<<

Against Metaethical Descriptivism: The Semantic Problem

Item Type text; Electronic Dissertation

Authors Mitchell, Steven Cole

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 30/09/2021 17:07:01

Link to Item http://hdl.handle.net/10150/202935 1

AGAINST METAETHICAL DESCRIPTIVISM: THE SEMANTIC PROBLEM

by

STEVEN COLE MITCHELL

______

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2011

2

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Steven Cole Mitchell entitled Against Metaethical Descriptivism: The Semantic Problem and recommend that it be accepted as fulfilling the dissertation requirement for the

Degree of Doctor of Philosophy

______Date: August 12, 2011 Mark Timmons

______Date: August 12, 2011 Michael Gill

______Date: August 12, 2011 Terry Horgan

______Date:

______Date:

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: August 12, 2011 Dissertation Director: Mark Timmons 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: Steven Cole Mitchell

4

ACKNOWLEDGEMENTS

My long-suffering dissertation director was Mark Timmons: a man of great learning and parts, endowed with an encouraging spirit and the patience of Job. My other committee members, Michael Gill and Terry Horgan, gave helpful feedback and flattered my ego at pivotal moments. David Owen went out of his way to help me begin work on the dissertation, and Houston Smit and Suzi Dovi helped me overcome personal disaster. Audiences at the University of Arizona and Washington University at St. Louis treated me to a kind and stimulating reception when chapters were presented. I discussed metaethics with virtually all of the Arizona professors and graduate students over the years—Ian Evans deserves special mention for some clutch midwifing. My interest in metaethics was sparked back at the University of Alabama, most especially by Stuart Rachels. And I should thank all the professors who put metaethics papers online: this was a godsend when I lived abroad and was first gripped by my current metaethical intuitions.

Friends sometimes make life marginally tolerable: thanks to everyone from Murfreesboro, Tuscaloosa, Litoměřice, Hradec Králové, and Tucson. Thanks to my family for never giving me a hard about this goofy career path, and for supporting me when things got extra stupid.

5

DEDICATION

To the fans.

6

TABLE OF CONTENTS

ABSTRACT...... 7 CHAPTER 1: INTRODUCTION...... 8 CHAPTER 2: ...... 16 2.1 Schroeder on normative reasons...... 16 2.2 Copp’s society-centered realist-expressivism...... 23 2.3 Critique of Copp...... 32 CHAPTER 3: NONNATURALISM...... 40 3.1 Textbook nonnaturalism...... 41 3.2 Semantic arguments against naturalism...... 44 3.3 Hume and Moore against nonnaturalism...... 50 3.4 Moral Twin Earth against nonnaturalism...... 57 3.5 Nonnormative Conceptual Role Semantics...... 63 3.6 Normative Conceptual Role Semantics...... 68 3.7 Two objections...... 74 3.8 Conclusion...... 83 CHAPTER 4: QUIETISM...... 85 4.1 Introduction...... 85 4.2 Dworkin’s theory...... 89 4.3 Dworkin on error-theory...... 94 4.4 Dworkin on expressivism...... 105 4.5 Dworkin and metaphysical quietism...... 113 CHAPTER 5: ERROR-THEORY...... 119 5.1 The possible of moral properties...... 119 5.2 Preview of the argument...... 123 5.3 Mackie’s error-theory...... 125 5.4 The argument in detail...... 129 5.5 Mackie’s proposed explanation of supervenience...... 135 5.6 Error-theories in other domains...... 139 5.7 Moral properties as impossible?...... 140 CHAPTER 6: FICTIONALISM...... 146 6.1 Joyce’s fictionalism...... 146 6.2 Kalderon: an overview...... 162 6.3. Kalderon: a critique...... 174 CHAPTER 7: BLACKBURN’S SUPERVENIENCE ARGUMENT...... 183 7.1. Preliminary points...... 184 7.2. How the argument works...... 186 7.3. Countering objections with conceptual competence...... 190 7.4. Natural kinds...... 200 7.5. Other targets...... 207 7.6. Conclusion...... 215 WORKS CITED...... 216

7

ABSTRACT

In my dissertation I argue that prominent descriptivist metaethical views face a serious

semantic problem. According to standard descriptivism, moral and discourse

purports to describe some of moral properties and/or : e.g., the term

‘good’ purports to refer to some or cluster of properties. Central to any such

theory, then, is the recognition of certain items of ontology which, should they actually exist, would count as the referents of moral terms and . And since one commonly

accepted feature of moral thought and discourse is a supervenience constraint, descriptivists hold that any ontology suitable for morality would have to supervene upon non-moral ontology. But this lands descriptivists with the task of providing a semantic account capable of relating this ontology to moral terms and concepts. That is, they must explain why it is that certain items of ontology and not others would count as the referents of moral terms and concepts, in a way that is consistent with the supervenience

constraint. I argue that this important explanatory task cannot be carried out. And because

the problem generalizes from metaethics to all normativity, we are left with good reason

to pursue alternatives to descriptivist accounts of normative semantics.

8

CHAPTER 1: INTRODUCTION

In my view, the central problem for metaethics lies in understanding the mysterious of normativity. Morality is where the phenomenon and its mysteries are most visible. Moral obligation provides a clear case of standards to be followed that are invested with objective normative authority—standards whose credentials are perfectly unclear. And when ethics ranges beyond the deontic realm, comparing better and worse ways of life, it may eschew the stern commands of duty but it its ideals are still presented as something to aspire to. Ethics can be taken broadly, encompassing the prudential subdomain of self-interest and the instrumental subdomain of means-end reasoning: again, pursuing what contributes to one’s own personal advantage, and taking suitable means to one’s ends are presented as somehow more appropriate than the alternative. The upshot is that normativity pervades the practical domain that is the natural home of metaethics. Indeed, metaethical interest in normativity arguably extends far further, into the domain of epistemic reasons and linguistic .

One prominent way of understanding the normative mystery involves normative ontology. What gives normative evaluation its distinctive character, on this approach, is its distinctive : properties or relations that somehow point the way to a way of life worth living. This normative ontology finds its way into normative evaluation by way of its descriptive content, with normative terms and concepts purporting to refer to this ontology. This is the descriptivist approach to understanding normativity. 9

The approach is embraced by a wide variety of metaethical theories. Realist

theories form the most prominent example, with naturalists characterizing moral

judgments as beliefs whose content refers to normative ontology belonging to the natural

world, and nonnaturalists locating their favored ontology somewhere beyond the natural

world. But anti-realists also routinely take this approach: error-theorists and fictionalists do not commit their theories to the actual existence of normative ontology, commonly rejecting it as too ‘queer’ and otherworldly to be credible, but they understand normative judgments as purporting to refer to it all the same.

It is my gut conviction that this approach is deeply misguided. Ontology is not the right place to look for a proper understanding of normativity. And thus I find the very of normative ontology—ontology that could somehow tell us the distinctive character of normative evaluation—not just mysterious, not just ‘queer’, but nigh- incomprehensible. But of course, gut convictions are not arguments, and not everyone shares my bafflement. This study, therefore, is an attempt to identify and develop a general form of argument against descriptivist metaethical theories that understand normativity in terms of normative ontology.

The argument I have settled upon is, perhaps disappointingly or perhaps reassuringly, nothing all that new. In short, it challenges descriptivists to provide a semantic account capable of explaining why normative terms and concepts refer to the ontology they do. The challenge gets its bite from a couple of general features of normative evaluation. First, it is extremely flexible, admitting of drastic disagreement and denial: no matter how overwhelmingly plausible it is that x is better than y, it always 10

seems to be at least conceptually coherent to think y is better than x, a case of genuine normative disagreement rather than conceptual versus conceptual incompetence.

And indeed, just as it seems coherent to deny any normative status attributed to any evaluated item, it seems coherent to go nihilist and deny all normative statuses to all items. This flexibility means conceptual are hard to come by in the normative domain, depriving would-be semantic accounts of their lifeblood. But second, there is a conceptual truth that paradoxically ends up only causing more trouble: the conceptual constraint that an item’s normative status supervenes upon its other characteristics. This aggravates the burden in a few ways, as we will see, most prominently by forcing the question of why descriptive concepts would have such a constraint built into them.

The semantic problem of making sense of normativity in terms of the ontology referred to by descriptive concepts received its classic expression in the most well-worn topics of metaethics: Hume’s is-ought gap and Moore’s open question argument (backing his ‘naturalistic fallacy’ accusation). And in contemporary metaethics, it is Horgan and

Timmons’s Moral Twin Earth argument that focuses discussion of the problem.

Following recent developments in the philosophy of language and their implementation in various naturalization projects, metaethical naturalists offered semantic accounts intended to explain reference while allowing for the open questions that have worried naturalists since Moore. But the semantic accounts that seem so plausible in the case of natural kinds run into a serious problem when applied to moral concepts. Whatever the would-be naturalistic reference-fixing relation is (e.g., the ‘causal regulation’ relation proposed by Richard Boyd), it does not seem capable of picking out any particular piece 11

of naturalistic ontology in a way that respects the flexibility of moral concepts. With

natural kind terms and concepts like ‘water’, two communities related to different

properties (e.g., H2O and XYZ) seem to be either using different concepts or at least talking about different natural kinds, so that any apparent disagreement between the two groups over what counts as falling under the term or in question turns out to be a pseudo-disagreement, where both sides can be right. But the same does not seem to hold

for moral terms and concepts: when one community uses its terms in relation to one

property (e.g., that picked out as morally significant by a consequentialist theory) and the

other community in relation to another property (e.g., that picked out as morally

significant by a deontological theory), then as long as both communities give their terms

the right sort of a practical role in their life, there seems to be enough semantic common

ground to see them as genuinely disagreeing about morality, and any attempt to explain it

away as a pseudo-disagreement where both sides can be right seems implausible and

unmotivated.1

This problem is commonly seen as a thorn in the side of naturalism about moral

properties, but I think it generalizes to jeopardize all forms of metaethical descriptivism

and all normative ontology. It is not just naturalism that needs a semantic account capable

of explaining reference, and it is not just naturalism that is jeopardized by the flexibility

of normative concepts. Nonnaturalism, error-theory, and fictionalism all need a way of explaining the descriptive content they ascribe to normative concepts so as to redeem the view that there is some possible ontology that is (or at least would be if it were actual)

1 See Horgan and Timmons (1990–91), and also Horgan and Timmons (2000). 12

plausibly seen as the referent of those concepts. Without such an account, they labor

under the same burden as naturalism, and inherit the same implausibility.

Note that pressing this problem this does not only serve as an argument against

descriptivist theories. It also serves as an indirect argument for descriptivism’s greatest

rival: metaethical expressivism. Expressivists propose to understand normativity in terms

of distinctive psychological factors (having to do with sentiments, , and

decision-making), rather than in terms of ontology forming a distinctive subject-matter

for ordinary descriptive . And so any argument against descriptivism redounds to

the benefit of expressivism. But, given the dialectical situation in contemporary

metaethics, the semantic problem can do more than that. For expressivism’s toughest

problem is itself semantic: viz., the Frege-Geach problem that challenges expressivists to

find an expressivist-friendly, nondescriptive-attitude–based account of embedded

normative judgments that also applies to freestanding normative judgments so as to

underwrite the validity of basic normative reasoning. And thus, if expressivism’s

alternatives are all stuck with an equally tough semantic problem, then this provides

welcome relief for expressivists. It levels the metaethical playing field, saddling both camps with serious semantic difficulties.

In the following chapters, I will pursue this line of argument against some of the

most prominent descriptivist metaethical theories. I begin in Chapter 2 with the work of

naturalists Mark Schroeder and David Copp. Schroeder’s work provides an opportunity

to consider the semantic problem outside the moral domain, and show that it generalizes

to all normative reasons for action. Copp’s work, on the other hand, arguably represents 13

the most well-developed attempt to answer the semantic problem on behalf of metaethical naturalists.

In Chapter 3, I argue that the problem generalizes to nonnaturalism. the same semantic objections to naturalism that give nonnaturalism its plausibility serve equally well as objections against nonnaturalism. I review the is-ought gap and the Open

Question Argument in order to show how they cast doubt on direct nonnaturalistic analyses of moral terms and concepts. I then develop versions of the Moral Twin Earth argument that challenge nonnaturalism. Metaphysically modest versions of nonnaturalism like that defended by Russ Shafer-Landau are especially prone to this problem, since their differs little from naturalism. But metaphysically ambitious versions of nonnaturalism are also in trouble: here I examine the work of

David Enoch and Ralph Wedgwood. They defend an irreducibly normative version of conceptual role semantics intended to solve the semantic problem and escape the Moral

Twin Earth argument. But I argue that the supervenience constraint demanded of all normative domains brings their semantic account back into the line of fire. In Chapter 4, I consider versions of nonnaturalism which seek to avoid ontological commitments altogether. According to these ‘quietist’ views, moral judgments are descriptive, but nevertheless there are not any robust truth-conditions to explain the semantics of these judgments. Focusing on the recent work of Ronald Dworkin, I argue that such views face difficulties in distinguishing themselves from their putative metaethical alternatives: they can explain moral semantics in practical psychological terms and collapse into expressivism, explain them in ontological terms and abandon quietism, or else be left 14

without any account of moral semantics whatsoever. In the end, Dworkin’s view (and any

kindred quietism) is left in need of an account of descriptive content that somehow makes

do without reference to ontology.

In Chapter 5, I consider metaethical error theory and present an argument against

it based on the supervenience constraint. According to error theory, the ontology that

would count as a suit- able referent for moral terms and concepts does not actually exist.

But given the arguments pro- vided by error theorists, they seem to allow that such ontology could exist. This constitutes an immediate breach of supervenience, inasmuch as there could be a world exactly like ours, only with moral ontology added to it. But even worse, there is a dilemma turning on whether the possible worlds containing moral ontology conform to a supervenience constraint: if they do not, then the ontology does not count as moral ontology after all, and if they do, then error theorists are left with the same unexplained supervenience relations criticized by Mackie. Here I criticize Mackie’s appeal to theistic explanations of supervenience relations. Finally, I consider the possibility of a more radical error theory which denies the very possibility of moral ontology. I argue that such a view will have trouble maintaining a descriptive account of moral semantics.

In Chapter 6, I consider metaethical fictionalism and argue that it is vulnerable to the same problems afflicting error theory. Fictionalism, unlike most versions of descriptivism, denies that moral judgments are (or ought to be) descriptive beliefs.

Instead, they are non-belief attitudes directed at descriptive content. But the semantic problem is centered on content, not attitude, which means fictionalism can fare no better 15

than conventional forms of descriptivism. I press this point against the work of Richard

Joyce and Mark Kalderon.

Finally, in Chapter 7, I present an interpretation and defense of Blackburn’s supervenience argument against moral realism. I argue that it is at heart a semantic argument drawing on the deeply embedded conceptual necessity of the supervenience constraint, and that it applies not only to realist views, but descriptivist views generally.

This means that Blackburn’s argument is much more similar to the semantic problem I am pressing than is generally thought—thus the argument merits consideration for more than interpretive reasons. In developing this interpretation, I consider a variety of objections appearing in the literature that have grown around the argument and show that

Blackburn’s challenge is a forceful one that has yet to be answered.

16

CHAPTER 2: NATURALISM

As we have seen in the introductory chapter, metaethical naturalism faces a serious semantic problem: there does not appear to be any semantic account capable of explaining why moral terms refer to the naturalistic ontology they allegedly do. In this chapter, I pursue recent attempts to develop metaethical naturalism, arguing that they cannot provide a satisfactory solution to the problem.

2.1 Schroeder on normative reasons

I begin with Mark Schroeder’s recent defense of a Humean theory of normative reasons for action.2 According to his ‘Hypotheticalism’, these reasons are desire-based: a fact’s counting as a normative reason for an agent to perform an action consists in that fact’s part of what explains why the action promotes something desired by the agent.

Thus, supposing that Ronnie has a standing desire to go dancing, the fact that there will be dancing at a party counts as a normative reason for Ronnie to go to the party because it is part of what explains why going to the party promotes something desired by Ronnie.3

Schroeder develops the view with an eye towards accommodating the agent-neutral

reasons we expect of morality within a Humean framework. But I will be focusing on the

foundations of the view rather than on its development. For though metaethics is

concerned with moral reasons especially, the traditionally broad scope of ‘ethics’ allows

metaethics to comprise all normative reasons for action: moral, prudential, instrumental.

2 Schroeder (2007). 3 Schroeder (2007: 29). 17

And so regardless of whether Schroeder can successfully accommodate agent-neutral reasons, his theory is a piece of naturalistic metaethics. And as I shall argue, it is subject to all the same metaethical concerns as more familiar morality-focused naturalistic

approaches.

A convenient way to present my concerns is to consider Schroeder’s response to a

prominent line of anti-naturalist critique pressed by both Jean Hampton and Christine

Korsgaard.4 It turns on the assumption that a Humean theory is committed to attributing

to each agent a general reason to promote her desires, with this general reason underlying and explaining all particular reasons to perform particular actions that promote particular

desires. Schroeder puts the critique in the form of a dilemma. On the one hand, if this

general reason is to be consistent with the Humean theory, then it too must be based on

the agent’s desires. But this is impossible, for it is viciously circular to explain a reason in

terms of itself.5 On the other hand, if this general reason is an isolated exception to the

Humean theory, and therefore not based on the agent’s desires, then Humeans have no

principled basis for rejecting all other non-desire-based reasons and are guilty of

chauvinism.6

Schroeder’s response is to challenge the assumption that Humeans must posit a

general reason to promote one’s desires. To be sure, Humeans must accept the intuitively

4 Schroeder (2007: 41–60). 5 Why does Schroeder deem this a case of impossible circularity? I’m not entirely sure. Admittedly, such a view seems fairly unstable: if the general reason to promote one’s desires is successfully grounded in a general desire to promote one’s desires, then evidently this desire is capable of explaining this reason without outside help, which leaves it unclear why all an agent’s particular desires are not themselves capable of explaining associated particular reasons without outside help. Alternatively, if no desire can explain a reason without some further background reason, then we are off on a regress. But despite this serious problem, I see no circularity in the view. 6 Schroeder (2007: 46–50). 18

plausible but ambiguous claim that “[t]here is a reason for everyone to do what promotes

her desires”. But they need only accept it in the sense that “[f]or all agents x and actions

a, if doing a promotes one of x’s desires, then there is a reason r for x to do a”. And thus

they need not posit one general reason per agent (“to do whatever promotes one of x’s

desires”), much less some sort of cross-agent general reason (“for anyone to do whatever

promotes one of her desires”).7

Thus Schroeder rests content with a basic Humean biconditional, not to be explained in terms of any further or more general reason: “For R to be a reason for X to do A is for

there to be some p such that X has a desire whose is p, and the truth of R is part of what explains why X’s doing A promotes p”.8 But here the original objection resurfaces,

in the form of what Schroeder calls the “Revived Chauvinism” objection. If Humeans are

willing to accept basic desire-based conditions for reasons, then what principled basis do

they have for rejecting all other non-desire-based conditions for reasons? Why not an

additional condition which posits, say, reasons to be friendly to one’s neighbor?9

Schroeder’s final response is disarmingly straightforward. First, he stresses that his

theory is intended to provide a constitutive account of normative reasons, telling us “what

it is for R to be a reason for X to do A”. But his means that any additional conditions would compromise the metaphysical unity of the account, driving us to the conclusion

that “reasons are like pieces of jade”, with disjunctive accounts deemed tantamount to “a

kind of skepticism about reasons”. Second, he argues (throughout his book) that his

7 Schroeder (2007: 55). 8 Schroeder (2007: 57). 9 Schroeder (2007: 58–59). 19

Hypotheticalism “provides a better theory of what reasons are than do competing accounts”.10

Now, as for this second point, I do not intend to evaluate the merits of competing first-order normative accounts of what reasons we have. Schroeder might well be capable of showing that Hypotheticalism gives us the best first-order normative account on offer.

But this alone would presumably do nothing to resolve the metaethical questions at stake—not without considerable additional argument, in any case. Indeed, as we will see shortly, Hypotheticalism as a first-order normative account is compatible with any number of rival metaethical theories.

But the first point also stalls out. Even if we concede to Schroeder that disjunctive accounts of normative reasons are unacceptable, it hardly follows that Hypotheticalism carries the day by default. After all, many other reasonably plausible accounts of normative reasons can be dreamed up that are as unified as Schroeder’s: an account where my reasons depend on anyone’s desires (not just my own), an account where my reasons depend on my own pleasure, or on anyone’s pleasure, etc. Each of these accounts could be formulated in a biconditional like that of Schroeder’s, and could also be taken as providing a constitutive account of what it is to be a normative reason. And then the Re- vived Chauvinism objection would be restored in full force: what principled basis would

Humeans have for rejecting these other accounts?

The underlying problem, in my view, is that Schroeder’s theory is every bit as metaethically questionable, and every bit as in need of metaethical explanation, as any

10 Schroeder (2007: 59–60). 20

other theory. Thus it does not deserve any privileges or special treatment. Here I have

two points to make.

First, normative reasons call for just as much metaethical explanation as moral

properties. Just as we can ask “What sort of property could goodness or rightness be?”, likewise we can ask “What sort of entity is a normative reason?” To this, naturalists can

of course respond that moral properties are natural properties (e.g., pleasure), or that

normative reasons are natural entities (e.g., facts about actions promoting desires). But

just as we can ask for a semantic account capable of explaining why moral terms and

concepts refer to the naturalistic ontology they allegedly do, likewise we can ask for a semantic account capable of explaining why other normative terms and concepts refer to the naturalistic ontology they allegedly do. To see this more clearly, we can look to the

familiar predicates used in moral discourse and work up similar predicates for discourse

about normative reasons: e.g., ‘reason-supported’, ‘choiceworthy’, or ‘fitting’. Just as we

say an action is right or wrong for a person, we can say an action is choiceworthy or

unfitting for a person. And then the old metaethical questions are easily framed: what sort

of thing—choiceworthiness or fittingness—do these terms and concepts refer to? what

semantic account explains this reference?

