
Moral Utterances, Attitude Expression, and Implicature 189
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:
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.
190 H aving I t B oth Ways
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
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.
Moral Utterances, Attitude Expression, and Implicature 191
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.
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.
192 H aving I t B oth Ways
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:
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.
194 H aving I t B oth Ways
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.
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
Moral Utterances, Attitude Expression, and Implicature 195
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:
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
196 H aving I t B oth Ways
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
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.
Moral Utterances, Attitude Expression, and Implicature 197
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
Metaethics, vol. 4, ed. R. Shafer-Landau, 133–166. Oxford: Oxford University Press.
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.
Copp, D. 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. R. Shafer-Landau, 167–202. Oxford: Oxford University Press.
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/.
Finlay, S. 2004. “The Conversational Practicality of Value Judgement.” Journal of Ethics
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.
Hay, R. 2013. “Hybrid Expressivism and the Analogy between Pejoratives and Moral
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.
198 H aving I t B oth Ways
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: 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:
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
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:
Multipropositional Sentences
(4) Granny, who, by the way, is extremely unlikely to get drunk, is drunk.
202 H aving I t B oth Ways
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
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 203
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:
2. See Barker 2003 and 2004 for discussions of the intersubjective dimension in various
contexts.
204 H aving I t B oth Ways
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:
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.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 205
Pragmatic Presupposition
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.
206 H aving I t B oth Ways
(7) Buggsy knows that even the best philosophical minds can get confused.
(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
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.
(9) Even Granny was drunk. If even Granny was drunk, it must have been
a wild party. Therefore it was a wild party.
208 H aving I t B oth Ways
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.
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:
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
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 209
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.
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.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 211
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:
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
212 H aving I t B oth Ways
7. This is how Daniel Boisvert (2008), Stephen Finlay (2005), and Mark Schroeder (2009)
take the view.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 213
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
214 H aving I t B oth Ways
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
MO T I VAT I ON AL E X TE R N ALI SM
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.
216 H aving I t B oth Ways
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.
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.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 217
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.
218 H aving I t B oth Ways
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:
Canonical Input
Unlikely [O is F]
Testimony state
Noncanonical
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.
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:
220 H aving I t B oth Ways
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:
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.
Pure versus Hybrid Expressivism and the Enigma of Conventional Implicature 221
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:
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
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).
222 H aving I t B oth Ways
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.
Boisvert, Daniel. 2008. “Expressive-Assertivism.” Pacific Philosophical Quarterly
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.
———. 2009. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies in
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
I N T R OD U C TI O N
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.
224 H aving I t B oth Ways
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
• 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
A P L E A F O R C O N SE R VATI SM I N SE M A NTICS
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.
(How) Is Ethical Neo-Expressivism a Hybrid View? 227
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).
228 H aving I t B oth Ways
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.
(How) Is Ethical Neo-Expressivism a Hybrid View? 229
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.
230 H aving I t B oth Ways
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:
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
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
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
234 H aving I t B oth Ways
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
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.)
236 H aving I t B oth Ways
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.)
(How) Is Ethical Neo-Expressivism a Hybrid View? 237
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.
238 H aving I t B oth Ways
28. This problem has been called “the specification problem” for expressivism. See Björnsson
and McPherson forthcoming; Köhler 2013.
(How) Is Ethical Neo-Expressivism a Hybrid View? 239
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.
240 H aving I t B oth Ways
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.
(How) Is Ethical Neo-Expressivism a Hybrid View? 241
35. For further discussion, see Bar-On and Chrisman 2009, 143–150.
242 H aving I t B oth Ways
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.
(How) Is Ethical Neo-Expressivism a Hybrid View? 243
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 ?
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.
244 H aving I t B oth Ways
References
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.
———. 2014. Impassioned Belief. Oxford: Oxford University Press.
Schroeder, M. 2008. Being For: Evaluating the Semantic Program of Expressivism.
Oxford: Oxford University Press.
———. 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
Phenomenological Research 29:506–527.
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
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
C O N C E P T U A L STR UCTUR E
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).
250 H aving I t B oth Ways
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
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
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.
254 H aving I t B oth Ways
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.
256 H aving I t B oth Ways
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:
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
258 H aving I t B oth Ways
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
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.
260 H aving I t B oth Ways
(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.
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.
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.
262 H aving I t B oth Ways
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
264 H aving I t B oth Ways
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
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.
266 H aving I t B oth Ways
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
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.
268 H aving I t B oth Ways
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
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.
270 H aving I t B oth Ways
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.
274 H aving I t B oth Ways
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
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.
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.
B I G H Y P OT H E SI S SE M AN TI CS
7. Hay (2011) argues persuasively that a more promising analogy for moral terms are what
he calls “general pejoratives,” like “jerk” and “asshole.”
278 H aving I t B oth Ways
Connectives:
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.
Predicates:
New Connective:
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:
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.
280 H aving I t B oth Ways
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”:
C1 Mark is a cheesehead.
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:
A D VA N TA GE S O V E R C O G NI TI V I SM
11. I will turn to a different pattern of inference involving the word “true” in the fifth section.
12. Copp 2001, 2009, this volume.
The Truth in Hybrid Semantics 283
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.
284 H aving I t B oth Ways
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.
The Truth in Hybrid Semantics 285
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:
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 Truth in Hybrid Semantics 287
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”:
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:
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
288 H aving I t B oth Ways
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 Truth in Hybrid Semantics 289
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:
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:
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
290 H aving I t B oth Ways
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?
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
The Truth in Hybrid Semantics 291
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:
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
292 H aving I t B oth Ways
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.
The Truth in Hybrid Semantics 293
References
Bach, Kent. 1999. “The Myth of Conventional Implicature.” Linguistics and Philosophy
22 (4): 327–366.
Barker, Stephen. 2000. “Is Value Content a Component of Conventional Implicature?”
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.
———. 2009. “Realist-Expressivism and Conventional Implicature.” In Oxford Studies in
Metaethics, vol. 4, ed. Russ Shafer-Landau, 167–202. Oxford: Oxford University Press.
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.
Hay, Ryan. 2011. “Hybrid Expressivism and the Analogy between Pejoratives and Moral
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
Metaethics, vol. 2, ed. Russ Shafer-Landau, 51–76. Oxford: Oxford University Press.
———. 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
“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