Another way of illustrating the point is to note that, just as moral theories can be

grounded in a variety of metaethical theories, likewise theories about what normative

reasons we have can be grounded in a variety of metaethical theories. Humeanism about

normative reasons, for example, can take a nonnaturalist form, a naturalist form, a con-

structivist form, and an expressivist form. Nonnaturalist Humeans will say that there are 21

nonnatural normative properties belonging to actions that promote an agent’s desires only, naturalists that normative properties are nothing over and above the natural properties of these actions only, divine command theory constructivists that their fittingness consists in God’s recommending these actions only, and expressivists that their endorsement of these actions only is more a matter of practical attitude than of descriptive belief. Humeanism by itself, or (presumably) any theory of what normative reasons we have, does nothing to settle the question of which metaethical theory is correct.

Second, a theory that picks out actions which promote an agent’s desires as the supervenience base for normative reasons is a substantive theory that cannot be taken for granted, no more than we can take substantive ethical theories such as utilitarianism or

Rossian pluralism for granted. After all, it is not uncommon in the history of philosophy for thinkers to denigrate human desire as a mere product of the animal side of our , or even of the corrupt and damnable side of our nature. Of the mere fact that an action will promote an agent’s desire, these thinkers might well say that it provides no reason whatsoever for the action, or even that it provides reason against the action (perhaps be- cause this frustrates the desire and mortifies the passions). And even those with a more positive outlook on human desire might deny that it always gives agents a normative reason for actions which promote it: some desires might run contrary to prudence, some desires might be beneath one’s dignity, some desires might have any would-be normative import entirely silenced by the exigencies of the situation. Perhaps all these departures from Humeanism are wrong, but we certainly cannot take that for granted. 22

But this means that Schroeder’s theory is susceptible to the semantic problem I am

pressing against descriptivist theories. Just as traditional metaethical naturalists need a

semantic account, so too does Schroeder. And neither kind of theory appears more

promising than the other.

Consider a standard Moral Twin Earth scenario applied to Schroeder’s theory. One

community’s normative discourse is causally regulated by the facts about desire-

promoting actions that go with Hypotheticalism. Another community’s normative

discourse is causally regulated by different facts tied to a different and incompatible

theory of normative reasons—one in which desire isn’t everything. If causal regulation

fixes reference, then these two communities should be interpreted as not truly disagreeing

with each other. But this seems implausible: they appear to be in a fairly clear case of

normative disagreement. And the problem seems to remain, whatever reference-fixing

relation we bring in.

This conclusion reaffirms the thrust of Hampton and Korsgaard’s argument. As

Hampton puts it, hypothetical imperatives involve “the same mysterious objective

authority that attends the categorical imperative”. And thus Humean views about normative reasons get no free pass to metaethical complacency. They too must explain

the mysterious objective authority essential to normativity, starting with providing an

account of why certain natural facts ought to be seen as the true referents of our

normative terms and concepts.

23

2.2 Copp’s society-centered realist-expressivism

Next I turn to the work of David Copp.11 His work warrants examination for three

reasons. First, Copp is probably the most prominent and active defender of metaethical

naturalism in recent years. Second, he can stand as a representative of an emerging trend:

the development of ‘hybrid’ views that incorporate elements of both realism and

expressivism. And third, Copp arguably does the most of any naturalist to face up to the

semantic problem I am pressing against naturalism.

In critiquing Copp’s view, I will draw on a lesson taken from the previous section.

The lesson is that the semantic problem cannot be easily restricted to the domain of

morality, but instead has a strong tendency to carry over into all normative domains, or at

least domains concerned with normative reasons for action. And this imposes a plausible

constraint on attempts to provide a naturalistic semantic account for moral terms and

concepts: the account must not presuppose truths about other domains concerned with

normative reasons, not without then taking on the burden of providing a similar semantic

account for the relevant normative terms and concepts, or instead abandoning the

naturalization project halfway in. Thus it would be misguided to propose that ‘good’

refers to whichever natural property is most choiceworthy, or whichever natural property we ought to , or have normative reason to desire: this would only relocate the

question of normative semantics, without any significant metaethical having

been made.

11 Copp, D. (2007). 24

We should begin with a brief overview of Copp’s first-order moral theory, for it significantly informs his metaethical work. It is a relational “society-centered moral theory” that assigns moral properties in relation to a particular society and its needs. As

Copp himself puts it:

[The moral theory] identifies the property of rightness—that is, the property of being the right action in a context C (in relation to the relevant society S)—with the property of being required by the code of rules, whatever it is, the currency of which in S actually would best contribute to S’s ability to meet its needs—its needs, inter alia, for social stability, for peaceful cooperative interaction among its members, and for its members to be able to contribute to the overall flourishing of the society. Call this code the “best code” for the society in question. On this account, the facts as to which actions are right depend on which code of rules actually is the best code for the relevant society. But since different societies have the same basic needs, the best code for one society will be very similar to, if not the same as, the best code for another society, as long as their circumstances are basically the same in all relevant respects.12

This is at least a close cousin of rule-consequentialism, with societal needs occupying the role otherwise occupied by pleasure or desire-satisfaction, and an allowance that the best rules might vary somewhat from society to society. Notice that this theory by itself could be combined with any number of metaethical theories: nonnaturalism, naturalism, constructivism, expressivism, etc. Notice also that it is a substantive and controversial theory, and indeed Copp admits that “hardly anyone accepts the society-centered theory”.13

The first piece of metaethical work done by this theory emerges in Copp’s

incorporation of expressivist elements into his naturalist moral realism.14 He holds that though moral judgments are themselves ordinary descriptive beliefs about robust and

12 Copp (2007: 237). 13 Copp (2007:190). 14 See his “Realist-Expressivism: A Neglected Option for Moral Realism”, reprinted in Copp (2007: 153– 202). 25

naturalistic moral properties, they are also typically accompanied by the overtly practical attitude of subscribing oneself to a moral standard, and that consequently there are linguistic conventions in virtue of which moral discourse conventionally implicates such

accompanying attitudes.15 But why would moral beliefs typically be accompanied by

practical ‘subscription’ attitudes? Because moral beliefs are about behavioral standards

whose currency would serve the needs of the society. And if a behavioral standard were

not practically subscribed to, so as to influence behavior and decision-making, but only

believed in, then the needs of the society would never be served by its currency.16 But

then why would moral discourse conventionally implicate these attitudes? Because the

point of moral teaching, and also moral discourse generally, is to encourage others to

subscribe to the standards one sees as correct. And thus it is only natural to expect others

employing moral discourse to have the corresponding practical attitudes, and thus only

natural for linguistic communities to develop a on which moral discourse

implicates these attitudes.17

These expressivist elements might help Copp develop a response to the semantic

problem. After all, one difficulty faced by naturalists engaging with Moral Twin Earth is

that of accommodating the intuition that the two different communities have a genuine

moral disagreement with each other. And if, as Copp’s theory suggests, the two

communities can be expected to differ in the same sort of practical ‘subscription’

attitudes, then this practical opposition might form a sufficient basis for genuine moral

15 Copp (2007: 181–85). 16 Copp (2007: 186–87). 17 Copp (2007: 190–93). 26

disagreement. So hybridizing one’s naturalism by bringing in expressivist elements might

save naturalism from the semantic problem.

When Copp directly addresses Moral Twin Earth,18 he does not explicitly pursue this

line of response. But he begins with something very close in substance.19 In this

preliminary response, Copp points out that though the two communities may be referring

to different properties, they still have a great deal in common when it comes to their

moral terms. Namely, the terms “play the same role in people’s lives”. And this practical

commonality might be enough to deliver the intuitively favored result: they really do

have a moral disagreement over what to do. As Copp puts it:

[T]hey still disagree about something morally substantive, for they disagree about whether to lie in the imagined situation. They also would disagree about whether to urge people in similar situations not to lie, about whether to object to lying in such situations, about whether to teach their children not to lie in such situations, and so on…. Important matters can be at stake here, including whether to avoid actions that are wrong or whether instead to avoid actions that are twin-wrong, whether to teach children the moral code or instead to teach the twin-moral code, and so on. Our intuition that the Earthlings and the Twin Earthlings have something morally substantive to disagree about is therefore true, even if corresponding moral and twin- moral terms refer to different properties.20

In short, the practical elements shared by the two communities put them in practical

opposition, which may form a sufficient basis for genuine moral disagreement. Copp

concludes that the two communities’ terms are at least similar enough to pass the test of

inter-translatability, and that there are at least genuine “disagreements between the

Earthlings and Twin-Earthlings about what to do”. Even if expressivist hybridization is not explicitly noted as doing the work, I see no more than a nominal difference between

18 See his “Milk, Honey, and the Good Life on Moral Twin Earth”, reprinted in Copp (2007: 203–29). 19 Copp (2007: 212–16). 20 Copp (2007: 215). 27

the practical role played in people’s lives and the practical ‘subscription’ attitudes of his

realist-expressivist project.

But, as Copp recognizes, mere inter-translatability might be not enough of an achievement to satisfy our intuitions. We want more than the rough convergence of a bilingual dictionary, we want the two communities to disagree about the truth-value of one and the same proposition.21 To accommodate this stronger intuition, Copp pursues a

slightly different line of response. Here the general point is that, when it comes to

Putnam-style reference-fixing semantic accounts, there is more to semantics than the

piece of ontology serving as the referent: there are also the two important factors of

referential and speaker interests. But then, once we bring these factors into

account, not only will the two moral communities have a great deal of overlap when it

comes to the semantics of their moral terms, but we will also see that they are in fact

referring to one and same piece of moral ontology—for the two semantic factors in

question provide us with enough to determine a referent for moral terms. And this would

clearly be sufficient to establish genuine moral disagreement of the strongest sort.22

His example of ‘milk’ illustrates the two factors at work. Though different kinds of

animal milk have are heterogeneous in structure at the level of chemistry, there is still a

shared nature “at the level of functional and genetic properties”.23 This fact enters into our semantic , so that we use the term ‘milk’ with a referential

“to refer to whatever has the relevantly same nature as the local samples”.24 Likewise,

21 Copp (2007: 216). 22 Copp (2007: 216–27). 23 Unaccountably, Copp chooses not to pursue any puns about “homogenized” milk. 24 Copp (2007: 218). 28

since our interest in milk has mainly to do with new mothers feeding their young, the relevant similarities across the different chemicals we call ‘milk’ are again functional and genetic.25 This means we can make sense of a Milky Twin Earth scenario thus: if one community is unacquainted with sheep milk, and the other is unacquainted with goat milk, and they disagree over whether dog milk is ‘milk’ because of a slight color difference, we can explain why this is a genuine disagreement about milk, and indeed settle the question of whether dog milk is ‘milk’, simply by invoking the shared referential intentions and speaker interests of the two communities. Convergence in these semantic factors is sufficient to determine the reference of ‘milk’, and therefore sufficient to secure genuine disagreement about milk, despite the fact that the relevant liquids on the different planets do not share the same chemical nature.26

Of course, to put this point to work on behalf of metaethical naturalism, Copp needs to offer an account of the referential intentions and speaker interests associated with moral terms that can safely be generalized across all moral communities. Using ‘wrong’ as an example, he has this to say about the referential intentions associated with moral terms:

[I]n using “wrong” (in a nondeviant case), an English speaker intends to refer to actions that are of the kind, or that have the property, that he and most speakers in his linguistic community intend to refer to in using “wrong.” Or, more simply, in using “wrong,” an English speaker intends to refer to actions that are of the kind, wrong, or that have the property of being wrong or of being not-to-be-done. And speakers of Twin English would have to have the same intentions. If they did not, the relevant terms in their language would not be moral terms.27

25 Copp (2007: 221). 26 Copp (2007: 225–27). 27 Copp (2007: 223). 29

And for the speaker interests associated with moral terms:

Their interest is to pick out the kind of action or property of actions, whatever that might be, that is of primary importance morally in deciding which actions to avoid— the kind that is especially to be avoided, or the property of being morally “to-be- avoided” or “not-to-be-done.”28

Thus our referential intentions for ‘wrong’ are to refer to whatever has the same nature as

the intended referent of our linguistic community. And our speaker interests for ‘wrong’

are to pick out whatever has the most primary negative normative bearing on action.

It is plausible enough that this account, lean as it is, can be generalized across all

moral communities. And if so, then all moral communities can be reckoned to have a

great deal of semantic common ground. But is this common ground enough to determine

a common referent for a moral term like ‘wrong’? Here Copp appeals to the stipulated

near-indiscernibility of Earth and Twin Earth. Drastically different environments might differ in all sorts of morally relevant respects, but since these two environments seem to

be alike in all morally relevant respects, then the terms of the different communities

should end up referring to one and the same property of wrongness. But then genuine

moral disagreement is secured, and the threat of Moral Twin Earth is averted.29

In a follow-up paper, Copp aims to “fill in some of the details” of his proposal to

make clearer still why it does not run afoul of Moral Twin Earth.30 The major move is to

bring his first-order moral theory into his account of the referential intentions associated

with moral terms. Copp writes:

28 Copp (2007: 224–25). 29 Copp (2007: 225). 30 See his “Referring to Moral Properties: Moral Twin Earth, Again”, in Copp (2007: 230–45). 30

The basic would be that “right” is used with the semantic intention of ascribing to an action or a kind of action the property of being required by the code of rules, whatever it is, the currency of which in the society in question would best contribute to the society’s ability to meet its needs. Or, more colloquially, it is the intention of ascribing the property of being required by the code that would work best for “our” society—the code, the currency of which among us would enable us—the society comprising all of us—to cope as well as can be with our common problems. The idea then is that the facts about the nature of the relevant society and the circumstances it faces determine which code is actually the best code for the society and so determine which property is ascribed by “right.”31

Notice that this account of referential intentions goes well beyond the lean and formal

‘same nature of the intended referent’ account given earlier. Something like rule- consequentialism has now been enshrined in the very semantics of moral terms.

A second, minor move is to make the case for allowing a measure of indeterminacy.

Most ordinary terms are bound to be vague, Copp reminds us, given the -boggling complexity of and the limited resources of human and natural language.

Thus it would be no surprise if an otherwise successful semantic account were stuck with vagueness and indeterminacy in assigning referents to moral terms. Thus if applying a

Putnam-style semantic account leaves us with a great number of eligible referents for moral terms, this is no problem, so long as the results are no worse off than those for

“water” or “milk”.32

Finally, Copp draws on these moves in replying to three possible Moral Twin Earth

objections. First, what if the two communities have different referential intentions, and

yet they still intuitively seem to have a genuine moral disagreement? Copp thinks it begs

the question against the Putnam-style account to assume that is a real possibility. For,

31 Copp (2007: 238). 32 Copp (2007: 234–36). 31

“given [the] intuition [that the two communities’ terms have the same meaning”, the

Putnamian account would take the Earthlings and Twin Earthlings to have the same

semantic intentions in using their terms”. And if we stipulate different semantic

intentions, then Copp is confident that holding this stipulation in mind will yield the

intuition that the groups are not genuinely disagreeing.33

Second, what if the terms play “the same kind of practical role” in the communities’

lives, despite the (stipulated) different semantic intentions? Wouldn’t this bring us back

to the intuition that there is a genuine moral disagreement? To some extent, Copp replies,

but only enough to attribute sameness of meaning in the low-level sense of inter- translatability. As before, practical similarity might be enough to satisfy the rough-and- ready standards of a bilingual dictionary, but not enough to show that the terms have the same meaning in any stronger sense.34

Finally, what about the indeterminacy that might remain when identifying the best

code for a society? Isn’t this a likely failure of Copp’s account? No, he replies, for even if

there is not a single best code for a society, there will still be “a (non-empty) of codes that are tied as ‘best’”. To introduce an example, it is unlikely that there will be a single best code when it comes to May-December relationships. And thus there may be no fact of the matter about the general rightness or wrongness of a 60-year old dating a 35-year old, since the top codes disagree with each other. But despite this indeterminacy, perhaps the top codes will all tie for best, converging on e.g. the wrongness of a 51-year old man

33 Copp (2007: 240–41). 34 Copp (2007: 241–43). 32

marrying a 16-year old girl, so that moral terms are no worse off than “water” or

“milk”.35

2.3 Critique of Copp

We may organize Copp’s responses into three main categories, depending on what factors he draws on to secure genuine moral disagreement: (i) appeal to practical similarities; (ii) appeal to lean referential intentions and speaker interests; and (iii) appeal to substantive referential intentions involving his society-centered moral theory. I will critique each in turn.

First, there is the appeal to practical similarities. The moral terms in both communities “play the same role in people’s lives”, and this is an important commonality that might pave the way for genuine moral disagreement. Copp himself notes that this will probably only secure a low level of cross-community engagement: only enough for inter-translatability. But it is still worth seeing why this kind of response will not solve the semantic problem.

One problem is the threat of . As Horgan and Timmons note in their response to Copp’s first Moral Twin Earth paper, it is not enough to have a practical opposition of the sort Stevenson famously called a “disagreement in attitude”. For if the two communities’ moral terms refer to different properties, then we get community- relativized truth-conditions. Both sides could be saying something true even when they have a genuine moral disagreement. And this is an unacceptable result for metaethical

35 Copp (2007: 243–44). 33

naturalists.36 Of course, one might think that Copp would be happy to accept relativistic

consequences, owing to his ‘relational’ society-centered moral theory. But this would be

a mistake: Copp allows for rightness and wrongness to vary across communities only to

the extent that the communities have different needs and different codes of rules capable

of meeting those needs. He, like any realist, would be unwilling to allow for relativism

about judgments directed at one and the same fixed scenario, which is what we end up

with when different communities refer to different properties due to a genuine moral

disagreement.37

Moreover, there is the problem of collapsing into expressivism. Copp wishes to

characterize the expressivist elements of his account as inessential to moral judgment itself, but typically associated with it to such an extent that conventional implicature of these elements results. But if the elements are inessential, then presumably there could be a community that lacks these elements: they make moral judgments and refer to moral properties, but with only a weak or nonexistent connection to practical ‘subscription’ attitudes. If such a community could exist, then there are no practical similarities that could bring them under the umbrella of morality, so that they could be interpreted as genuinely agreeing or disagreeing about morality. And since (ex hypothesi) this is our only way of accounting for disagreement, then they would be excluded from the moral game: their judgments would not be moral judgments after all. But this would mean that what brings a community’s terms and concepts under the umbrella of morality has nothing to do with the descriptive side of Copp’s story. What makes a moral term a moral

36 Horgan, T. and M. Timmons. (2000: 143). 37 Copp (2007: 236–37). 34

term, or a moral concept a moral concept, would be its practical nondescriptive role. And this, of course, would be metaethical expressivism, with relativist trimmings.

Copp could avoid this conclusion if he were to switch tactics and make a case for treating the expressivist elements as essential to moral judgment, and necessarily following from the descriptive side of his account. That is, if no community could manage to make descriptive moral judgments and refer to moral properties without giving moral evaluation a certain kind of practical relevance in their lives, then the community mentioned above could not possibly exist. But this is a tall order. It is at least common for descriptive beliefs to lack any necessary connection to practical decision-making—we can take it or leave it, with different people responding with positivity or negativity or indifference. And it is highly controversial whether any descriptive beliefs could have such direct practical implications that no community could have those beliefs without also having certain practical attitudes.

To make the case, Copp could draw on his first-order moral theory relating rightness and wrongness to the needs of society. Perhaps if moral evaluation were essentially a matter of classifying actions according to need-based social codes, as Copp suggests in his most recent account of referential intentions, then we could expect moral evaluation to carry direct practical implications. But this faces two serious problems. First, it rests on the that all communities engaged in moral evaluation could be expected to care about the needs of their own society in an especially strong way. And this presupposition, though plausible enough for the communities most of us are familiar with, might easily falter when it comes to communities that have internalized a moral 35

code of humility and self-abnegation. And second, it rests on the presupposition that moral evaluation essentially incorporates something like Copp’s first-order moral theory.

As we will see shortly, this presupposition appears to be untrue.

Second, Copp could appeal to the lean referential intentions and speaker interests of his first Moral Twin Earth paper. Recall that our referential intentions for ‘wrong’ are to refer to whatever has the same nature as the intended referent of our linguistic community, and our speaker interests for ‘wrong’ are to pick out whatever has the most primary negative normative bearing on action.

The problem here is one of extreme indeterminacy, owing to circularity and regress.

Consider: in order to determine what satisfies our referential intentions, we must first identify what our linguistic community intends to refer to with ‘wrong’. But then we are faced with virtually the same semantic problem we began with: explaining why our community’s term ‘wrong’ refers to what it does by bringing in additional about what our community intends to refer to. And if all we know is that they intend to refer to what is ‘not-to-be-done’, then we are off on a regress: this is a brazenly normative term, in need of exactly the same sort of semantic explanation as ‘wrong’, which means nothing has been gained.

Or consider speaker interests: in order to determine what satisfies our interest in avoiding what we ought to avoid, we must identify some negative normative property.

But this is tantamount to identifying wrongness—there is nothing easier about the one task as opposed to the other. Without additional information about what might qualify as 36

‘to-be-avoided’, we have nothing to go on and are incapable of making even the grossest discrimination between different candidate referents.

Copp might reply that he is willing to accept a reasonable measure of indeterminacy,

of the sort found in ‘water’ and ‘milk’. But the current proposal leads to extreme

indeterminacy: virtually every first-order moral theory imaginable could claim to be vindicated by these referential intentions and speaker interests, since each could claim to have identified what is ‘not-to-be-done’ with their theory of wrong action. The resulting indeterminacy is thus not the sort that could be lived with: there would be no fact of the matter whether to serve society’s needs or destroy society entirely, whether to eat animals or torture them or crown them King of England.

Lastly, Copp could appeal to his more substantive account of referential intentions.

Perhaps we intend to refer to what is required or forbidden by the code of rules best at meeting society’s needs. This would at least rule out moral theories according to which one’s society should be destroyed.

But could it do more than that? Could it achieve a reasonable measure of determinacy? It would seem not. The term ‘need’ is one of those contested terms that straddles the boundary between normative and non-normative. Consequently, opinions will vary across persons and across communities as to what a society’s needs truly are.

Hedonists and perfectionists, for example, can be expected to differ sharply about a whole range of scenarios and what counts as meeting the needs of the societies envisioned. But if there is no fact of the matter as to whether Brave New World is a utopia or a dystopia, some vital moral questions are left dangling. 37

But matters are worse than that. It cannot be safely taken for granted that morality has

anything to do with meeting the needs of societies. Copp himself has acknowledged, as

earlier noted, that “hardly anyone accepts the society-centered theory”. And any

community drawn to a ‘Fiat iustitia, pereat mundus’ take on morality is unlikely to warm

up to anything approaching rule-consequentialism. But then it seems that Copp will have to rule out embarrassing numbers of moral thinkers as not even making moral judgments in the first place.

To bring this point home, I have to overcome the objection that I am simply begging the question. After all, if we have the intuition that the two communities have a genuine moral disagreement, doesn’t it beg the question against Copp’s semantic account to assume that they have different referential intentions rather than the same referential intentions? I think not. First, I think it is very clear that we can imagine a Twin Earth community with referential intentions other than those Copp proposes: e.g., a Kantian community. The question, then, is how plausible it is to interpret the apparent moral disagreement between this community and a Copp-friendly community as a genuine moral disagreement. If it is highly plausible, as I think it is, then this is bad news for

Copp’s semantic account. It does not beg the question against an account to apply it to a clear situation and see whether it yields counterintuitive results.

To make this still easier, we can simply imagine a Twin Earth community that has no referential intentions whatsoever. This nullifies the question of whether the two communities’ referential intentions are the same or different. And if the Twin Earth community has all the practical trappings of moral evaluation and apparently moral terms 38

to go with it, it again seems plausible to interpret the apparent moral disagreement as

genuine.

But instead of engaging the question-begging charge directly, we can take the route of parody. A textbook utilitarian could build utilitarian referential intentions into their account of moral semantics, a Kantian could do the same, and so on for Aristotelian virtue ethicists and Rossian pluralists and Thomist natural law theorists and moral relativists and Stoics and Hobbesians. Imagine a diverse line-up of communities featuring these diverse referential intentions. Here we are stipulating different referential

intentions, as different as all the different first-order moral theories, which according to

Copp should mean that the genuineness of the moral disagreement should seem implausible as we keep in mind the stipulation. But the opposite seems true: keeping in mind the stipulation of different referential intentions corresponding to different first- order moral theories only fuels the intention that we are dealing with genuine moral disagreement. Moreover, the charge that these different referential intentions lie outside the umbrella of morality could always be returned with equal force, for there is no apparent reason to favor Copp’s society-centered moral theory over any of the dozens of other theories regularly taught in ethics classes.

Thus none of Copp’s responses are capable of solving the semantic problem.

Hybridizing one’s view with expressivism only secures moral disagreement at the cost of relativism and ultimately a collapse into an expressivist account of moral evaluation.

Bringing in such additional semantic factors as referential intentions does nothing to 39

secure a reasonable measure of determinacy unless one loads them with first-order moral claims of the sort explicitly denied by many moral thinkers.

40

CHAPTER 3: NONNATURALISM

Metaethical nonnaturalism is commonly rejected for rather programmatic reasons. Its

nonnatural properties—located in a mysterious metaphysical realm beyond nature—

offend our naturalistic scruples and scientific . And our knowledge of these otherworldly properties is no less suspect: none of our ordinary ways of finding out about the world seem to apply, leaving only a quaint faculty of intellectual intuition that clashes with even the broadest sort of .38 But for nonnaturalists, these worries can

seem question-begging at best and mere prejudice at worst. If morality resists

naturalization, and downright is literally unbelievable, why shouldn’t we expand

our horizons and break with naturalistic orthodoxy? Surely it is not a compelling

criticism of an unabashedly anti-naturalistic view that it fits poorly with naturalism.39

I hope to provide an argument against nonnaturalism that steers clear of contentious

naturalistic presuppositions. The problem with nonnaturalism, I contend, lies in its

semantic commitments. Nonnaturalists have always scorned their naturalist rivals for

failing to explain why moral terms and concepts refer to the naturalistic ontology they

allegedly do: such semantic arguments as the is-ought gap, the Open Question Argument,

and Moral Twin Earth all present an obstacle to metaethical naturalism. But I will argue

that the same problems apply to nonnaturalism, and that no advantage is gained by

having moral terms and concepts refer to nonnaturalistic ontology. And because self-de-

scribed ‘nonnaturalists’ vary in their metaphysical commitments (with some disclaiming

38 See Mackie (1977: 30–42) for a well-known presentation of these concerns. 39 See Huemer (2005: 228–30, 239–48) for a presentation of this sort of response. 41

any metaphysical commitments),40 and because versions of nonnaturalism that depart significantly from naturalism stand a better chance of avoiding naturalism’s problems, I make a point of focusing on the most metaphysically ambitious form of nonnaturalism, what I call ‘textbook nonnaturalism’.

3.1 Textbook nonnaturalism

Textbook metaethical nonnaturalism is the view that there are objective facts lying beyond the natural world and forming the subject matter of morality. This distinguishes it from other textbook views: metaethical naturalism (on which there are objective facts belonging to the natural world which form the subject matter of morality), from construc- tivism (on which morality is about facts that are less than objective, because constructed from stances taken by subjects), from error-theory (on which the objective facts lying beyond the natural world forming the purported subject matter of morality do not really exist), and finally from expressivism (on which morality does not even purport to have a factual subject matter, since it is primarily a matter of nondescriptive action-guiding attitudes).

The glory days of nonnaturalism can be found in the first half of the twentieth century, in the works of G. E. Moore, C. D. Broad, W. D. Ross, H. A. Prichard, and A. C.

Ewing. And it is from this era that what I am calling ‘textbook nonnaturalism’ finds its origin. Of course, the same sort of view can be arguably be found in Henry Sidgwick’s

Methods of Ethics and perhaps even in the rationalist British Moralists Ralph Cudworth,

40 See Dworkin (2011), Scanlon (2003), Parfit (2006), and Dancy (2006). 42

Samuel Clarke, John Balguy, and Richard Price. But the view I wish to examine was first

explicitly developed as a metaethical rival to naturalism and expressivism in the twentieth

century.

And though the glory days are gone, textbook nonnaturalism may be making a

comeback. In what is perhaps the most prominent sign, Russ Shafer-Landau has declared

himself a nonnaturalist. Unfortunately, it is quite debatable whether Shafer-Landau’s

view counts as a genuine case of textbook nonnaturalism. For while Shafer-Landau con- tends that “moral properties are sui generis, and not identical to any natural properties”,41

this non- is less dramatic than it might seem: instantiations of moral properties are

still “exhaustively constituted”42 by instantiations of the descriptive natural properties

studied by the natural sciences, and the resulting metaphysics is deemed nonnaturalistic

only because it rejects a very strong type-type identity version of reductionism of the sort rejected by many physicalists in the .43 Assuming that these

physicalists cannot be said to countenance any mental reality over and above the physical

world, then Shafer-Landau cannot be said to countenance any moral reality lying beyond

the natural world. Of course, textbook nonnaturalism follows Moore and the emergentists

in recognizing the supervenience of the moral on the natural, but it seems incompatible with any moral-natural relation so close as exhaustive constitution. And indeed Shafer-

Landau himself acknowledges that his best claim to the title of nonnaturalist lies in his a

priori-friendly moral .44

41 Shafer-Landau (2003: 66) 42 See Shafer-Landau (2003: 74–6) for discussion of exhaustive constitution. 43 Shafer-Landau (2003: 74–5) 44 Shafer-Landau (2003: 61–5) 43

But other nonnaturalists, fitting more comfortably under the textbook nonnaturalist

umbrella, have begun to show themselves. Michael Huemer, who terms himself an

‘intuitionist’, contends that “moral properties are irreducible, non-natural properties”,45

and rejects any synthetic scientific reduction of moral properties akin to such standard

cases as the reduction of heat, water, and sound.46 This would seem to separate his view

from that of Shafer-Landau, who appears to accept a comparison of the status of morality

with that of such special sciences as biology.47 William J. FitzPatrick explicitly

distinguishes his view from Shafer-Landau’s, arguing for supervenience without exhaustive constitution,48 proposing “fundamental metaphysical facts of value ladenness

at [a] basic level” at which “there may not be anything more for philosophy to say”,49 and commenting that his view “would count as non-naturalistic by any criterion I am aware of”.50 Alas, FitzPatrick terms his view more Aristotelian than Platonic,51 which makes

David Enoch’s nonnaturalism even more welcome: though he modestly disavows the

historical competency needed to embrace the term, he “will not be offended if you call

[him] a Platonist”,52 and posits “normative truth[s] out there, as it were, in ’s

heaven”.53 This is textbook nonnaturalism.

Now, such a view is virtually defined by its ontological extravagance, or (more nicely put) its ontological affluence. Consequently, due to the minimalist aesthetic currently in

45 Huemer, (2005: 67) 46 Huemer (2005: 83–94) 47 See Shafer-Landau (2003: 72) for this comparison. 48 FitzPatrick (2008: 190–4) 49 FitzPatrick (2008: 190–7) 50 FitzPatrick (2008: 192 n.77) 51 See e.g. FitzPatrick (2008: 196) 52 Enoch (forthcoming: 16) 53 Enoch (forthcoming: 295) 44

fashion, the view acquires much of its appeal by pointing out the flaws of its rivals. And

the rival that makes nonnaturalism a distinct metaethical view of its own is of course

naturalism. Thus nonnaturalists have traditionally argued that there are important semantic considerations which go to show that moral facts cannot belong to the natural world. I will now begin arguing that these semantic considerations cannot be escaped by a flight to Plato’s heaven.

3.2 Semantic arguments against naturalism

Why can’t moral facts belong to the natural world? Perhaps the most commonly given

reason comes down to this: naturalistic facts are not what people actually have in mind

when they make moral judgments. When sincere claims about Santa Claus are construed

as mere claims about Christmas cheer, we may smile indulgently, but we do not take the

construal seriously. When the word ‘God’ is taken as a mere label for the natural

universe, we may welcome the redefinition, but we still recognize the possibility of

serious misunderstanding. Likewise, we might think, it is just not accurate to claim that

moral judgments are merely descriptions of the natural world and the facts belonging to

it. This rough idea finds expression in two classic arguments against metaethical

naturalism: the is-ought gap and the Open Question Argument.

The is-ought gap is grounded in the claim that no moral judgments, and indeed no

normative judgments generally, can be validly derived from mere descriptive statements

of natural fact. After all, any such derivation would require a bridge premise assigning

some normative status to the relevant naturalistic facts, and any such bridge premise 45

would itself be a full-blooded normative judgment, as opposed to a mere descriptive

statement of natural fact or even the sort of definitional premise that might be acceptable

in a valid derivation. On the (controversial) assumption that metaethical naturalism

requires the validity of such is-ought derivations, the is-ought gap shows that metaethical

naturalism is false.54

Thus the gap is often seen as a thorn in the side of naturalism. David Brink, in a

prominent book-length defense of naturalism, spends an entire chapter considering the gap.55 Michael Huemer’s recent critique of naturalism presses the gap as a reason to

doubt certain forms of naturalism.56 Stephen Finlay’s recent survey of moral realism

brings it up in much the same way, as a reason to doubt that naturalism can accommodate

the normativity of morality.57 The is-ought gap is recognized as a problem that naturalists must respond to.

The Open Question Argument is that since moral terms (indeed, evaluative terms generally) cannot be given a naturalistic analysis, therefore moral facts cannot belong to the natural world. The problem with naturalistic analyses of moral terms is that, for any naturalistic predicate N, it will make perfect sense for a competent speaker to wonder “x

is N, but is x good?”—i.e., such questions will seem open. And since a competent

speaker’s sense of the openness of a question is a telltale sign of the feasibility of an

54 For a recent critique consideration of the is-ought gap, see Sinnott-Armstrong (2006: ch. 7) 55 Brink (1989: ch. 6) 56 Huemer (2005: 72–82) 57 Finlay (2007: 15–16) 46

analysis, it follows that naturalistic analyses are unsuccessful. And as go naturalistic

analyses, the argument concludes, so goes metaethical naturalism itself.58

The Open Question Argument can fairly be deemed the argument against naturalism.

Brink’s discussion of the is-ought gap quickly turns into a discussion of the Open

Question Argument,59 and Finlay’s presentation ties the two together.60 Huemer’s

critique of naturalism begins with a defense of the argument.61 Frank Jackson’s recent de-

fense of naturalism devotes a section to the argument.62 And Shafer-Landau’s critique of

naturalism begins with a qualified defense of the argument, as successfully shifting the

burden of proof onto naturalists.63

Both of these arguments can be justly branded semantic arguments against naturalism. The is-ought gap relies on such semantic notions as validity and definition.

The Open Question Argument draws its metaphysical conclusions from semantic premises about analysis. What both have in common is the view that there is a semantic gap between moral concepts and naturalistic concepts: otherwise the is-ought gap could be closed with a definitional bridge premise, and certain naturalistic predicates could close the open question.

But of course naturalism is not so easily defeated. Even if it is true that moral concepts and naturalistic concepts are separated by a semantic gap, it simply does not

58 See Moore (1903, 1993: ch. 1), and also Ayer (1936: 106–08) 59 Brink (1989:150–70) 60 Finlay (2007: 15–16) 61 Huemer (2005: 67–72) 62 Jackson (1998: 150–153) 63 Shafer-Landau (2003: 56–8) 47

64 follow that metaethical naturalism itself is false. As the case of water and H2O

illustrates, even when two distinct concepts are separated by a semantic gap in the mind

of competent speakers, the two concepts may yet refer to the very same metaphysical

reality. Thus despite the semantic gap separating them, moral and natural concepts may

refer to the very same facts, belonging to the natural world, just as naturalism says. Thus

the is-ought gap can be acknowledged by naturalists without compromising their

naturalism. Likewise, naturalists can cheerfully agree with the Open Question Argument

that moral terms cannot be given a successful naturalistic analysis. The thesis definitive

of metaethical naturalism, after all, is a metaphysical thesis, not a semantic one, so why

couldn’t naturalists learn to live with the semantic gap?

Suppose, then, that naturalists can live with the semantic gap. Still, there is a price to

be paid. Naturalists must develop a semantic account explaining why it is that the

concepts separated by a semantic gap nevertheless refer to the same metaphysical reality.

After all, in cases like water and H2O, we have a plausible enough semantic account on hand: viz., the term ‘water’ has its reference given by its historical and external causal connections, which (as a matter of empirical fact) links it to the substance H2O. If

naturalism is to live with a semantic gap, it needs a semantic account that can perform the

same sort of explanatory task.

Naturalists acknowledge this price and they have tried to pay it. Still following the

case of water and H2O, Richard Boyd has argued that moral terms have their reference given by their historical and external causal connections. Thus the term ‘good’ refers to

64 See Durrant (1970), Brink (1989: 156–67) 48

whichever naturalistic property in our environment causally regulates our use of the term.

If it turns out to be regulated by pleasure alone, then ‘good’ refers to pleasure, even if “x

is pleasant, but is x good?” is an open question, and even if ‘good’ cannot be given a

naturalistic analysis in terms of pleasure. Moral terms, like natural kind terms, do not need analyses to capture what competent speakers have in mind. Instead, they need syn-

thetic “natural definitions”, indicating the underlying scientific nature of what the term

refers to.65

But Boyd’s semantic account runs in trouble, and this trouble brings us to the final

semantic argument against naturalism: the Moral Twin Earth argument.66 If moral terms,

like natural kind terms, have their reference given by their historical and external causal

connections, then there should be a Moral Twin Earth story for moral terms that runs

parallel to the Twin Earth story for natural kind terms. Now, in the Twin Earth story,

normal Earthlings have our term ‘water’ causally regulated by H2O, Twin Earthlings

have their term ‘water’ causally regulated by XYZ, and the result is that our terms refer

to different substances, so that any apparent interplanetary disagreement over whether a

particular sample counts as ‘water’ would in fact be a pseudo-disagreement. Thus, in a

Moral Twin Earth story, where normal Earthlings have our term ‘good’ causally

regulated by one set of naturalistic properties (a set tied to a deontologist moral theory),

and where Twin Earthlings have their term ‘good’ (a term used in apparently evaluative

ways) causally regulated by a different set of naturalistic properties (a set tied to a

consequentialist moral theory), the result ought to be the same: the terms refer to different

65 Boyd (1988: 194–5) 66 Horgan and Timmons (1992), and also Horgan and Timmons (2000) 49

properties, so any interplanetary disagreement is a pseudo-disagreement, with each side

speaking truly. But since it seems plausible that such a disagreement would be a genuine

moral disagreement, the stories do not run in parallel like they should. Thus the causal

theory of reference that seems so plausible for natural kind terms cannot be applied to

moral terms.

And the Moral Twin Earth argument aspires to generality: it is not merely intended as

a problem for Boyd-style causal theories of reference, but as a problem for all semantic

accounts that might interest naturalists learning to live with a semantic gap.67 For,

whatever term-to-property relation, causal or non-causal, one proposes as the relation that

fixes reference, it is plausible that a disappointing Moral Twin Earth story can be

concocted for that relation: i.e., a story in which two societies use similar terms so that

they bear the allegedly reference-fixing relation to different properties, and yet intersociety disagreement cannot be plausibly explained away as pseudo-disagreement.

And, if every semantic account is undone by the Moral Twin Earth argument, then naturalists can never square their view with the semantic gap. What the is-ought gap and the Open Question Argument started, the Moral Twin Earth argument threatens to finish.

Nonnaturalists, of course, are not unhappy with this development. Shafer-Landau

endorses the Moral Twin Earth argument against Boyd’s view, and argues that without a

plausible semantic account for naturalism, “the burden created by open-question

considerations becomes commensurately more powerful”.68 Huemer notes that though

certain naturalists might “appear to escape the Open Question Argument”, he finds the

67 See Horgan & Timmons (2000: 139–42) 68 Shafer-Landau (2003: 71) 50

Moral Twin Earth argument “persuasive” evidence of the contrary.69 If naturalists have

no semantic account enabling an escape from the combination of the Open Question

Argument and the Moral Twin Earth argument, then nonnaturalism gains an advantage

over its oldest metaethical rival.

3.3 Hume and Moore against nonnaturalism

Though the three semantic arguments just considered have most commonly been treated as anti-naturalist arguments only, I think they have a much broader scope. In particular, I think they work equally well as anti-nonnaturalist arguments. The Moral Twin Earth

argument has the most contemporary relevance, and I intend to focus on it. But I want to

start with a historical review of the other two arguments, as illustrations of how doubts

about the naturalizing of moral facts can carry over into doubts about the nonnaturalizing

of moral facts.

Start with the is-ought gap. Suppose that it is a genuine gap, and that no normative

judgment can be validly derived from descriptive statements of natural fact. But what

about descriptive statements of nonnatural fact? Take ‘ectoplasm’, the hypothetical

nonphysical ontology sometimes discussed by philosophers of mind.70 Could a normative

judgment be validly derived from a description of all the ectoplasmic facts in the world—

e.g., what objects ectoplasmic properties are instantiated in? Not obviously! As before,

we would seem to be in need of a bridge premise assigning some normative status to the

ectoplasmic facts, and any such bridge premise would itself be a full-blooded normative

69 Huemer (2005: 265 n.29) 70 See e.g. Churchland (1984), Lewis (1988) 51

judgment, as opposed to a mere descriptive statement of nonnatural fact or the sort of

definitional premise that might be acceptable in a valid derivation. Naturally, one could

try simply stipulating that the ectoplasm in question is itself normative, so that no

derivation is needed. But this tactic seems no more justified than the parallel stipulative

tactic would be for naturalists helping themselves to the normativity of certain natural

facts.

This problem is already well-known in discussions of the most famous exception to

naturalism in all of philosophy: God.71 For it is unclear how normative judgments are to

be validly derived from descriptive statements of theological fact. Discussions of divine

command theory very frequently bring in Hume’s is-ought passage in order to make

trouble for the view,72 and a return to Hume himself underscores the point:

In every system of morality, which I have hitherto met with, I have always remark’d, that the author proceeds for some time in the ordinary way of reasoning, and establishes the being of a God, or makes concerning human affairs; when of a sudden I am surpriz’d to find, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not.73

Note that Hume’s only two examples of is-propositions have as their subject matter “the

being of a God” and “human affairs”: the latter natural and the former nonnatural

(indeed, supernatural).

Nor does the matter rest with God. The is-ought paragraph comes at the end of a section of the Treatise whose thesis is that mere reasoning is incapable of performing moral evaluation. And Hume’s main target is the moral of such Christians as

71 Cf. Plantinga (2010: 250–1) construing naturalism thus: “there is no such person as God… or anything like God.” 72 See e.g. Wierenga (1983: 397–9) 73 Hume (1740/2000: 302) 52

Samuel Clarke and John Balguy and such deists as Matthew Tindal,74 a moral rationalism

which posited the existence of abstract moral relations knowable only through intellectual

intuition. In other words, the section is a critique of the historical predecessor to

metaethical nonnaturalism. Indeed, apart from a single paragraph on causal reasoning

about matters of fact, Hume never even addresses anything like a naturalist position. And

at the end of the following section, Hume proudly notes that his sentimentalist account

makes sense of moral evaluation “without looking for any incomprehensible relations and

qualities, which never did exist in nature, nor even in our imagination, by any clear and

distinct conception.”75 While the is-ought paragraph is of course only an addendum to the anti-rationalist section, and while its exact role in Hume’s line of reasoning is not clear,

Hume is clear enough about the view threatened by attention to the is-ought gap: it is

supposed to “let us see, that the distinction of vice and virtue is not founded merely on

the relations of objects, nor is perceiv’d by reason”.76

It is perhaps no surprise that Hume’s point would target naturalists and nonnaturalists

alike—Hume himself is not known for defending the existence of objective moral facts of

any kind.77 But G. E. Moore is the father of nonnaturalism. How is it, then, that the Open

Question Argument could threaten naturalism and nonnaturalism alike?

Consider the argument in isolation. No naturalistic predicate seems capable of closing

the open question, which goes to show that no naturalistic analysis can succeed. But what

about nonnaturalistic predicates? The question “x is ectoplasmic, but is x good?” does not

74 See Russell (2008: ch. 17) 75 Hume (1740/2000: 306) 76 Hume (1740/2000: 302) 77 Though see Norton (1984) 53

seem to be closed. Likewise with “x is loved by God, but is x good?” (unless we build

goodness into our definition of God, a tactic that can be borrowed by naturalist hedonists

simply through defining ‘gpleasure’ as ‘pleasure that is essentially good’). Given his

discussion of divine command metaethics in Ethics,78 Moore himself would presumably

prefer a variant of the Open Question Argument here: “if God does not really exist, might

x still be good?” and “if ectoplasm does not really exist, might x still be good?” are open

questions, which defeats any proposed ectoplasmic or theological analysis of ‘good’. For

ectoplasm and God, then, the Open Question Argument is as effective as ever.

Of course, it is no disproof of nonnaturalism generally to show that two particular kinds of nonnatural facts are not fit for a successful nonnaturalistic analysis: other kinds of nonnatural facts might succeed where these two kinds have failed. But unless we simply stipulate the normativity of the nonnatural fact in question, a tactic which (as be-

fore) can be borrowed by enterprising naturalists, it is hard to see how other nonnatural

facts would fare any better: neither ghosts nor numbers nor universals nor Fregean

nor Meinongian objects seem capable of closing the open question. So, with the

same sort of generalizing approach nonnaturalists use on natural facts, we can conclude

that no nonnatural fact is fit for a successful nonnaturalistic analysis.

Strangely enough, Moore himself would agree with the bulk of the foregoing. Of

course, the canonical Open Question Argument passage occurs as part of a process-of-

elimination argument that ‘good’ is simple and indefinable, with the argument directed

78 Moore (1912/1965: 64–5) 54

against an analysis of ‘good’ in terms of pleasure.79 If the matter rested there it would seem to be an anti-naturalistic argument only. But the argument recurs as Moore prosecutes his case against rival views: thus Moore acknowledges that health is good, but insists “that this must not be taken to be obvious; that it must be regarded as an open question” and adds that “[t]o declare it to be obvious is to suggest the naturalistic fallacy”.80 And the naturalistic fallacy is not just a problem for “Naturalistic Ethics”, but

“Metaphysical Ethics” as well: though he “give[s] it but one name”, both Naturalistic

Ethics and Metaphysical Ethics are “based on the same fallacy”.81

What is this Metaphysical Ethics? For Moore, it is a metaethical view which defines

‘good’ in terms of conventional metaphysics—i.e., metaphysics which conjures up a timeless reality beyond the senses populated by “non-natural existents”82. Moore repudiates this picture, arguing that ‘good’ cannot be defined in terms of any reality or existence whatsoever, be it sensible or supersensible: “To hold that from any proposition asserting ‘Reality is of this nature’ we can infer, or obtain confirmation for, any proposition asserting ‘This is good in itself’ is to commit the naturalistic fallacy”.83 And in the course of critiquing Metaphysical Ethics, he again invokes the Open Question

Argument:

That such a reduction of all propositions to the type of those which assert either that something exists or that something which exists has a certain attribute (which means, that both exist in a certain relation to one another), is erroneous, may easily be seen by reference to the particular class of ethical propositions. For whatever we may have proved to exist, and whatever two existents we may have proved to be necessarily

79 Moore (1903/1993: 66–9) 80 Moore (1903/1993: 95) 81 Moore (1903/1993: 91) 82 Moore (1903/1993: 163) 83 Moore (1903/1993: 164–5) 55

connected with one another, it still remains a distinct and different question whether what thus exists is good; whether either or both of the two existents is so; and whether it is good that they should exist together. To assert the one is plainly and obviously not the same thing as to assert the other. We understand what we mean by asking: Is this, which exists, or necessarily exists, after all, good? and we perceive that we are asking a question which has not been answered. In the face of this direct that the two questions are distinct, no proof that they must be identical can have the slightest value.84

Naturalistic Ethics and Metaphysical Ethics, for Moore, are on an equal footing. Both are

based on the naturalistic fallacy, and both can be disproved by the Open Question

Argument.

But since Moore is himself a nonnaturalist, and if the Open Question Argument

defeats Metaphysical Ethics, how is his view supposed to escape his own argument? The

short answer is that he never says. In fairness, he does distinguish his own view from

Metaphysical Ethics. Eschewing the supersensible reality of conventional metaphysics,

Moore favors a metaphysics of non-natural objects which do not exist in any reality whatsoever. In Moorean metaphysics, numbers are but do not exist: “Two is somehow, although it does not exist.” Likewise, truths (most obviously “‘’ truths”) do not exist. And it is among such “non-natural objects” that we find the object referred to by

‘good’.85 So Moore is still a nonnaturalist: one who rejects the conventional metaphysics

of a supersensible reality of nonnatural existents, but is nevertheless happy to posit

nonnatural objects with being (if not existence).

But if the Open Question Argument is the sort of blunt object that cannot discriminate

between Naturalistic Ethics and Metaphysical Ethics, then it is completely unclear how it

84 Moore (1903/1993: 176) 85 See Moore (1903/1993: 161–3) 56

could discriminate between conventional metaphysics and Moorean metaphysics—

between nonnatural existents and nonnatural nonexistents. Indeed, Moorean metaphysics

seems to provide its own unfortunate example: surely we cannot define ‘good’ in terms of

numbers, and surely the Open Question Argument would defeat any such analysis, even

granting Moore’s suggestion that numbers do not exist but somehow are. Thus we have

at least one example of a kind of nonnatural nonexistent that is not fit for a successful

analysis. And thus if objects belonging to the natural world are all in trouble, and objects

belonging to a hypothetical supersensible reality are all in trouble, it is not at all clear

why nonexistent objects lying beyond any reality would fare any better.

Moore, of course, became deeply dissatisfied with his account of the nonnaturalness

of intrinsic value. He later wrote that he “did not give any tenable explanation of what

[he] meant by saying that ‘good’ was not a natural property”, and characterizes the

account he gave as “utterly silly and preposterous”.86 And his later efforts at characterizing the difference do not seem to help matters much: he distinguishes ‘intrinsic value’ properties from natural properties by contending that the latter “seem to describe

the intrinsic nature of what possesses them in a sense in which predicates of value never

do”, so that while natural properties thus describe the intrinsic nature of a thing and

consequently count as bona fide intrinsic properties, ‘intrinsic value’ properties merely

supervene with unconditional but non-logical necessity on the intrinsic nature of a

thing.87 And this does not explain why nonnatural facts should escape the Open Question

Argument: it is not as if this special supervenience relation is itself sufficient to single out

86 Moore (1942: 582) 87 See Moore (1922: esp. 274–5), and also Moore (1993: 21–27). 57

goodness (“x is F, and F-ness is not an intrinsic property of x but does supervene with un- conditional but non-logical necessity on x’s intrinsic nature, but is x good?” seems as open a question as any), and it is not clear how a predicate satisfying this later account of nonnaturalness could do better than any other kind of predicate at closing the open question.

3.4 Moral Twin Earth against nonnaturalism

The Moral Twin Earth argument is the offspring of a challenge to naturalists: provide a semantic account capable of explaining why it is that moral terms refer to the naturalistic ontology they allegedly do. This challenge seems to embody a legitimate request in any case, but it acquires urgency in the wake of the two classic semantic arguments against naturalism: if they do not disprove naturalism, they at least cast doubt on simple semantic accounts on which there are naturalistic predicates directly synonymous with moral terms, and fuel the question of what semantic account can rescue naturalism. Thus if I am right that the two semantic arguments against naturalism apply to nonnaturalism as well, then the challenge carries over with comparable urgency: nonnaturalists must provide a semantic account capable of explaining why it is that moral terms refer to the nonnaturalistic ontology they allegedly do. Again, there is a legitimate request here in any case, but nonnaturalists who think the two classic semantic arguments make serious trouble for naturalism (as so many do) must acknowledge that they too are in need of rescue. 58

Now the Moral Twin Earth argument applies with greater ease to naturalism than to

nonnaturalism. For in mounting an anti-naturalist Moral Twin Earth argument, one can take advantage of the fact that there are competing ethical theories which pick out different kinds of natural facts as morally relevant. And this means that Moral Twin Earth stories concocted with reference to these competing ethical theories automatically generate pressure to interpret the apparent disagreement as a genuine moral disagreement: after all, partisans of competing ethical theories offer an extremely clear case of genuine moral disagreement. But whereas the moral-natural identifications of any version of metaethical naturalism entail a specific ethical theory (thus if ‘good’ refers to pleasure, then hedonism is true), the moral-nonnatural identifications of metaethical nonnaturalism would seem to be neutral with respect to competing ethical theories (thus if ‘good’ refers to some nonnatural property, any number of ethical theories could be true, depending on the specific supervenience facts that obtain). Since the nonnatural facts merely supervene on the natural facts ethical theories might pick out as morally relevant, without being identified with any of those moral facts, nonnaturalism can avoid getting tied down to any particular ethical theory, and then getting disagreed with by competent moralizers related to different facts.

Nevertheless, I believe a Moral Twin Earth argument can be devised as a threat to nonnaturalism. I find three different approaches promising, and in this section I will briefly outline them, providing each with a working example of how to develop a Moral

Twin Earth story. What distinguishes one approach from another is the nature of the properties it selects as potential rivals to the nonnatural properties favored by the 59

nonnaturalist. Thus the first focuses on nonnatural properties other than normative

properties (e.g., mathematical properties): assuming that Plato’s heaven has more than

normative properties in it, then perhaps one community’s use of moral terms might fasten onto these other properties, with the other community’s use fastening onto the properties favored by the nonnaturalist. The second approach focuses on the diversity of normative properties populating Plato’s heaven: assuming that goodness and badness and being a

reason for and being a reason against are all nonnatural normative properties, one

community’s terms might fasten onto one set of normative properties, and the other

community’s terms might fasten onto another set. Finally, the third approach focuses on

the natural properties people might find morally significant: given a reference-fixing

relation incapable of discriminating between nonnatural and natural properties, one

community might fasten onto certain nonnatural properties, and the other community’s

terms might fasten onto certain natural properties. Of course, just given this brief de-

scription, I do not expect it to be obvious how to generate Moral Twin Earth stories with

these three approaches. But the following examples should help to spell it out clearly.

To illustrate the first approach, I’ll take the simple example of nonnatural mathemati-

cal properties. Let’s suppose that the nonnaturalist has a semantic account of moral terms

and concepts that centers on some reference-fixing relation (e.g., the causal regulation of

Boyd’s account), and suppose that our use of moral terms is thus related to certain nonnatural properties, so that the nonnaturalist considers these properties to be what our moral terms refer to. Now, imagine another community that uses lookalike terms, and does so in apparently evaluative ways, so that these terms are related to certain other 60

nonnatural properties: viz., mathematical properties. If the proposed semantic account is

adequate, then their terms should refer to these mathematical properties, and any apparent

intercommunity disagreement can be dismissed as mere pseudo-disagreement. And there-

fore, if such disagreement seems like bona fide moral disagreement, then the semantic

account is inadequate.

Now this example is not very convincing. First, it is open to serious doubt whether

lookalike moral terms could be used in apparently evaluative ways so that the terms are

somehow related to mathematical properties. Mathematical properties, after all, seem to

be instantiated by different sorts of things than moral properties, and it’s not implausible

to think that this—the sorts of things instantiated by moral properties—would be an

essential feature of any would-be moral terms used in apparently evaluative ways.

Second, and not unrelatedly, there is not much pressure to interpret intercommunity disagreement as bona fide moral disagreement: even if you could somehow ‘evaluate’ your way into referring to mathematical properties, it is open to doubt whether you would thereby be engaging in bona fide moral evaluation.

But I think more convincing examples might be found. Consider a supernatural being, and the nonnatural properties having to do with it and its will. If our moral terms bear the allegedly reference-fixing relation to the nonnaturalist’s favorite nonnatural properties, but another community’s apparently evaluative lookalike moral terms bear the same relation to these theological properties, it looks like we have a plausible Moral Twin

Earth story on our hands. It is all too plausible that a community might tie evaluative practices to the will of a supernatural being, and not terribly plausible to dismiss any 61

intercommunity disagreement as mere pseudo-disagreement.88 If their supernatural being

is a consequentialist (taking the point of view of the universe), and our nonnatural

properties supervene in a deontological kind of way, then our respective evaluative

practices will be pitted against each other in a way best interpreted as genuine moral

disagreement. And if this feat can be pulled off for the reference-fixing relation of any se-

mantic account, then nonnaturalists are left without an adequate moral semantics, and left

without an answer to the challenge posed by the classic semantic arguments. Call this the

theological case.

Now I’ll walk through a case for the second approach. Suppose our moral terms

‘right’ and ‘wrong’ refer to certain nonnatural properties in virtue of some reference-

fixing relation. Now suppose another community has things reversed: their term ‘right’

relates to what our term ‘wrong’ relates to, and vice versa. As a result, we are in what ap-

pears to be extreme moral disagreement (e.g., we call animal torture ‘wrong’, they call it

‘right’): can this be plausibly dismissed as mere pseudo-disagreement?

Perhaps it can be dismissed. After all, if a would-be moral standard departs too far

from morality as we know it, it ceases to be a ‘moral’ standard at all: consider the anti-

utilitarian view that the right action is the one which maximizes the overall balance of

pain over pleasure. It arguably belongs to the very concept of morality that certain

values—perhaps well-being and impartiality—be incorporated, and thus a completely

reversed evaluative system scarcely qualifies as a system of morality. Thus perhaps we

88 Cf. Adams (1999) on how a divine command theory of moral obligation can be grounded in a Boyd-like semantic theory of ethical terms. 62

can say that there is no genuine moral disagreement between us and the reversed community, simply because they have no genuine morality at all.89

But I think this is too quick. Even if the concept of morality has certain values built into it, the broader concept of normativity seems perfectly neutral. Thus even if anti- utilitarianism does not count as an ethical theory, it can still count as a theory of correct action, and an anti-utilitarian can insist that those who disregard its requirements are act- ing against the one true authoritative and binding practical system. And this means that the Moral Twin Earth story can simply be recast as a Normative Twin Earth story: if our normative terms ‘is a reason for’ and ‘is a reason against’ refer to certain nonnatural properties in virtue of some reference-fixing relation, it seems there could be a community who has things reversed, and the resulting apparent disagreement seems best interpreted as a genuine normative disagreement. This casts doubt on any semantic account centering on that alleged reference-fixing relation, and the problem would seem to generalize. Call this the intranormative case,

Finally, I will give an example of the third approach. On this approach, another community is related to natural properties in the same way we are related to nonnatural properties. Thus we might be related to nonnatural properties that supervene in a deontological kind of way, and they might be related to a set of natural properties tied to a consequentialist theory. It seems most plausible to interpret the resulting apparent disagreement as a genuine moral disagreement, which in turn casts doubt on any semantic account centered on the relation in question. Call this the natural case.

89 Cf. Foot (1958) 63

I should note that this last approach has a potential weakness: nonnaturalists may be

able to identify a reference-fixing relation that takes only terms and nonnatural properties

as relata, thereby excluding natural properties. For example, it may be that our moral

terms refer to nonnatural properties in virtue of a relation of intellectual intuitive

grasping, in which case it is doubtful that we could be thus related to natural properties.

Given such a relation, the third approach to generating Moral Twin Earth stories couldn’t

even get started, since the other community would be ruled out as impossible. But of

course, it is up to nonnaturalists to spell out the reference-fixing relation in virtue of which our moral terms refer to nonnatural properties, and this third approach at least serves to make it clear that the relation will have to be a rather discriminating one.

3.5 Nonnormative Conceptual Role Semantics

David Enoch is one nonnaturalist who has faced up to this problem—or, rather, something in its vicinity. What Enoch addresses is what I earlier called a “legitimate request”: nonnaturalists must provide a semantic account explaining why it is that moral terms and concepts refer to the nonnaturalistic ontology they allegedly do. He sees his view (“robust realism”) as especially vulnerable to this problem due to his conception of nonnatural normative properties as “causally inert and response-independent”. Given such a conception of the properties in question, it might seem that nonnaturalists have nothing satisfying to say, as if normative terms refer by “some kind of referential magic”.90 As Enoch writes:

90 Enoch (forthcoming: 365) 64

[H]ere, in our world, we have all these words and of use; and all the way out there in Plato’s heaven, there is a plethora of normative (and presumably other) properties and relations. In virtue of what, then, does the word “good” manage to find the property goodness and not some other property? This seems like a question that robust realists should at least have something to say about.91

Obviously, I agree. And since Enoch takes the problem seriously, I think a close

examination of his proposed solution is important to a defense of my attempt to press this

problem against nonnaturalists.

Enoch’s initial response to the problem is to look for safety in numbers. If Platonists

about moral properties have a problem, then Platonists about numbers and Platonists

about logical connectives have the same problem. And if it is plausible that these other

Platonists might have a solution, then this serves as a reasonable defense of nonnatura- listic metaethics.92

But Enoch does not stop at the level of generalities. He proposes a solution that might

work for both Platonists about logical connectives and Platonists about normativity:

conceptual role semantics (CRS). The basic idea of CRS is that a concept is individuated

by its role in cognition. That is, there is a limited selection of causal relations the concept

bears to other concepts and perhaps to input and output states—causal relations having to

do with inferences especially—and it is these relations that make it the particular concept

it is. Now, CRS is a natural fit for logical connectives. For example, what makes the con-

cept of logical conjunction the particular concept it is? Plausibly enough, it is something

about the way it enters into such inferences as ‘p and q, therefore p’ and ‘p, q, therefore p

and q’. That is, if making those inferences with this concept is especially likely or

91 Enoch (forthcoming: 366) 92 Enoch (forthcoming: 367) 65

especially appropriate, then we must be dealing with the concept of conjunction. And if

CRS applies to logical connectives, then perhaps CRS can also apply to normative terms

and concepts.93

Now, as I’ve just presented it, this line of thought is much too quick. For CRS is

primarily a theory of concept individuation, not a theory of reference.94 And if the theory

is restricted to narrowly psychological states and the causal-cum-inferential relations

between them, then it is hard to see how reference to external objects (much less nonna-

tural objects in Plato’s heaven) could be thus secured. After all, the ‘water’ concepts of

Oscar and Twin Oscar are alike in their narrowly psychological conceptual role, and yet

they refer to different substances. Typically, therefore, CRS must account for reference

by bringing in external mind-world connections—paradigmatically, causal connections.

In other words, CRS accounts for reference in much the same way Boyd accounted for reference.

This is commonly done in one of two ways. ‘Long-arm’ views like those of Gilbert

Harman build the external connections into the conceptual role: thus the ‘water’ concepts of Oscar and Twin Oscar have different conceptual roles, because they are connected to different substances in the world. ‘Short-arm’ two-factor views like those of Ned Block leave conceptual role narrow, but then conjoin it with an external factor which accounts for reference: thus the ‘water’ concepts of Oscar and Twin Oscar are alike in their conceptual role factor, and even alike in the nature of their referential factor (e.g., both being causally determined), but differing in their reference. Of course, for Block,

93 Enoch (forthcoming: 368–9) 94 See Block (1986), Fodor and Lepore (1992: ch. 6) 66

conceptual role is not completely irrelevant to reference: the conceptual role factor is

what determines the nature of the referential factor, and thus it is no coincidence that the

‘water’ concepts of Oscar and Twin Oscar are alike in the nature of their referential

factor.95 But the upshot is that, on either version of CRS, reference itself is determined by

external connections that go beyond the narrowly psychological.

This means that, even if CRS is a plausible theory for logical connectives when it

comes to concept individuation, it is not a promising theory of reference for Platonism

about logical connectives. Perhaps the concept of logical conjunction can be individuated

by some limited selection of the causal-cum-inferential relations it enters into, but that wouldn’t begin to explain why it refers to whichever piece of nonnaturalistic ontology it does. And the common way of supplementing the theory—by incorporating external mind-world connections—is a poor fit for Platonism, for it is not obvious how to implement the suggestion that an abstract truth-functional connective causally interacts with a particular concept of mine. And if CRS cannot account for the reference of logical connectives, why should we expect it to account for the reference of normative terms?

But for us the most important problem with a CRS-based nonnaturalism is its vulnerability to Moral Twin Earth. We have seen how Boyd-style causal theories of reference would be vulnerable to Moral Twin Earth, regardless of whether the alleged moral properties are natural or nonnatural. But, as mentioned above, when CRS theorists account for reference, they do so by incorporating Boyd-style theories of reference. Thus even if CRS has its own distinctive way of individuating normative concepts, it has an

95 Block (1986: 635–9, 643–4) 67

all-too-familiar way of fixing reference, which means Moral Twin Earth will still pose a threat.

Recalling the three approaches briefly developed in the previous section, it seems that all three (and their corresponding cases) will apply. The theological case, where the nonnaturalist’s favored properties compete with other nonnatural properties having to do with the will of a supernatural being, will apply: both kinds of properties seem equally capable of entering into Boyd-style reference-fixing relations. The intranormative case,

where the nonnaturalist’s favored properties compete with a complete reversal there of,

will apply: here, since the very same properties are at stake, it would seem impossible to deny that both communities might bear the same allegedly reference-fixing relation. And even the natural case, where the nonnaturalist’s favored properties compete with a set of natural properties tied to a ethical theory denied by the nonnaturalist, will apply: for it is not credible to deny that Boyd-style reference-fixing relations can have natural properties as relata.

Consequently, CRS as typically understood cannot escape Moral Twin Earth. If it is merely a theory of concept individuation, then it simply does not answer the question of reference: nonnaturalists are left with no account of why moral terms refer to the nonnatural ontology they do. But if CRS is made to account for reference in the typical way, then the resulting CRS-based nonnaturalism is threatened by Moral Twin Earth, for the same reason that theories based on Boyd-style theories of reference are threatened by

Moral Twin Earth. 68

Now, Enoch does not appear to recognize this problem. He takes it as unproblematic

that CRS can account for the reference of logical connectives, and therefore normative

terms as well. But this means his version of CRS is significantly different from that of

Harman and Block. For as I’ll now explain, Enoch’s version of CRS uses normative

reference-fixing relations.

3.6 Normative Conceptual Role Semantics

As we have seen, CRS individuates a concept in terms of certain especially likely or

appropriate inferences the concept enters into. In saying ‘likely or appropriate’, I papered

over a division among CRS theorists: some individuate concepts with inferences selected

by some natural criteron (e.g., likelihood, what inferences the subject is disposed to per-

form), whereas others select the concept-individuating inferences by some normative

criterion (e.g., appropriateness, what inferences the subject should perform).96 Enoch’s

CRS takes the normative side of this divide. Thus assuming the concept of logical

conjunction is individuated by inferences like ‘p and q, therefore p’, it is because these inferences are appropriate ones for a subject with that concept to perform.97

But this normativity shows up only when it comes to concept individuation. It is an

additional bit of normative machinery that makes Enoch’s CRS so significantly different

from the typical version: the reference of a concept, for Enoch, is what makes these

concept-individuating inferences appropriate. For example, assuming there are certain

inferences especially appropriate for the concept of logical conjunction (and therefore

96 See Whiting (2009) for this distinction. 97 Enoch (forthcoming: 368) 69

responsible for the individuation of this concept), Enoch would have the concept

referring to whichever Platonic object it is that makes those concept-individuating inferences appropriate. As he explains (using the example of the material conditional):

And when we have these rules in mind, it seems very plausible to say that there is a relation between the fact that “if… then…” refers to material implication and the fact that we use “if… then…” locutions roughly according to the rules of implication- introduction and implication-elimination (or modus ponens). Indeed, it seems plausible that our words “if… then…” manage to refer to material implication precisely in virtue of the fact that such reference will render our use of the relevant introduction and elimination rules rational, or valid, or justified, or truth-preserving, or in some other way okay. This is the kind of view I will be talking about when using the term “conceptual role semantics” (or CRS)—a view according to which what explains or grounds the fact that a certain linguistic expression has a certain referent is that this is the referent that will render the use of the linguistic expression largely okay.98

Thus Enoch is attempting to secure reference for logical connectives without appeal to

the external reference-fixing connections typically incorporated into CRS. These

connections are in effect replaced with normative relations: the reference is what it is

because it is related to the concept by the normative relation making the inferences that

individuate this concept appropriate. And thus, assuming that our normative concepts are

individuated by certain especially appropriate inferences they enter into, these concepts

refer to whichever Platonic object makes these inferences appropriate.

This is the heart of Enoch’s account. But he has one further emendation, which has

him turning to the version of CRS developed by Ralph Wedgwood. Suppose that the

concept-individuating inferences for a normative concept were all intranormative: for

example, the inference from ‘X would bring about the most good’ to ‘X is right’. In that

case, it is unclear that these inferences alone would be sufficient to distinguish between

98 Enoch (forthcoming: 368) 70

different normative concepts: as Enoch notes, “this strategy seems vulnerable to ‘inverted

normative spectrum’ counterexamples” where the allegedly concept-individuating

inferences for two perfectly opposed normative concepts (e.g., good and bad) come out

indistinguishable. So we may need more than mere intranormative inferences in order to

individuate normative concepts.

Thus Enoch brings in inferences that end in action: i.e., practical inferences from

(say) ought-beliefs to the corresponding actions. The idea is that perfectly opposed normative concepts can be distinguished by the different actions terminating certain especially appropriate practical inferences. And, happily enough, Wedgwood has a well-

developed account of CRS for normative concepts that explicitly incorporates practical inferences. For a practical ‘ought’, on Wedgwood’s account, is individuated by the

conceptual role given by the following rule of rational commitment: “[a]cceptance of the

first-person proposition ‘O(p)’—where ‘t’ refers to sometime in the present or near

future—commits one to making p part of one's ideal plan about what to do at t.” In other

words, a normative judgment like “I ought to buy shoes this week” rationally commits

me to making buying shoes part of my ideal plan for this week, and it is this that

individuates the ought-concept occurring in that normative judgment.99

As this point we have all we need to summarize Enoch’s proposed semantic account.

A normative concept is individuated by certain especially appropriate inferences it enters

into, including practical inferences that go beyond mere intranormative inferences. And

the reference of this normative concept is just whichever Platonic object it is that makes

99 Enoch (forthcoming: 369–70) 71

these inferences appropriate. This explains why a normative concept refers to the

nonnaturalistic ontology it does: because it is that piece of ontology that makes the

concept-individuating inferences appropriate.

But it is worth pursuing Wedgwood’s semantic account further. First, it is

Wedgwood’s account that Enoch hopes to tie his semantic wagon to. Second,

Wedgwood’s account offers a particularly clear-cut example of the sort of normative

CRS under investigation.

We have seen how Wedgwood would individuate normative concepts: in terms of a

conceptual role given by a rule of rational commitment. Thus it is a rule of rational

commitment that ought-judgments rationally commit you to corresponding adjustments

in your practical plans. But what is this ‘rational commitment’? For Wedgwood, it is just

a convenient abbreviation for talk of the rationality of certain mental states in relation to

each other: ‘S1 rationally commits you to S2’ is equivalent to ‘if S1 is rational, then it would be irrational of you not to be in S2’. And thus Wedgwood would say that if your

acceptance of the practical normative judgment “I ought to buy shoes this week” (the first

mental state) is rational, then it would be irrational of you not to make buying shoes part

of your ideal plan for this week (the second mental state).100 Concept individuation is

based on the rationality of mental states. And Wedgwood explicitly argues that this

‘rationality’ is an “essentially normative concep[t]”: it corresponds, he says, “to an in-

formation-relative ‘ought’”.101

100 Wedgwood (2007: 84) 101 Wedgwood (2007: 156) 72

But what about reference? Here Wedgwood shifts from rationality to ‘correctness’: the reference of a normative concept is what makes the rule of rational commitment (the same rule that played into individuation of the concept) a valid one, where validity is just the familiar condition that if the first mental state is correct, then it would be incorrect to be in the opposite of the second mental state. Thus suppose that the rule in question for the shoe-buying example is a valid one, and suppose also that I am correct in accepting “I ought to buy shoes this week”. In that case, it would be incorrect to make not-shoe- buying part of my ideal plan for this week, and the reference of the practical ‘ought’ in question is what makes this true. Wedgwood characterizes this reference as a property of the proposition occurring in the normative judgment: e.g., a property of the proposition I am buying shoes this week.102 Again, this ‘correctness’ is an essentially normative concept, but this time it corresponds to “a more objective [and less information-relative]

‘ought’”.103

For more on ‘correctness’, it is instructive to work through the example of belief.

Now, Wedgwood would contend that the correctness of a belief consists in its truth. But why truth? Because of certain distinctive normative standards essentially tied to this type of mental state: i.e., it is essential to the state-type belief that beliefs “are guided and re- gulated by those standards”. But there is something underlying these standards:

Conforming to these standards is not just important for its own sake, however; there is some further point or purpose in conforming to these standards. It is when a mental state achieves this further point or purpose that it counts as a correct mental state.

102 Wedgwood (2007: 99–102) 103 Wedgwood (2007: 156) 73

And when it comes to belief, the point or purpose underlying its distinctive normative

standards is a matter of truth: doxastic standards are worth conforming to because of their relation to truth. Thus it is when a belief is true that it achieves this point or purpose, which means it is when a belief is true that it counts as correct.104

Wedgwood thinks this holds in general. That is, the correctness of a mental state

consists in whatever a state must achieve in order to achieve the further point or purpose

underlying the normative standards essential to that type of state. What, then, about the

mental states belonging to practical reasoning? When does a plan or an intention count as

correct? Wedgwood suggests (mainly for purposes of illustration) that it is a matter of

choiceworthiness: the normative standards essential to these mental states are oriented towards the further point or purpose of choiceworthiness, and thus it is when a plan or intention is choiceworthy that it counts as correct.

And so to account for the reference of normative terms, Wedgwood relies on a

number of normative properties and relations. Certain mental states are objectively

correct, where this correctness consists in achieving some further point or purpose

underlying the normative standards essential to that type of mental state. And the

reference of a practical ‘ought’ is fixed by the normative relation making it the case that

the relevant rule of rational commitment is a valid one. That is, there is some property of the proposition occurring in one’s normative judgment that makes the relevant rule of rational commitment a valid rule, and it is this property that counts as the reference of the

104 Wedgwood (2007: 100–01) 74

practical ‘ought’ in question. And this is what explains why normative concepts refer to

the ontology they do.

3.7 Two objections

Thus far we have seen that nonnaturalism is vulnerable to the same semantic arguments it

frequently wields against naturalism, that conventional nonnormative CRS does not

escape this problem, and that nonnaturalists have recently developed explicitly normative

versions of CRS that aim to answer the question of why normative concepts refer to the

nonnaturalistic ontology they do. In this section, I will present two separate objections

intended to show that not even normative CRS can escape the problem.

Objection 1: Vicious circularity

One attraction of normative CRS is that it would seem to elude the grasp of Moral Twin

Earth. After all, its reference-fixing relations are not causal, or even natural, but instead nonnatural and normative, which thoroughly distinguishes it from the Boyd-style theories of reference that get metaethicists into trouble. Wedgwood states this explicitly:

It is because my account is not “naturalistic’” in this strong sense that it can escape the dilemma that Terry Horgan and Mark Timmons have deployed against all forms of “naturalistic moral realism”... This argument is plausible only if it is assumed that this relation R is a purely natural relation, and not itself a normative relation. But in my account, the relation in virtue of which these uses of ‘ought’ refer to the relevant property is itself a normative relation. In my account, this relation is the following: first, these uses of ‘ought’ express a concept whose essential conceptual role consists in the way in which certain beliefs involving this concept commit one to incorporating a certain proposition into one’s plans; and secondly, this concept refers to the property that makes this sort of practical reasoning valid—that is, the property of a proposition p that makes it correct for one to incorporate the proposition p into one’s 75

plans about what to do at t, and incorrect to incorporate the negation of p into one’s plans about what to do at t.105

But, as I’ll now argue, the use of normative reference-fixing relations turns out to bring problems of its own.

Enoch’s and Wedgwood’s semantic accounts are supposed to tell us why normative terms and concepts refer to what they do. But these accounts help themselves to normative truths about the correctness of mental states and what makes inferences appropriate or certain rules of rational commitment valid. But if the normative truths we began investigating need a semantic account, then it would seem that these additional normative truths need a semantic account as well. Now, if these truths call for a different kind of semantic account, then we are left with a seriously incomplete account of normative semantics. But if the same kind of semantic account is used, then they will only presuppose further normative truths, so that the need for a semantic account is never satisfied. Either way, we are left without a general semantic account for normative terms and concepts.

To see this, return to Enoch’s account. Why does the normative term “ought to” refer to the Platonic object it does? It is because this object is what makes certain key inferences appropriate: e.g., the inference “from the judgment that we have the strongest undefeated reason to Φ to an intention to Φ”.106 And this semantic account is supposed to make sense of such everyday normative truths as “I ought to take my cat to the vet”. But the account performs its task only by helping itself to such normative truths as “this

105 Wedgwood (2006: 141) 106 Enoch (forthcoming: 370): despite what this example might suggest, Enoch rejects motivational judgment internalism. 76

Platonic object is what makes this inference appropriate”. And these normative truths are

just as much in need of a semantic account as the more everyday ones. If the normative

terms occurring in these truths are to be provided with a completely different kind of

semantic account, then Enoch’s normative semantics is as yet incomplete: normative

conceptual role semantics offers only part of the story for the semantics of normative

terms. But if we return to the very same kind of semantic account, then we are not getting

anywhere.

Now, Wedgwood acknowledges something like this problem. He puts it in terms of

circularity:

Some philosophers will worry that if [correctness] is a normative concept, then my account is viciously circular. But in my account the notion of a correct mental state is used only in extensional contexts in the metalanguage; it is not used in specifying the contents of anyone’s mental states or utterances. Thus, the appearance of a normative term in the metalanguage of our account of the nature of a normative concept is no more viciously circular than the appearance of a logical constant (like ‘not’) in our account of the nature of a logical concept.107

And elsewhere:

Thus, in giving an account of what it is for a term to refer to Socrates, or to cows, it is perfectly permissible to talk about Socrates, or about cows. For example, it would be perfectly permissible to say that the meaning of the term ‘Socrates’ partially consists in the fact that the term stands in some relation to Socrates himself, or that the meaning of ‘cow’ partially consists in the fact that the term stands in some relation to actual cows.108

I agree with Wedgwood that there is no vicious circularity in the mere fact that a term

occurs in a semantic account for that very term. And indeed we might refer to the

individual(s) in question when stating which relation between the individual(s) and the

107 Wedgwood (2007: 102) 108 Wedgwood (2006: 143) 77

term serves to fix the term’s reference. But problems do arise when the term’s having the

reference it does is essential to the reference-fixing relation itself.

For example, it is relatively unproblematic to say that the term ‘Socrates’ refers to

whichever individual it is that is most closely causally connected to people saying the

term ‘Socrates’—or (more or less equivalently) that ‘Socrates’ refers to what it does

because of certain causal relations between the term ‘Socrates’ and the individual

Socrates. But consider the account which says that the term ‘Socrates’ refers to

whomever Socrates is pointing to when he says the term. This is seriously problematic: to

identify the term’s reference, it is necessary to identify the individual bearing the

proposed reference-fixing relation to the term, but it is impossible to identify that

individual without first identifying the term’s reference. Thus if there are several distinct

individuals who point to themselves when they say ‘Socrates’, the proposed semantic

account is powerless to determine which individual the term actually refers to. Such an

account is subject to a vicious circularity, despite the fact that the term ‘Socrates’ occurs

only in the metalanguage.

Likewise with normative terms. If ‘good’ refers to whichever property it is that is

most closely causally connected to people saying the term ‘good’, the problem does not

arise. But if we restrict it to good people saying the term ‘good’ (so that ‘good’ refers to

whichever property is most closely causally connected to good people saying the term

‘good’), the problem is clear. It is impossible to identify the property via its reference- fixing relation without first identifying the people who instantiate the property, and it is 78

impossible to identify which people instantiate the property without first identifying

which property it is they are supposed to instantiate. Again, this is vicious circularity.

Now recall Wedgwood’s account. He does not put the term’s having the reference it

does into the reference-fixing relation for that term, as in the previous cases of circularity.

But he does put other normative terms’ having the reference they do into his reference- fixing relation. Thus to identify the reference of the practical ‘ought’ via its reference- fixing relation, it is necessary to identify the property of the proposition occurring in the relevant normative judgment that makes the relevant rule of rational commitment valid.

But it is impossible to identify this property without first identifying the property that makes it the case that if the normative judgment is correct, then it would be incorrect to be in the opposite of the practical plan adjustment state. And since correctness is a matter of achieving the point or purpose underlying the normative standards essentially tied to the type of mental state in question, it is then necessary to identify these standards and that point or purpose. But here the semantic account runs out: it does nothing to specify the reference-fixing relation that explains why the normative term ‘the point or purpose of practical reasoning’ refers to what it does. Indeed, it is not clear how to get

Wedgwood’s existing semantic account to perform this explanatory task.

But suppose that the account could be applied to this last normative term: could it provide a satisfactory explanation of normative semantics? No: since no normative term could have its reference identified without first identifying the point or purpose of practical reasoning (this being the bottom-line of the semantic account), it follows that not even the particular normative term ‘the point or purpose of practical reasoning’ could 79

have its reference identified without first identifying the point or purpose of practical

reasoning. And this would be a clear case of vicious circularity.

This means that Wedgwood’s account is not straightforwardly circular: it is not like

the account on which ‘Socrates’ refers to whomever Socrates is pointing to when he says

the term. But it is in a similar jam: it is more like the account on which ‘Socrates’ refers

to whomever Plato is pointing to when he says the term, and ‘Plato’ refers to whomever

Aristotle is pointing to when he says the term, and it is not clear why ‘’ refers to

what it does. Or (more closely) the account on which ‘good’ refers to the property right- acting people most prefer, and ‘right’ refers to the property that makes an action rational, and it is not clear why ‘rational’ refers to what it does, or how to extend the account of

‘good’ to explain ‘rational’ without courting circularity. A semantic account of normative terms will always be incomplete as long as it relies on normative terms whose reference is essential to the reference-fixing relations proposed by the account.

Objection 2: Supervenience and Moral Twin Earth

My first objection was that the normativity that promised to help normative CRS escape

Moral Twin Earth ends up only condemning the view to vicious circularity. But I will now argue that this promise is itself hollow: normative reference-fixing relations are as prone to Moral Twin Earth as nonnormative reference-fixing relations.

To see why, it is useful to consider a version of normative CRS on which all the normative properties and relations can be naturalized. Thus the validity of a rule might consist in the subject’s being disposed to perform it, and the reference-fixing normative 80

relation of making these inferences appropriate might consist in certain causal relations between the subject and something in her environment. For example, one inference that helps individuate the ‘water’ concept may be the practical inference from ‘I am on fire’ to jumping in a pool of what falls under that concept: the appropriateness of this inference

(and others like it) might consist in the subject’s being disposed to perform it, and this appropriateness-making disposition (and others like it) came about through causal

109 interaction with H2O, so that this is what fixes the reference of ‘water’.

I hope it is clear that normative CRS simply collapses into a more traditional CRS when naturalized. And, just as a more traditional CRS is vulnerable to Moral Twin Earth, likewise for naturalized normative CRS: the reference-fixing relation may be normative, but if it is reducible to some naturalistic relation, then its normativity does nothing to rescue it from familiar Moral Twin Earth stories built around naturalistic reference-fixing relations.

Now, no nonnaturalists worth their salt would agree that the normative reference- fixing relations could be naturalized. If all normativity is nonnatural, then the normativity at play in semantics is nonnatural as well. So they have nothing to fear from the foregoing point. But even if nonnaturalists deny that normative reference-fixing relations are reducible to naturalistic relations, still they must acknowledge that these relations supervene upon naturalistic relations. And the problem is that this supervenience is enough to invite Moral Twin Earth stories built around naturalistic reference-fixing relations.

109 I take this example from Block (1986: 623–4). 81

To see why, suppose that the normative reference-fixing relation for ‘good’

supervenes upon some naturalistic causal relation. This means that it is necessary that any

individuals related by the naturalistic relation are related by the normative relation. This obviously means that if a term and an individual are related by the naturalistic relation, then they are related by the normative relation, which in turn means that the term refers to the individual. Consequently, the naturalistic relation is itself a reference-fixing relation: for, if a reference-fixing relation supervenes upon another relation, then that other relation is also a reference-fixing relation.

Returning to Enoch’s account, if a normative concept refers to the Platonic object it

does due to that object’s standing in the normative relation making the inferences that

individuate this concept appropriate, then there must be some underlying naturalistic

relation holding between the object and the concept. That much follows from

supervenience of the normative on the natural. Likewise, it must be that this naturalistic

relation is itself sufficient for the normative relation, and consequently sufficient for

reference. That too follows from supervenience of the normative on the natural. And thus

this underlying naturalistic relation is itself a reference-fixing relation, and we are back on Moral Twin Earth.

This means that all three approaches (and their corresponding cases) will apply: the theological case, the intranormative case, and the natural case. For reasons that will become clear shortly, I’ll walk through the natural case. Suppose we are related to the nonnaturalist’s favorite nonnatural properties by the nonnaturalist’s favorite normative

reference-fixing relation. This relation supervenes upon some naturalistic relation, which 82

means we are also related to these properties by the naturalistic relation. So suppose there

is another community related to a conflicting set of natural properties by this naturalistic

relation, and therefore by the normative relation as well—by ‘conflicting’, I merely mean

that we might be related to nonnatural properties that supervene in a deontological kind

of way, and they might be related to a set of natural properties tied to a consequentialist theory. It seems most plausible to interpret the resulting apparent disagreement as a genuine moral disagreement. And that means the normative relation in question is not plausibly construed as a reference-fixing relation.

Now, the objection I’ve just presented can divided into two stages. First, show how supervenience serves to domesticate normative reference-fixing relations by tying them to nonnormative reference-fixing relations. Second, run the Moral Twin Earth argument on the underlying nonnormative reference-fixing relations. And I consider the first stage to be the most crucial stage: if I am right that normative reference-fixing relations are tied by supervenience to nonnormative reference-fixing relations, then the second stage

should go through without significant trouble. By contrast, if the normative relations

floated free of naturalistic relations, and if they were incapable of linking different

communities to different properties, then they would be invulnerable to the Moral Twin

Earth argument.

And so I am encouraged by a recent exchange between Wedgwood and James

Lenman, wherein Lenman poses an objection similar to the natural case of Moral Twin

Earth presented above. In his example, another community responds to natural properties in a way reflecting their passion for torturing helpless animals, and yet it is most plausible 83

to say that they have the same normative concepts we do, and that we genuinely disagree

with them.110 Now, Lenman does not attempt to show that the reference-fixing relation of

Wedgwood’s story is capable of linking different communities to different properties, or

(in particular) that the other community really could stand in that relation to the natural properties they respond to in their practical lives. He simply presupposes it in the course of presenting the objection. But in reply, Wedgwood does not challenge this presupposition. Instead, he insists that what the other community admires really is admirable, since their admiration is directed at such qualities as “ingenuity, steadfast dedication, and tireless zeal”, which remain admirable “even when they are deployed in the service of less than noble ends”.111 Now, I think Lenman’s story can be modified to get around this reply: presumably we can imagine the community admiring sadistic pleasure as such. But what strikes me is that Wedgwood never denies that the relations he takes as central to normative semantics can link us to natural properties quite different from the nonnatural ones he thinks we are in fact linked to. And so if Wedgwood really intends to make this concession, then I do not have to work for the first stage of argument by appealing to the supervenience of the normative on the natural. I get the first stage for free, at which point a Moral Twin Earth problem seems almost unavoidable.

3.8 Conclusion

I conclude that textbook nonnaturalism fares no better than its naturalistic rival at accounting for moral semantics. The classic semantic arguments against naturalism apply

110 Lenman (2010) 111 Wedgwood (2010) 84

equally well for naturalistic and nonnaturalistic ontology, and the Moral Twin Earth

argument can be applied to nonnaturalistic ontology as well. Conceptual role semantics

incorporating familiar naturalistic reference-fixing relations cannot avoid the problem, and the recent work of Enoch and Wedgwood to merge conceptual role semantics with

normative reference-fixing relations succumbs to the two objections I’ve outlined. A major source of nonnaturalism’s appeal is that it offers a way to make sense of moral objectivity while avoiding the semantic problems that plague naturalism. I hope to have

shown that, on this point, nonnaturalism has nothing to offer.

85

CHAPTER 4: QUIETISM

4.1 Introduction

Moral objectivity is an aspiration of many metaethical theories. Naturalist and nonnaturalist realists hope to secure it by finding robust and stance-independent moral properties for moral terms and concepts to refer to. Constructivists do not always seek moral objectivity, but when they do, they turn to certain actual or hypothetical individuals, whose evaluative stances and perspectives are argued to be both definitive of morality and associated with a single invariant moral code. Finally, many ‘quasi-realist’ expressivists hope to make sense of moral objectivity without appeal to any (realist or constructivist) truth-makers, construing morality’s claim to objectivity as an expression of practical attitudes marked by a certain inflexible and socially assertive character.

This suggests two broad ways of understanding moral objectivity. One starts with objective truth-makers: either the stance-independent properties espoused by realists or the stance-dependent (but still suitably invariant) properties espoused by objectivist constructivists. “Torturing animals is wrong” comes out as objectively true because torturing animals has a property of wrongness whose patterns of instantiation are independent of anyone’s subjective beliefs or attitudes. The other looks to the psychology of objectivity: if my opposition to torturing animals is so deep that I also oppose anyone who does not share it (and even anyone who does not share that opposition), so deep that

I hold it regardless of public opinion, carrying the opposition over to the consideration of hypothetical scenarios where everyone favors it (even me!), then it is not implausible to 86

see my mind as one gripped by the thought that it is objectively wrong to torture animals.

Instead of seeking an informative account of what it is for torture to be objectively wrong, we seek an informative account of what it is to think (or sincerely say) that torture is objectively wrong.

The two approaches are, of course, not mutually exclusive. Indeed, success at the first approach will shed much light on the second approach. For, if we had a good account of objective moral truth-makers, then presumably we could understand the psychology of moral objectivity primarily in terms of descriptive beliefs about those truth-makers.

Indeed, it is not clear what moral truth-makers could be, if not suitable truth-conditions for descriptive moral beliefs. If, however, we are convinced that the first approach is misguided, and that objective moral truths cannot be understood in terms of objective truth-makers, then we would seem to need something like the second approach, lest we have no informative account of anything to do with moral objectivity.

But metaethical quietists are different: they utterly reject the first approach, and are not happy with the expressivism associated with the second approach. They yield to no one in their defense of objective moral truths, but they deny that these truths have anything to do with any realist or constructivist truth-makers—moral objectivity is not to be understood metaphysically. Nor do they follow expressivists in jettisoning the assumption that moral judgments are descriptive beliefs—indeed, as we will see, some quietists see expressivism as a fundamentally skeptical view that cannot be reconciled with moral objectivity. Quietists want moral objectivity on their own terms. 87

As quietism has gradually emerged onto the metaethical scene, it has spent most of

its time targeting the realists and constructivists who seek truth-makers for objective

moral truth. No such metaphysical backing is needed for objectivity, at least when it

comes to morality. Thus Nagel writes that “[i]t is important not to associate [my view]

with an inappropriate metaphysical picture: it is not a form of Platonism” and that “[t]he

picture I associate with [my view] is not that of an extra set of properties of things and

events in the world”.112 And Putnam pours scorn on “the idea that if a statement is true

then it must be a description of some part of reality”.113 Likewise with Dworkin on the idea that “the most fundamental questions about morality are not themselves moral, but rather metaphysical, questions”.114 Or, as Scanlon puts the point, “moral judgments as I

understand them do not conflict with the claims of science, nor do they need the support

of, nor could they be supported in a plausible way by, metaphysical claims”.115

But it is not just that metaphysical backing is unnecessary: as Scanlon’s quote

suggests, it is positively inappropriate. Theorists seeking truth-makers are making a serious mistake, quietists often say, for the allegedly second-order questions about the existence and nature of moral truth-makers cannot be understood except as first-order moral questions in disguise. Thus the question of whether torture is objectively wrong cannot be understood except as the question of whether the claim that it is wrong is (as

Nagel puts it) “true or false independently of our beliefs and inclinations”—so that “no

112 Nagel (1986: 139–40). 113 Putnam (2004: 60). 114 Dworkin (2011: 25). 115 Scanlon (2003: 9). 88

other kinds of truths could imply the reality of values”.116 To get more details on moral truth, “all we can do is to refer to the arguments that persuade us of the objective validity of a reason or the correctness of a normative ”.117 Thus apparently second-order

metaethical questions about the metaphysical backing of moral truths are rejected as

misguided when not reinterpreted as first-order moral questions about which moral

claims are true.

This reinterpretation approach has some obvious affinities with the quasi-realist

expressivist account of moral objectivity. Expressivists do not only account for simple

moral judgments in terms of practical attitudes, but also account for claims to moral

objectivity in terms of practical attitudes. Thus both first-order and apparently second-

order claims are given the very same treatment. But quietists do not follow expressivists:

their main goal is to defend objective moral truths, and they are therefore suspicious of

the suggestion that morality cannot have orthodox truth-conditions for its judgments.

But it is not clear how quietists might have moral objectivity on their own terms. We

have seen two ways of understanding moral objectivity. Quietists repudiate the first

approach entirely and have no interest in following expressivists in the second approach.

This leaves them with the following options: finding a third approach, finding a descrip- tivist version of the second approach, or abandoning any attempt to give an informative account of moral objectivity. The last option strikes me as a surrender, and I cannot think of a third approach. Thus I will examine the possibility of a descriptive version of the

116 Nagel (1986: 144). 117 Nagel (1986: 139). 89

second approach: i.e., a descriptivist account of the psychology of moral objectivity that

carries no commitment to any truth-conditions.

In examining this problem, I will not examine quietism as a whole, but will instead focus on the work of Ronald Dworkin. I have two reasons for doing this. First, Dworkin’s

defense of quietism is arguably the most developed in the literature, and certainly the

most engaged with contemporary metaethics. Second, Dworkin is defending the most ex-

treme version of quietism in metaethics: quietism about metaethics itself. As I said above,

quietists and quasi-realist expressivists alike reject the notion that claims to moral

objectivity harbor metaphysical commitments. This could be called quietism about moral metaphysics, and though it obviously rules out any metaethical questions about the nature of moral metaphysics (except when reinterpreted as first-order moral questions), it is compatible with other metaethical questions (e.g., about the nature of moral judgment).

But Dworkin aspires to reject all metaethical questions (except when reinterpreted as first-order moral questions). Thus an examination of his view will amount to an examination of the lengths to which quietism can be pressed, as well as the metaethical questions that might (contra Dworkin) distinguish quietism from its rivals.

4.2 Dworkin’s theory

Dworkin aspires for his theory to respect what he calls “the ordinary view”. I find it helpful to separate the ordinary view into two subtheses. First, what I will call objectivism: the view that there are objective moral truths. Dworkin illustrates this subthesis with an example where you that it is wrong to stick pins in babies for the fun of 90

hearing them scream, and that it is wrong regardless of what anyone thinks.118 Second, a grab-bag of subtheses I will generously call ‘face value’-ism. Ordinary people don’t think their moral views can be proven scientifically (even in principle) or are genuine observations causally downstream of efficacious moral properties, but instead that moral argument is needed to see the moral truth. Moreover, ordinary people don’t think that your moral views are disqualified from truth-aptness due to being practical attitudes rather than descriptive beliefs. With these ‘face value’ additions, Dworkin is in effect insisting that the ordinary view clashes with all forms of reductionist naturalism and expressivism.119

All theories which reject the ordinary view are, for Dworkin, suspect. But his focus is on those ‘skeptical’ theories which reject objective moral truths, and in particular those forms of skepticism which would reject objective moral truths in toto by drawing on external truths about morality alone. Thus Dworkin does not much concern himself with forms of ‘internal skepticism’ which would draw on general moral truths to reject more particular moral truths. For such views are themselves committed to objective moral truths, at least counterfactual ones such as “If God existed, then those moral claims would be objectively true”. Dworkin concerns himself with forms of ‘external skepticism’, which would eschew all commitment to objective moral truths, and attempt to make a purely non-moral case against the entire category of objective moral truth.120

118 Dworkin (2011: 26–27). 119 Dworkin (2011: 27–28). 120 Dworkin (2011: 30–32). 91

Dworkin recognizes two forms of external skepticism: (i) ‘error skepticism’, which

deems all moral judgments erroneous for making claims that are not true (commonly

called ‘error-theory’), and (ii) ‘status skepticism’, which holds that moral judgments are

not even making the kind of descriptive claims that are capable of truth (commonly called

‘expressivism’).121 There is a small puzzle about Dworkin’s taxonomy: if skeptical

theories are the ones which reject objective moral truths, and if expressivist status

skepticism is a form of skepticism, then why did Dworkin have to add explicitly anti- expressivist subtheses to his ‘face value’ characterization of the ordinary view? Wouldn’t expressivism have already been ruled out by the objectivism of the ordinary view? The answer, I believe, lies in the fact that expressivism is perfectly compatible with objectivism as Dworkin characterizes it: expressivists can and do hold that certain things are wrong regardless of what anyone thinks, while also holding that moral judgments are not descriptions and thus do not possess orthodox truth-conditions. Thus, to preserve his taxonomy, Dworkin must either relax his ‘skepticism’ to include views which allow for non-descriptive objective moral truths, or else tighten his ‘objective moral truths’ to include a descriptivist condition. And, happily for Dworkin, neither modification seems objectionable.

Dworkin hopes to show that external skepticism is a lost cause: you can’t make a non-moral case against morality, which means you can’t make a case against morality without accepting some objective moral truth. But importantly, this isn’t just a point against external skepticism: Dworkin thinks the same point applies against all external

121 Dworkin (2011: 32–33). 92

views, including those positive views which think there is a non-moral case to be made for morality. Realists who think that objective moral truth could be vindicated by metaphysical considerations alone, in his view, are just as wrong as error skeptics who think that objective moral truth is refuted by metaphysical considerations, or expressivists who think it is refuted by semantic considerations. Questions of objective moral truth cannot be settled by considerations from outside the domain of morality alone.122

Thus Dworkin rejects the realist view that “there is some kind of entity or property in

the world—perhaps morally charged particles or morons—whose existence and

configuration can make a moral judgment true”.123 Indeed, the realist approach is simply

off on the wrong track:

The ordinary view insists that moral judgments are not made true by historical events or people’s opinions or emotions or anything else in the physical or mental world. But then what can make a moral conviction true? If you think the Iraq war immoral, you can cite various historical facts… that you believe justify your opinion. But it is hard to imagine any distinct state of the world—any configuration of morons, for instance—that can make your moral opinion true the way physical particles can make a physical opinion true. It is hard to imagine any distinct state of the world for which your case can be said to be evidence.124

Moral truths, it is clear, cannot possess subjective truth-makers without abandoning the

ordinary view’s insistence on objectivity. But nor can these truths possess the objective

truth-makers (perhaps ‘morons’) posited by realists. Such truth-makers are not only of

questionable reality, but are positively “hard to imagine”. But this is no problem for

moral objectivity: the ordinary view does not contain the metaphysical commitments

attributed to it by realists. But if the ordinary view explicitly denies subjective truth-

122 Dworkin (2011: 36–37). 123 Dworkin (2011: 26). 124 Dworkin (2011: 29). 93

makers, and if it has no truck with incomprehensible objective truth-makers, then

Dworkin’s theory arrives at an important claim: objective moral truths, it seems, must do

without any truth-makers at all.

But it is also a puzzling claim, for how can there be truths without truth-makers? Is

there nothing that makes a moral judgment true? Now Dworkin has an official answer to

this: “moral judgments are made true, when they are true, by an adequate moral argument

for their truth.” And this raises the specter of constructivism, suggesting that moral truths

are made true by being the output of an argument that satisfies some formal rules of adequacy for moral argument. But in fact he has nothing of the sort in mind: adequacy is not some formally discernible feature of moral argument, but instead a feature possessed by a moral argument only in virtue of “a further moral argument for its adequacy”.

Indeed, Dworkin candidly offers that “[w]e can say nothing more helpful than what I just said”. Thus we are roughly back where we began: perhaps Dworkin can offer uninformative minimalist truth-makers, either directly (e.g., what makes p true is the fact

that p) or postponed a few steps (e.g., what makes p true is an adequate argument for p), but his theory recognizes no robust truth-makers in virtue of which objective moral truths count as true.125

But this rejection of the robust truth-makers posited by moral realism puts Dworkin in unwelcome company: the ‘skeptics’ who embrace error-theory or expressivism. For they are in metaphysical agreement that the actual world contains no robust moral properties, leaving them disagreeing only about the fate of objective moral truth in a world without

125 Dworkin (2011: 37). 94

morons. And if Dworkin is right, this residual disagreement must be a moral disagreement. Consequently, his theory needs a way to distinguish itself from these two forms of anti-realism on moral grounds, lest it end up with no better claim to objective moral truth.

4.3 Dworkin on error-theory

Dworkin’s case against error-theory turns on the key proposition that you cannot deny all moral claims without making a moral claim, or (better), you cannot deny all objective moral truths without presupposing an objective moral truth. If this proposition were true, then error-theory would be self-contradictory: it would be committed to exactly the sort of objective moral truth it purports to reject. And this would certainly break the back of error-theory.

Now, since no clear-headed error-theorist would ever tempted to accept this proposition on its own, Dworkin needs to make a non-question-begging case for it. He needs to make it plausible that the error-theorist’s denial of objective moral truths must itself be interpreted as a moral judgment, i.e. an aspiring objective moral truth. And

Dworkin’s discussion seems to indicate two lines of support.

The first line of support looks no further than the denial itself. “There are no objective moral truths” seems to offer an opinion about what categorical moral reasons people have, namely none, and thus can certainly in some sense count as belonging to the moral domain. And if the claims of permissibility we make every day are denials of categorical moral reasons (for some agent in some circumstance), then the error-theorist’s denial is as 95

moral as any claim of permissibility. Such are the reasons Dworkin offers for interpreting

the denial itself as a moral judgment.126

But the second line of support brings in the rationale behind the denial. An error-

theorist might point to the nonexistence of God, or the nonexistence of intrinsically

motivating morons, or the implausibility of moral convergence among all rational , etc. as a compelling rationale for rejecting objective moral truths. But this rationale is virtually always wed to some hypothetical thesis to the effect that there would be objective moral truths if only things were different: e.g., if only God actually existed.

And such a hypothetical thesis can be plausibly interpreted as a moral judgment. For example, the hypothetical thesis that if God existed then it would be morally wrong to disobey his commands arguably commits you to the moral judgment that it is morally wrong to disobey God. Thus error-theorists who appeal to a rationale wed to such a moral judgment are in trouble: they turn out to accept (at least some) moral judgments as objective moral truths, despite their best efforts at rejecting them all.127

Now, error-theorists have a quick and easy way of responding to Dworkin’s

argument. They could say that even if their denial of objective moral truths is itself a

moral judgment with objective purport, it is a different kind than they originally intended

to deny. They can distinguish between the ‘good’ objective moral truths accepted by their

theory, or its rationale, and the ‘bad’ objective moral truths targeted for rejection by their

theory. Of course, Dworkin might protest that anyone who accepts some moral truths to

126 Dworkin (2011: 40–44). 127 Dworkin (2011: 46–47). 96

rule out others is not an external skeptic, but only an internal skeptic. But error-theorists

could presumably live with that label, so long as their theory remains the same.

But this quick response won’t work on its own. It stands in need a principled way to

draw the distinction between the ‘good’ and the ‘bad’ objective moral truths. Otherwise

error-theorists will have no answer for the question “What is it that makes the ‘good’

objective moral truths true, and why doesn’t it work equally well to make the ‘bad’ ones

true?” On the one hand, no self-respecting error-theorists would allow mysterious objec-

tively prescriptive properties into their ontology, just to serve as truth-makers for the few objective moral truths they accept. And on the other, if they allow for something less that could make objective moral truths true, then they compromise their own interpretation of moral evaluation as committed to mysterious objectively prescriptive properties. So they need a way of accommodating these few objective moral truths that doesn’t open the floodgates to objective morality, overwhelming and drowning their own theory.

A return to Dworkin’s first line of support might spell relief. Consider his claim that we can classify “the proposition that astrology is bunk” and “that there is no [planetary influence on human lives]” as a bona fide “astrological judgment”. Or that atheism can be classified as a “religious position” inasmuch as it “offers an opinion about the existence or properties of divine beings”. Or, indeed, that even if “no claim anyone makes about the shape or color of unicorns is true because there are no unicorns”, then nevertheless this claim itself is a true “proposition of unicorn zoology”.128 If error-theory is in as good a shape as these views, and if error-theory takes a position on morality only in the same

128 Dworkin (2011: 40–42). 97

sense that we take a position on unicorn zoology, then surely the error-theorist should

celebrate. To be sure, there may be a contradiction in carelessly formulated debunkings

of astrology, gods, or unicorns, but presumably this is a problem that can be easily

overcome. And so just as atheists can deny all religious propositions carrying the

presupposition that God exists while allowing that their own position is religious in a

broader sense, error-theorists can deny all moral propositions carrying the presupposition

that objective moral properties exist while allowing that their own position is moral in a

broader sense. This parallel also does away with the ‘open the floodgates’ worry: just as atheists can say that their religious truth is true because it accurately reports the non- existence of God, whereas others erroneously presuppose the existence of God, likewise for error-theorists and objective moral properties.

But what about Dworkin’s argument from permissibility? He envisions a disagreement over abortion: A says abortion is always immoral, B says abortion is sometimes morally required, C says abortion is always permissible, and D (the error- theorist) says abortion has no moral status whatsoever, neither forbidden nor required nor permissible. Dworkin thinks it is clear that C is taking a moral position, and argues that D and C are saying the same thing: that there are no categorical reasons one way or the other when it comes to abortion. It follows that D is also taking a moral position, which means error-theory contradicts itself.129

I think the argument is guilty of something like equivocation. C is taking a moral position, and D and C are saying the same thing, but these two claims can be understood

129 Dworkin (2011: 42–44). 98

differently. What D and C are saying about abortion may be the same in terms of their

literal semantic content, for both are denying the presence of categorical reasons when it

comes to abortion. But it is unclear whether to count this alone as “taking a moral position”. For what clearly makes it the case that C is taking a moral position is what C’s statement pragmatically communicates: namely, that C is willing to count certain actions as forbidden or required, but wishes to distinguish abortions from these other actions.

This distinction is communicated by the word ‘permissible’, which does more than deny

the presence of categorical reasons in the case at hand, but also pragmatically commits

the speaker to recognizing other actions as forbidden or required by categorical reasons

and recognizing some morally significant difference between these actions and cases of

abortion. For imagine that C went on to say “Of course, I agree with D that all things are

morally permissible, because I don’t believe in any categorical reasons in any case.” Such

an addendum would cast doubt on the idea that C was taking a moral position in the first

place. This is why D feels the need to say that abortion is not even permissible: this is D’s

way of canceling the pragmatic implicature of C’s statement, and goes some way to

communicating D’s unwillingness to allow categorical reasons in any case.

Consequently, if D is taking a moral position only in the same sense that C’s literal

content expresses a moral position, then it is far from clear that D is taking any moral po-

sition at all. After all, it is no more clear that the literal content of “abortion is always

morally permissible” is any more of a moral judgment than a provocative and sweeping

skeptical claim like “all things are morally permissible”, which leaves us where we

began. 99

So Dworkin’s first line of support poses no serious threat to error-theorists. If error-

theory’s denial of objective moral truths or categorical reasons is itself a moral judgment

only in the same sense that atheism’s denial of objective religious or theological truths is

itself a religious judgment, then there is no realistic danger of self-contradiction. And if

this denial is every bit as moral as a claim of permissibility only when it comes to what is

literally stated (and not pragmatically communicated) by a claim of permissibility, then it

is not clear that any moral claim is being made in the first place. Of course, we have not

shown that these claims are not moral, and that error-theory is not self-contradictory. But then, such a showing is not necessary for a successful rebuttal of Dworkin’s argument.

Dworkin’s second line of support is stronger and weaker than the first. It is stronger in its interpretation of certain claims as bona fide moral judgments aspiring to objectivity.

But it is weaker in its attempt to impute these claims to error-theory. For while many error-theorists may in fact back up their theory by bringing in apparently moral claims about God or morons or convergence, it is by no means clear that an error-theorist must do so.

For one thing, an error-theorist might do without any rationale at all. This might not win the view any popularity contests, but it certainly clears it from Dworkin’s charge of self-contradiction. For if the view itself has not been shown to be a moral judgment (as

I’ve just argued), and if the error-theorist refuses to offer any rationale that might itself be shown to be a moral judgment (as we are now stipulating), then Dworkin is left without any way of doing damage to the view, much less charging it with self-contradiction. 100

But moreover, an error-theorist might offer a rationale different in kind from the sort

Dworkin considers. Instead of appealing to a metaphysical connection between moral

objectivity and God (or morons, etc.) the error-theorist might offer an epistemic case: “It

is reasonable to reject claims of objective truth in the absence of compelling empirical

evidence for their truth. The purported objective truths of morality are not supported by

any compelling . Therefore, it is reasonable to reject the purported

objective truths of morality.” Dworkin would agree that empirical evidence provides no

support for moral objectivity, and would indeed insist that it is a mistake to even treat

moral objectivity as a matter to be decided by empirical evidence. But without some

additional argument, any error-theorist who evaluates all claims in terms of empirical

evidence would be completely unfazed. The charge of self-contradiction certainly would

not hit home, not unless Dworkin could make the case for assimilating the epistemic

rationale provided by the error-theorist to a moral claim.130

But because error-theorists frequently do provide the sort of metaphysical rationale considered by Dworkin, it is worth seeing whether they can escape his argument. And unfortunately for the error-theorist, I think Dworkin is on stronger ground here. For example, if objective moral truth is metaphysically connected with God’s will, then the error-theorist is committed to the claim that if God existed and forbade abortion, then abortion would be immoral. This certainly looks like a moral claim: just the sort of claim

that could be discussed and evaluated by the ordinary rules of moral argumentation. Thus

you might question the moral relevance of God’s will, or of sanctions in the afterlife, or

130 I am actually sympathetic to this: I think epistemic reasons raise the same metaethical worries as reasons for action. 101

press concerns about counterintuitive results (“what if God commanded mandatory third- trimester back-alley abortions for all?”). Some ‘sic semper tyrannis’-type might even hold the contrary position, contending that we would be obligated to defy God’s

commands, in which case the resulting disagreement would be plausibly construed as a

moral disagreement. The judgment, then, looks genuinely moral.

And if the error-theorists try to make heavy weather of the hypothetical character of

the judgment, thereby clearing themselves of having made any bona fide moral

judgments, we can remind them of Mackie’s insistence on explaining supervenience

relations. Supposing that is objectively true that objective moral truths supervene upon

God’s will, which is why the nonexistence of God spells doom for objective moral truths,

then there must be some explanation for why they supervene this way and not some other

way (e.g., the ‘sic semper tyrannis’ way). And if error-theorists reject the call for such an

explanation, then Dworkin will seemingly be entitled to reject the call to explain the

supervenience relations picked out by whichever ethical theory he subscribes to. This

would vindicate his charge that error-theorists’ background commitments are ultimately

self-defeating.131

So what these error-theorists need is a way of accounting for the truth of their

metaphysical claim that could not be borrowed to account for the objective moral truths

Dworkin is interested in. And mirabile dictu, I think there is such a way. Recall that

error-theorists base their claims on an interpretation of the practice of moral evaluation.

They think moral evaluation includes a claim to objectivity that commits us to objectively

131 For more on this point, see Chapter 5 on error-theory. 102

prescriptive moral properties (or to God or to convergence, etc.). But if this interpretation

is correct, then when they go on to connect objective moral truth with these

metaphysically robust properties, they have an explanation of this connection which is

not at all mysterious. Namely, they have a semantic explanation in terms of the correct interpretation of moral evaluation. There is a truth-maker for the claim “if objectively prescriptive moral properties existed, then we morally ought to follow their lead”: namely, the human practice of moral evaluation. Just as the atheist can explain why

God’s existence is needed for the truth of certain religious claims by providing a theologically-laden interpretation of those claims, the error-theorist can explain why special metaphysics is needed for the truth of ordinary moral claims by providing a metaphysically-laden interpretation of those claims.

Now, not all interpretations of moral evaluation are equally plausible. It is rather doubtful that moral evaluation is itself committed to any supernatural agency, much less the Western conception of God. And it is at least questionable whether moral evaluation is itself committed to any intrinsically motivating moral properties: the wilder sort of

‘moron’. But surely there is nothing obviously false about the idea, shared by error- theorists and realists alike, that moral evaluation is itself committed to the bare and unadorned claim that some or other robust truth-makers actually exist, metaphysically robust properties who presence or absence determines what we morally ought to do. As more details are added to this picture (whether the properties are naturalistic, whether they are intrinsically motivating, whether they are ever encountered in ), it becomes less and less plausible that moral evaluation carries such commitments. But the 103

bare claim of moral truth-makers as an essential commitment of moral evaluation is something worth considering, and it is in any case something error-theorists were already sold on.

But this means that error-theorists have a response to Dworkin’s argument. They can admit that they accept something in the vicinity of an objective moral truth, viz. that the existence of moral truth-makers would indeed provide agents with categorical reasons.

But they can offer an explanation for this ‘objective moral truth’ that does not com- promise their error-theory, and does not lend assistance to Dworkin. They can say that this ‘objective moral truth’ is a mere consequence of the correct interpretation of the very practice of moral evaluation. This separates it from such moral claims as “Abortion is wrong” or “Pleasure is the only good” or even “We ought to obey God’s commands”. For those claims cannot be construed as mere consequences of any plausible interpretation of the practice of moral evaluation: those who deny them are not plausibly construed as leaving the practice.

Of course, Dworkin would deny that moral evaluation is correctly interpreted as containing any metaphysical commitments. So would many philosophers, and I agree that their view is highly plausible. But to take sides on this dispute would clearly beg the question against error-theorists, and thus disqualify the argument. Allowing them their own view on how to interpret moral evaluation, they can provide a coherent account in which they accept certain claims that Dworkin would term moral but still retain their sweeping rejection of all other moral judgments. 104

Finally, we should address the question we started with: given that both reject the

actual existence of moral metaphysics, what exactly distinguishes Dworkin’s view from error-theory? The main point of difference concerns the topic we just discussed: whether moral evaluation should be interpreted as containing metaphysical commitments. As we have seen, though Dworkin’s “ordinary view” is committed to a form of objectivity, it contains no metaphysical commitments, and indeed Dworkin finds the very idea of any such metaphysics “hard to imagine”. This is why Dworkin feels free to embrace objective moral truths and error-theorists feel constrained to deny them: they disagree about whether ordinary belief in objective moral truths carries with it a commitment to moral metaphysics.

But then it looks for all the world like we have a second-order metaethical disagreement on our hands, viz. a disagreement about the semantics of moral evaluation.

Now, Dworkin recognizes legitimate second-order questions about morality: e.g., whether “our economic or other circumstances explain why we are drawn to moral convictions that other cultures with different circumstances reject” or whether a certain moral opinion “is almost universally held”.132 But of course, as a metaethical quietist, he

is inclined to reclassify traditional metaethical questions under the domain of first-order

morality. And thus in order to maintain his position, he will have to defend the view that

this question of moral semantics is unlike the questions he recognizes as second-order

and instead belongs to the first-order moral rather than the second-order metaethical

132 Dworkin (2011: 31). 105

domain. This seems an exceptionally hard road to hoe, and none of the moves Dworkin

has offered thus far seems capable of doing the trick.

4.4 Dworkin on expressivism

We have just seen that Dworkin’s theory distinguishes itself from error-theory at the cost

of recognizing the legitimacy of an apparently second-order metaethical question about

moral semantics. But then it will be still more difficult to distinguish his theory from

expressivism: for expressivism is on Dworkin’s side of that crucial question, agreeing

that moral evaluation harbors no metaphysical commitments. As we review and critique

his case against expressivism, it will become clear that Dworkin has misunderstood the

basic workings of expressivism, especially in its quasi-realist form, and thus is left with

very little space for his theory to occupy.

He first considers what it is that expressivists—as status skeptics—are rejecting. If

the rejected claims are themselves moral claims, then their rejection is itself a moral

position and thus the thesis of expressivism would have to give itself an expressivist

treatment, thereby undermining itself. Thus if someone claims that abortion is objectively wrong, meaning that it is wrong for everyone in all circumstances, then expressivists would be taking a moral position when rejecting such a claim of objective wrongness.

And yet if the rejected claims are not themselves moral claims, then it is unlikely that

they are part of the ordinary moral evaluation that skeptics wish to impugn. Thus if they 106

wish to reject the existence of God, that’s perfectly harmless, because since ordinary claims to moral objectivity harbor no theistic commitments.133 Dworkin writes:

[An expressivist] can remain both external and a skeptic only if he can find something else in my further claims [claims to objectivity in one’s moral judgments], something that is not itself a moral claim and yet whose denial has skeptical implications. I shall call these the twin conditions of semantic independence and skeptical pertinence.134

Since Dworkin is convinced that the two conditions cannot be met, we can put his view thus: there is nothing belonging to ordinary moral evaluation and its claims to objectivity

(the only pertinent target of skeptics) other than first-order moral claims (which enjoy no semantic independence).

To this, I think, there is a quick and successful reply from expressivists: we are not interested in skeptical pertinence. Expressivism’s chief negative thesis, after all, is that moral judgments are not the sort of descriptive beliefs that have orthodox truth- conditions. In other words, moral judgments are not (at least primarily) bearers of descriptive content whose truth-conditions could be provided by robust metaphysical facts. And this negative thesis has skeptical implications only given the assumption that ordinary moral evaluation carries a commitment to its own descriptive character. But this assumption is not much different from the assumption that ordinary moral evaluation carries robust metaphysical commitments. And in rejecting this assumption both Dworkin and expressivists agree. So expressivists never wanted to be status skeptics: they only wanted to deny the metaphysically-oriented sort of objective moral truth at the center of the debate between error-theorists and realists, leaving open the possibility of providing a

133 Dworkin (2011: 52–55). 134 Dworkin (2011: 55). 107

satisfactory anti-realist account of objective moral truth (the so-called project of ‘quasi-

realism’).135

Of course, Dworkin has ways of making expressivists into skeptics. As earlier noted,

he has included an anti-expressivism clause in his account of “the ordinary view”: an

ordinary person “would be puzzled if someone told you that when you express a moral

opinion you are not really saying anything”, or that “you are only venting your spleen, or

projecting some attitude, or declaring how you propose to live, so that it would be a

mistake to think that what you had said is even a candidate for being true”.136 And if

ordinary moral evaluation really does rule out the view that moral judgments are

(expressions of) practical attitudes rather than descriptive beliefs, then expressivists really are skeptics.

But this move would seem to give expressivists a solution to Dworkin’s own challenge. If ordinary moral evaluation contains not only first-order moral claims, but also an anti-expressivism clause, then this clause is what expressivists reject. The clause is semantically independent, for it does not appear to be a moral claim itself, and it is skeptically pertinent, assuming that Dworkin is right to attribute it to the ordinary view.

Of course, if the clause really is semantically independent, it is a second-order metaethical view of the sort Dworkin is loath to recognize. And the more independent it is, the more dispensable it is to the practice of moral evaluation, and the less skeptical a quasi-realist expressivist revision of moral evaluation will seem to be.

135 See Blackburn (1993), Blackburn (1998), Gibbard (1990), Gibbard (2003), Timmons (1999). 136 Dworkin (2011: 27–28). 108

Consequently, I suspect that Dworkin intends the anti-expressivism clause to be a moral claim itself. If I am not merely declaring how I propose to live, for example, perhaps this is because I will hold other people to what I take to be objective moral truth.

Or if I see my moral judgments as truth-apt, perhaps this is because I am willing to subject them to the crucible of careful reasoning and open discussion, where truth and falsehood are the name of the game. But then expressivists need not count as skeptics.

Interpreting such claims to moral objectivity as first-order moral claims is a time-honored expressivist tactic, and there is no obvious reason why the anti-expressivist clause understood morally could not be given the same expressivist treatment given any other moral claim.

The same drama between non-skeptical expressivism and the anti-expressivist clause recurs throughout Dworkin’s discussion. Consider his case against Richard Rorty’s ‘two language games’ version of status skepticism.137 Rorty’s position is supposed to be

comparable to treatments of fictionalist discourse: in “playing the world-of-fiction

game”, I can say that Lady Macbeth was married more than once, while going on to deny

that Lady Macbeth even exists, when playing the “real-world game”. Now, a feature of

this account is that what we say in one game comes into surface contradiction with what

we say in another game, so that we look to be just plain inconsistent. But any such

problem can be resolved by giving the two apparently conflicting statements interpreta-

tions that reveal their true content, showing that their true content does not actually come

into conflict. Thus we could interpret world-of-fiction claims about Lady Macbeth as

137 Dworkin (2011: 59–61). 109

nothing more than claims about what should be said according to the world-of-fiction game, so that such claims do not conflict with the claim that she is merely a fictional creation of Shakespeare.138

Rorty wishes to apply the same sort of approach to a variety of ordinary claims:

Dworkin focuses on geological claims and moral claims. In everyday life, we say that

mountains exist independently of humans and that torture is wrong independently of

humans. But as philosophers, according to Rorty, we should deny that “Reality as It Is In

Itself” contains mountains or wrongness. Here, as in the Macbeth case, Dworkin sees a

potential conflict and looks for a way to resolve it through careful interpretation. But he

cannot find any way to do this, due to the two conditions of semantic independence and

skeptical pertinence:

[Rorty] hoped to display a difference [in meaning between the two propositions] through capitalization: the second proposition sports capital letters that the first lacks. But that device is not helpful. If we give the sentence “Mountains are part of Reality as It Is In Itself” the meaning it would have if anyone said this, then it means nothing different from “Mountains exist, and would exist even if there were no people,” and the contrast Rorty needs disappears. If, on the other hand, we assign some novel or special sense to that sentence—if we say, for example, that it means that mountains are a logically necessary feature of the universe—then his argument loses any critical force or philosophical bite because no one would or could think that mountains are logically necessary.139

To secure semantic independence, Rorty must know of some difference in metaphysical

strength between what he affirms in the geology game (something weak and harmless)

and what he denies in the philosophy game (something strong and serious). But if what

138 Dworkin (2011: 60). 139 Dworkin (2011: 61). 110

he denies is something so ludicrously strong that nobody ever “would or could” think it true, then he has failed to secure skeptical pertinence. The same dilemma recurs.

But, unsurprisingly, Dworkin runs into the same problem: his target is not interested in skeptical pertinence in the sense of challenging the ordinary view. Rorty thinks his account is at peace with the ordinary view, and only poses a skeptical challenge to the metaphysically-oriented philosophical claims made in realist accounts of the ordinary view. Now, it is surely controversial to deny that the ordinary view of mountains harbors no serious metaphysical commitments. But this appears to be Rorty’s view of the matter, and such a view enjoys more popularity (and indeed prima facie plausibility) when applied to ordinary moral evaluation. Rorty and Dworkin can agree that everyday moral life is not committed to morons, and they can agree that realists are wrong to posit moral metaphysics to make sense of moral objectivity. But then what do Dworkin and Rorty have to disagree about when it comes to morality?

At this point, Dworkin could insist that the ordinary view harbors commitments that

Rorty wishes to reject—something analogous to the anti-expressivist clause. Then the difference between he and Rorty would lie in those commitments. But if they are moral commitments, then it is unclear why Rorty could not accommodate them as belonging to the moral game, right alongside other moral commitments (e.g., mind-independence).

This would again bring Dworkin and Rorty into full agreement. Alternatively, if those commitments are non-moral commitments (e.g., metaphysical commitments), then

Dworkin would have to give up on his entire quietist approach: instead of dismissing all metaethical issues as nothing more than misunderstood moral issues, he would have to 111

recognize a genuinely metaethical divide between himself and Rorty, and then argue that

the ordinary view supports his side of the disagreement. Either way, the prospect of a

distinct quietist objectivity slips away.

Finally, consider Dworkin’s case against the quasi-realist expressivism associated with Blackburn and Gibbard.140 He regards them as defending a ‘two language games’

view somewhat similar to Rorty’s. But their strategy works by explicitly classifying any

claim to moral objectivity as belonging to the game of ordinary moral evaluation, leaving

their skepticism confined to the philosophy game. Thus Dworkin has a “self-described

‘projectivist’” making the philosophical claim that “in reality moral convictions must be

understood as emotional projections onto a morally inert world”, then the moral claim

that torture is wrong “no matter what attitudes or emotions anyone had about it”, and then

the further philosophical claim that the previous claim of moral objectivity is itself “only

the projection of an attitude”.141

As before, Dworkin sees a surface contradiction between what is said in the two

games—about whether morality does or does not depend on attitudes—and cannot find

any way to resolve it through careful interpretation:

[The projectivist] can’t replace what he says in the morality game with any statement, while still in that game, that implies or allows that wrongness is only a matter of projection. He can’t replace his statement in the philosophy game by declaring or implying there that wrongness does not depend on projection.142

But there is no contradiction here, not even a superficial one. What is said in the

philosophy game is about the nature of evaluation, and what is said in the moral game is

140 Dworkin (2011: 62–63). 141 Dworkin (2011: 62). 142 Dworkin (2011: 62–63). 112

about the object of evaluation. For example, the moral claim that torture is wrong is about torture, and it carries no obvious implications about the nature of moral evaluation itself—e.g., whether moral judgments are more like descriptive beliefs or practical attitudes. Likewise, the philosophical claim that moral judgments are not descriptive beliefs is about moral evaluation itself, and it carries no obvious moral implications about torture or abortion or any other topic we might take a interest in—e.g., it does not say that torture’s wrongness depends on anyone’s attitude about torture.

In other words, expressivists are saying that moral judgment depends on or consists in attitudes, not that anything’s moral status depends on or consists in attitudes. And this distinction is an important one, for it is needed by virtually any theorist: whether we are expressivists or whether we see the distinctive attitude of moral judgment as descriptive belief, no extant account of the nature of our moral attitudes should be seen as committing us to the view that anything’s rightness or wrongness depends on our moral attitudes. That is, without such a distinction, we cannot give any account of moral evaluation—expressivist or otherwise—without thereby becoming moral relativists of the

most implausible kind.

Of course, Dworkin could simply force a contradiction. By insisting on the anti- expressivist clause in his account of the ordinary view, he could say that any moral claim is in conflict with the expressivist’s claim in the philosophy game. But this only returns us to the original dilemma. Taken as a moral claim, the anti-expressivist clause can pre- sumably be accommodated by expressivists and no disagreement between Dworkin and expressivists remains. Or, taken as a non-moral metaethical claim, the clause both 113

provides expressivists with something to be skeptical about, and also compromises

Dworkin’s quietism.

In sum, then, Dworkin’s case against expressivism comes to the same end as before.

He cannot sustain the charge that expressivism undermines itself, for expressivists are not

interested in rejecting any non-moral thesis belonging to the ordinary view as they

understand it. They are only interested in rejecting the metaphysical claims of metaethical

realism and the associated descriptivist interpretation of moral evaluation. Nor can

Dworkin rain thunderbolts on them from a quietist heaven, for the only way to

distinguish his own view from theirs is to recognize a metaethical issue they disagree

about: viz., whether moral judgments are correctly understood as (expressions of) nonde- scriptive practical attitudes. Thus he must once again compromise his metaethical quietism: distinguishing his view from error-theory via the metaethical issue of whether moral evaluation harbors metaphysical commitments, and now distinguishing his view from expressivism via the metaethical issue of whether moral evaluation consists in descriptive beliefs.

4.5 Dworkin and metaphysical quietism

Thus far we have been pushing Dworkin to retreat from his thoroughgoing metaethical quietism to a more modest metaphysical quietism. The denial of realism puts his theory in league with error-theory and expressivism. To distinguish his theory from these two versions of anti-realism, he must recognize issues to disagree on them with. But the issues available are not well-disposed for a quietist treatment: they look metaethical on 114

their face, and it is difficult to see how to interpret them as first-order moral questions.

Thus Dworkin is left with a more modest form of quietism that recognizes objective

moral truths, but denies robust moral truth-makers (distinguishing itself from realism and

constructivism), denies that moral evaluation harbors any such metaphysical

commitments (distinguishing itself from error-theory), and yet insists that moral

evaluation consists in descriptive beliefs (distinguishing itself from expressivism).

Note here that it is not enough for Dworkin to insist that moral evaluation consists in

beliefs. For quasi-realist expressivists are happy to characterize moral judgments as

beliefs, just so long as they are not ordinary descriptive beliefs, but rather a sort of belief

that essentially incorporates practical attitudes. According to the most developed attempt

to make expressivism safe for beliefs, the so-called “cognitivist expressivism” of Horgan

and Timmons,143 a moral judgment is a kind of belief whose overall content is not descriptive, but instead incorporates a distinctive sort of “affirmatory commitment” state directed at a “core descriptive content”.144 Thus the moral judgment that Carthage ought

to be destroyed is a moral belief whose overall content, that Carthage ought to be

destroyed, is not descriptive, but instead a special nondescriptive normative content

which incorporates a distinctive action-guiding “ought-commitment”—in this case

directed at the core descriptive content that Carthage is destroyed. This is contrast to e.g.

the ordinary descriptive belief that Carthage has been destroyed, whose overall content is

descriptive content “that represents, or constitutes, a way the world might be”.145 Now,

143 Horgan and Timmons (2006). 144 Horgan and Timmons (2006: 270). 145 Horgan and Timmons (2006: 261). 115

such a view nominally agrees with Dworkin in characterizing moral judgments as beliefs.

But because it takes pains to distinguish these expressivist-friendly moral beliefs from ordinary descriptive beliefs, it is incompatible with Dworkin’s purely descriptivist account of moral evaluation.

This leaves Dworkin characterizing moral judgments as purely descriptive beliefs directed at purely descriptive content which harbors no metaphysical commitment to robust truth-conditions: descriptivism without metaphysical commitments. And this presents a challenge: viz., the challenge of making sense of such a view. For the most natural way to understand purely descriptive beliefs with purely descriptive content is in terms of their robust truth-conditions: e.g., the belief that Carthage has been destroyed is most naturally understood in terms of the destruction of an actual city. Indeed, the purely descriptive beliefs Horgan and Timmons against which they define their view are explicitly understood in terms of “ways the world might be”.146 This would seem to rule

out Dworkin’s position virtually by definition: you can’t have descriptivism without

metaphysical commitments, when description just is an attempt to describe how things are metaphysically.

Of course, it is important not to overstate the point. There are ways of understanding descriptive content which involve no direct appeal to any particular robust truth- conditions. The de dicto belief that the largest pig in the world has foul-smelling breath can be understood quite well independently of any information about which pig it is who satisfies the description. Similarly, an Australian-style indirect role-specification of all

146 Horgan and Timmons (2006: 262). 116

the a priori platitudes implicitly accepted by individuals who have mastered the concept

in question can be given without identifying the actual role-occupant. Finally, one might

pin down a concept by listing the characteristic causal and/or inferential relations the

concept enters into, all while knowing little to nothing about the actual referent (if any) of

the concept.

But even when descriptive content is understood via these indirect means, one cannot

take away robust truth-conditions without taking away its descriptive character. De dicto

descriptive beliefs, descriptive beliefs incorporating implicitly-defined theoretical terms, and descriptive beliefs incorporating conceptual-role-defined terms still seem to harbor metaphysical commitments to existence of their robust truth-conditions, whatever they turn out to be. Thus “there is water in that glass” harbors clear metaphysical commitments whether the concept of water entering into the descriptive content is individuated via some conceptual role, some implicitly-accepted Ramsey sentence, or some explicitly-represented “watery stuff” description. However we understand descriptive content, via direct or indirect means, descriptive beliefs by their nature seem to carry metaphysical commitments to robust truth-conditions.

Sometimes the point is begged off with an appeal to reasons. Immediately after eschewing Platonism, Nagel says that “[t]he claim is that there are reasons for action” and adds later that “the question… is not ‘What can we see that the world contains, considered from this impersonal standpoint?’ but ‘What is there reason to do or want, considered from this impersonal standpoint?’”147 In a similar vein, Scanlon writes:

147 Nagel (1986: 139–40). 117

If moral judgments do not make, or need the support of, metaphysical or scientific claims, then what kind of claims “about the world” do they make? Why isn't it idle to take them as true? The answer is that moral judgments are not idle because they make claims about what we have reason to do, and this is something of importance to us, as rational beings. The kind of gravitas that they require is thus not metaphysical but normative.148

But an appeal to reasons does nothing to alleviate the original worry.149 Supposing the

claim that there are reasons (or that we have reasons) to have descriptive content, then the

most natural understanding of the claim involves a metaphysical commitment to the real

existence of a kind of entity: reasons. And if such a metaphysical commitment is denied,

then we are left with no obvious way of understanding the allegedly descriptive content

in claims about reasons.

Dworkin himself seems pessimistic about the possibility of an informative account of the content of moral judgments. Consider what these remarks about expressivism:

Some status skeptics insist that when ordinary people declare that torture is morally wrong, they themselves intend nothing more than to turn their thumbs down on the practice. That semantic story seems obviously wrong. What ordinary people mean when they say that torture is wrong is that torture is wrong. No restatement of what they mean can be as accurate.150

Now, it is unsurprising that Dworkin would reject any “restatement of what they mean”

in other terms. Such a restatement is redolent of the sort of semantic reductionism he

steadfastly rejects. But he seems to taking a stronger stance: for the expressivist account

being rejected is not a semantic reductionist account that tells us what people mean by

effecting a substitution of descriptive content, but a psychological account that tells us

what people mean by telling us what it is to think that torture is wrong. And if Dworkin

148 Scanlon (2003: 9). 149 For an attempt to diagnose the contrary assumption, see Olson (2009). 150 Dworkin (2011: 56). 118

will accept no account other than an uninformative minimalist account, then he is rejecting all informative accounts of the nature of moral evaluation and the descriptive content involved in it.

And this would at last reduce Dworkin’s theory to nonexistence. He cannot give us an informative account of moral objectivity in terms of robust objective truth-makers, for he denies that moral judgments have anything to do with robust objective truth-makers. He cannot give us an informative account of moral objectivity in terms of psychology, for he rejects all accounts but the most trivial. The theory, then, gives us no information at all about moral evaluation or its content: it only denies all other theories. But then we are provided with nothing to help us understand of how purely descriptive content could possibly carry no metaphysical commitments. But a theory beset by an apparent contradiction which refuses to say anything positive has no way of defusing the contradiction. And then it is scarcely a theory at all.

119

CHAPTER 5: ERROR-THEORY

Error-theorists deny the existence of moral properties. But unlike certain expressivists, they don’t go so far as to rule moral properties out as incoherent or impossible. In this chapter, I consider and criticize a version of error-theory according to which moral properties do not actually exist but do possibly exist. I present a dilemma turning on the question of whether, in worlds with moral properties, the moral supervenes on the nonmoral. I finish by arguing that J. L. Mackie’s discussion of explaining supervenience cannot provide a solution to this dilemma.

5.1 The possible existence of moral properties

Error-theorists and expressivists disagree about the nature of moral evaluation. To put it roughly, error-theorists model moral evaluation upon ordinary description, whereas expressivists reject this model and look for different (perhaps sui generis) conceptions of moral evaluation. Admittedly, this rough statement conceals complications. On matters of psychology, error-theorists take moral judgments to be ordinary beliefs, whereas expressivists take them to be other states of mind—emotions, acceptances of norms, special nondescriptive beliefs, or something altogether sui generis. On matters of semantic content, error-theorists take moral judgments to have ordinary truth-apt descriptive content, whereas expressivists take them to have other kinds of content— content that cannot be truth-apt except perhaps in a minimalist way. And on matters of language, error-theorists take moral statements to perform the function of ordinary 120

assertion, whereas expressivists take them to perform some other function—emoting,

prescribing, or a special nondescriptive kind of asserting. But I think the rough statement

is a good way of packaging these complications.

Despite this disagreement on the nature of moral evaluation, error-theorists and

expressivists are supposed to agree on matters of metaphysics. To wit, both sides reject

any moral ontology, denying the existence of moral properties.151, 152 This metaphysical

agreement, of course, gives rise to a new disagreement on the point of systematic error.

Error-theorists have moral judgments systematically falling short of truth, since they

make reference to moral properties that do not exist. And expressivists have moral

judgments protected from any such error, since these judgments never purported to

describe moral properties in the first place. But both sides are united in denying the

existence of moral properties.

But this alleged agreement can be questioned. To be sure, both sides reject the actual

existence of moral properties. But there is still room for disagreement over whether moral

properties possibly exist. And indeed, I think the two sides naturally end up disagreeing

on this issue. Error-theorists naturally end up asserting the possible existence of moral

properties, whereas expressivists naturally end up holding that moral properties couldn’t

even possibly exist. If I’m right about this, error-theorists and expressivists don’t even agree on matters of metaphysics.

151 It might make better sense to frame the issue not in terms of whether moral properties exist, but instead whether they are instantiated. But I’ll overlook this concern for ease of exposition. 152 Of course, some expressivists are willing to speak minimalistically of moral properties; thus “Abortion has the property of wrongness” says no more than “Abortion is wrong” (which might in turn say no more than “Boo, abortion!”). But I assume that there is a significant difference between such minimalist ascription of properties and a more ontologically robust realist ascription of properties, and my discussion is intended to go with the latter. 121

But why should the two sides naturally end up disagreeing? To see why, look at the

analogies used by the two sides. Error-theorists take moral judgments to be somewhat like claims concerning the activities of the Homeric gods, whereas expressivists take moral judgments to be somewhat like imperatives. Start with the first analogy. Claims about Homeric gods are claims about the world that could have been true, if only the world (especially Mount Olympus) had been different. There could have been entities serving as truth-makers for these claims. But now turn to the second analogy. There is

simply no sense to be made of entities somehow serving as truth-makers for imperatives.

No entities could possibly fit the bill, because imperatives aren’t the right kind of thing to

admit of truth-makers. But moral properties would have to be, if anything, suitable truth-

makers for moral judgments. Thus the expressivist’s analogy suggests that moral

properties are impossible, whereas the error-theorist’s analogy suggests that they are

possible.

To be sure, error-theorists have not been clear about where they stand. I can locate no

affirmation or denial of the possibility of moral properties in Mackie (1977) or Joyce

(2001). The issue mainly comes up in the expressivist literature, when expressivists

directly confront the idea of robust moral properties of the kind affirmed by realists and

denied by error-theorists. Blackburn, for example, has a hard time making sense of such

moral ontology. He draws an analogy to color:

So suppose someone did interpret Locke, not entirely implausibly, as holding that ‘we talk as if there are colours, although there are none really’. It seems then that this Locke would owe us a number of explanations. One would be, that he would need to explain in what way this world is deficient in terms of colour—how does it differ from some other world in which there are colours, really?... So: what would it be [for it to be] bad to neglect the needs of children? It is not so in this world, evidently, so 122

what is different about worlds in which it is? Do children in that world suffer more? But why would that cast doubt on it being bad to neglect ours?153

For Blackburn, the error-theorist154 is saddled with the view that moral properties do not

exist in the actual world, but do exist in some possible world. This allows Blackburn a

way of framing his explanatory demand: what exactly are these moral properties

supposed to be, such that our world lacks them but another world has them?

R. M. Hare makes similar remarks:

Think of one world into whose fabric values are objectively built; and think of another in which those values have been annihilated. And remember that in both worlds the people in them go on being concerned about the same things—there is no difference in the ‘subjective’ concern which people have for things, only in their ‘ob- jective’ value. Now I ask, ‘What is the difference between the states of affairs in these two worlds?’ Can any answer be given except ‘None whatever’?155

Hare is making the case that there is no real issue at stake between those who affirm and

those who deny the existence of moral properties: it is a pseudo-issue.156 Mackie quotes

this passage, insisting that there is a real issue, and that the first-order views we hold and we have do not (contra positivism) decide the second-order issue of moral ontology. He goes on to spell out the difference between the two worlds, writing: “If there were something in the fabric of the world that validated certain kinds of concerns, then it would possible to acquire these merely by finding something out, by letting one’s thinking be controlled by how things were” (Mackie 1977, 22).

153 Blackburn (2005: 324–25). 154 Here Blackburn is discussing fictionalism, not error-theory. But I am confident that this makes no difference. 155 Hare (1957/1972: 47) 156 Hare actually writes that “the quarrels between objectivists and subjectivists are purely verbal” (1957/1972: 41, emphasis added), and yet earlier in his essay writes, “There are no pseudo-problems in philosophy; if anything causes philosophical perplexity, it is the philosopher’s task to find the cause of this perplexity and so remove it” (1957/1972: 38). But even given these statements, it still seems acceptable to use the term ‘pseudo-issue’. 123

This lends credence to my earlier claims. Expressivists tend to hold that moral

properties are hard to even make sense of, and thus will naturally end up holding that

they couldn’t possibly exist. Error-theorists, on the other hand, tend to think that there is

good sense to be made of moral properties, and thus will naturally end up holding that

they could possibly exist.

I have been careful to avoid saying that expressivists and error-theorists are

committed to any particular position on this issue. In particular, I think there could be a

kind of error-theory which denies that moral properties are so much as possible. I would

like to see such a view developed in detail. But I would insist that orthodox error-theory

most naturally leads to the conclusion that moral properties are possible. And it is that

view I will now criticize.

5.2 Preview of the argument

Suppose an error-theorist agrees that moral properties possibly exist. Then she is faced

with the following dilemma: throughout the class of possible worlds in which moral

properties exist, the moral either supervenes on the nonmoral or it doesn’t. That is, either

no two items taken from any of those worlds are exactly nonmorally alike and yet

morally different, or else there are two items taken from somewhere in those worlds that

are exactly nonmorally alike and yet morally different.157 On the first horn of the

dilemma, the error-theorist accepts supervenience, but is landed with a worrisome

explanatory burden: she must explain why the moral supervenes on the nonmoral in the

157 For ease of exposition, I am relying on strong supervenience in its standard possible worlds formulation. Other varieties of supervenience would, I am confident, do just as well. 124

way it does rather than some other way—why, for instance, rightness and wrongness are not systematically inverted. On the second horn of the dilemma, the error-theorist denies supervenience, and is therefore faced with the problem of denying a thesis so plausible and non-controversial that many have seen it as an a priori truth or even a truth constitutive of conceptual competence. Either way the error-theorist is in trouble.

Take St. Francis as an example. Even if the actual St. Francis lacks moral properties

(such as the property of being good) due to the actual nonexistence of moral properties, still (let’s suppose) there is a possible world where moral properties exist and where St.

Francis is good. Now we ask, is there another possible world where moral properties exist and where an exact nonmoral duplicate of St. Francis is somehow evil? Is there a possible world with St. Francis’s evil twin?158 If so, then the moral fails to supervene on the nonmoral: two people that are exactly alike in their nonmoral properties are somehow different in their moral properties. But if not, then we need an explanation of why this collection of nonmoral properties, in every world where moral properties exist, goes with being good rather than with being evil. So the error-theorist ends up either denying supervenience or stuck with a worrisome explanatory burden.

One horn of the dilemma is centered on the idea of explaining supervenience relations. This is a particularly pressing concern for error-theorists, since the explanatory burden I mention is one that error-theorists would like to press against realists. Thus my argument is not just targeting the isolated thesis that moral properties possibly exist but

158 Note that, unlike typical evil twins, this evil twin has the same kind heart and gentle nature of St. Francis. It’s just that this perfectly pleasant collection of nonmoral properties somehow goes with being evil in that world. 125

do not actually exist. It is targeting the package of commitments typically held by error-

theorists. Thus before developing the argument in more detail, I will review some key

parts of Mackie’s classic presentation and discussion of his error-theory.

5.3 Mackie’s error-theory

Mackie holds that ordinary moral thought and discourse includes “a claim to

objectivity… to objective, intrinsic, prescriptivity”: values with motivational and/or

normative import holding independent of our desires. This much is revealed by

“linguistic and conceptual analysis”. However, such claims turn out false:

[T]he denial of objective values will have to be put forward not as the result of an analytic approach, but as an ‘error theory’, a theory that although most people in making moral judgments implicitly claim, among other things, to be pointing to something objectively prescriptive, these claims are all false.159 (Mackie 1977, 35)

He gives two arguments for his negative claim: the ‘argument from relativity’ and the

‘argument from queerness’. The first is based on disagreement: “the well-known

variation in moral codes from one society to another and one from one period to another,

and also the differences in moral beliefs between different groups and classes within a

complex community” (Mackie 1977, 36). This provides “indirec[t] support” for anti- realism by way of an argument to the best explanation: such variations “are more readily explained by the hypothesis that they reflect ways of life than by the hypothesis that they express , most of them seriously inadequate and badly distorted, of objective values”.160

159 Mackie (1977: 35). 160 Mackie (1977: 37). 126

The second argument ramifies twice over: into a epistemological branch and a

metaphysical branch, the latter into a complaint about motivation and/or normativity and

a complaint about explaining supervenience. The epistemological queerness is this: none of our ordinary faculties or methods of gaining knowledge are capable of making us

“aware of this authoritative prescriptivity”. Thus realists must ultimately turn to “a special sort of intuition”—the sort of rationalist-friendly intellect that might be required for knowledge of “some supposed metaphysical necessities or ” that Mackie the empiricist is equally willing to jettison.161

The metaphysical queerness is first a matter of motivation and/or normativity. Mackie

gives a picture of moral properties based (directly or indirectly) on Plato and Clarke:

Plato’s Forms give a dramatic picture of what objective values would have to be. The Form of the Good is such that knowledge of it provides the knower with both a direction and an overriding motive; something’s being good both tells the person who knows this to pursue it and makes him pursue it. And objective good would be sought by anyone who was acquainted with it, not because of any contingent fact that this person, or every person, is so constituted that he desires this end, but just because the end has to-be-pursuedness somehow built into it. Similarly, if there were objective of right and wrong, any wrong (possible) course of action would have not- to-be-doneness somehow built into it. Or we should have something like Clarke’s necessary relations of fitness between situation and actions, so that a situation would have a demand for such-and-such an action somehow built into it.162

It is unclear whether Mackie is making a claim about motivation or normativity or both.

He could be claiming that moral properties are such that belief in them is itself capable of

motivating us (independent of our desires)—a form of motivational internalism. Or he

could be claiming that moral properties (or relations) are such that they themselves land

us with normative reasons (“demand[s]”) to perform some action (again, independent of

161 Mackie (1977: 39). 162 Mackie (1977: 40). 127

our desires)—a form of moral rationalism.163 On either reading, he thinks that such

properties are too unusual to be believed: realism “involves the postulating of value-

entities or value-features of quite a different order from anything else with which we are

acquainted, and of a corresponding faculty with which to detect them”.164

The metaphysical queerness is also a matter of explaining supervenience. Mackie

writes:

Another way of bringing out this queerness is to ask, about anything that is supposed to have some objective moral , how this is linked with its natural features. what is the connection between the natural fact that an action is a piece of deliberate cru- elty—say, causing pain just for fun—and the moral fact that it is wrong? It cannot be an entailment, a logical or semantic necessity. Yet it is not merely that the two features occur together. the wrongness must somehow be ‘consequential’ or ‘supervenient’; it is wrong because it is a piece of deliberate cruelty. But just what in the world is signified by this ‘because’?165

It is unclear whether Mackie means to say that he cannot even understand what these

supervenience relations are supposed to be in the first place, or instead that these relations

are in some way too queer (too otherworldly) to be believed. But in his discussion of

moral arguments for theism in The Miracle of Theism, all signs point to the latter

reading.166 Mackie has just conceded that Plato (or, rather, Platonists like Cudworth) is right that God cannot create obligations by command—for any such creation would presuppose an more fundamental obligation to obey God’s commands. But he insists that there is still a good sense in which God can create morality: namely, God can create

163 I suspect he has the former in mind, considering his subsequent discussion of Hume’s motivational anti- rationalism in 2.3.3 of the Treatise. Joyce (2001), however, takes the latter to be the telling consideration against realism. For discussion of motivational internalism and moral rationalism, see Shafer-Landau (2003). 164 Mackie (1977: 40). 165 Mackie (1977: 41, emphasis original). 166 Mackie (1982). 128

supervenience relations. And this fact inspires a defensible argument for God’s existence:

First, supervenience relations are not grounded in analytic meaning-connections:

Now what is the logical character of this supervenience or making? Swinburne takes it to be analytic… But this cannot be right. Objective wrongness, if there is such a thing, is intrinsically prescriptive or action-guiding... But the natural features on which the moral ones supervene cannot be intrinsically action-guiding or reason-giv- ing in this way. Supervenience, then, must be a synthetic connection.167

Thus there must be some other explanation for the holding of these relations. He writes:

[W]e might well argue (borrowing, perhaps, from my own discussion elsewhere) that objective intrinsically prescriptive features, supervening upon natural ones, constitute so odd a cluster of qualities and relations that they are most unlikely to have arisen in the ordinary course of events, without an all-powerful god to create them. If, then, there are such intrinsically prescriptive objective values, they make the existence of a god more probable than it would have been without them. Thus we have, after all, a defensible inductive argument from morality to the existence of a god.168

Mackie, of course, rejects the existence of moral properties and thus rejects the argument.

But the argument sheds light on his complaint about supervenience. Since these relations

are not analytic, it would be intolerably odd for them to hold without further explanation.

And with God we have just the sort of explanation that might do the trick:

If we adopted moral objectivism, we should have to regard the relations of supervenience… as synthetic; they would then be in principle something that a god might conceivably create; and since they would otherwise be a very odd sort of thing, the admitting of them would be an inductive ground for admitting also a god to create them. There would be something here in need of explanation, and a [god] might well be the explanation we require.169

There are two lessons I draw from this review of Mackie’s error-theory. The first lesson

is that Mackie’s objections to realism about moral properties are not deep principled

objections to the very idea of moral properties, but instead the sort of objections that sit

167 Mackie (1982: 115). 168 Mackie (1982: 115). 169 Mackie (1982: 118). 129

well with the view that moral properties are at least possible. The argument from relativ- ity is an argument to the best explanation. The argument from queerness rests on a strong presumption of empiricism and naturalism. As Mackie sums up in The Miracle of

Theism, “[m]oral values, their objectivity and their supervenience, would be a continuing miracle in the sense explained in Chapter 1, a constant intrusion into the natural world.

But then our post-Humean scepticism about miracles will tell against this whole view”.170

Note that for Hume and for Mackie both, miracles are not impossible, just implausible.

The second lesson is that Mackie is unwilling to tolerate unexplained supervenience

relations. An explanation in terms of analytic truths or in terms of the creative acts of a

god will perhaps placate Mackie’s demand. But there must be some explanation.

Otherwise we are left wondering why these nonmoral properties go with those moral

properties.

With these lessons in hand, let’s now turn to my argument against error-theory.

5.4 The argument in detail

The argument looks to those worlds in which moral properties exist and poses a dilemma: in those worlds, does the moral supervene on the nonmoral or not? But what is it for the moral to supervene on the nonmoral? What is the thesis that the dilemma turns on? My answer is that it depends—it depends on the extent to which context matters. If the supervenience thesis in question concerns local supervenience, then context doesn’t matter: two individuals exactly alike in their nonmoral properties should be exactly alike

170 Mackie (1982: 118). 130

in their moral properties, even if the individuals are in vastly different contexts. Thus this

would rule out moral theories according to which the surrounding norms of one’s culture

or the commands made by some god can make a significant difference in the moral status

of something. One way of bringing context into the fold is to move to global

supervenience, according to which two worlds exactly alike in their distribution of

nonmoral properties should be exactly alike in their distribution of moral properties. But

this can seem like an overreaction, for global supervenience seems to allow for seriously

objectionable intra-world variations in moral status: e.g., a world might contain both an

Earth with a good St. Francis and a Twin Earth with an evil St. Francis, and global

supervenience would remain silent.171 It is thus a delicate matter to say what sort of

supervenience relation is appropriate for moral-on-nonmoral supervenience.

I think I can avoid these delicacies. I will run my argument with strong local

supervenience simply for ease of exposition. If the occasion warrants another variety of supervenience, then I will move to it. I think my argument will stand or fall independently of these matters. But for the sake of completeness, here is my canonical statement of the supervenience thesis:

For any possible worlds w1 and w2 such that w1 and w2 belong to the class of worlds in which moral properties exist, and for any individuals x in w1 and y in w2, if x in w1 is exactly alike to y in w2 with respect to every nonmoral property, then x in w1 is 172 exactly alike to y in w1 with respect to every moral property.

171 Of course global supervenience might not allow for intra-world evil twinnery. For, given the world I described, global supervenience definitely wouldn’t allow a world exactly like the first in its nonmoral properties where both St. Francises are evil—that would be a straightforward failure of global supervenience. And it might be that if the first world is possible, then the second one must be possible as well—after all, if one St. Francis can be evil, why couldn’t both be evil? 172 I use the standard possible worlds formulation rather than the standard model operator formulation because I do not want to explicitly rule out ‘free-floating’ moral properties. I do not think I need to, because 131

In other words, there are no individuals from any worlds where moral properties exist

that are exactly nonmorally alike and yet morally different. Leaving complications aside,

we can say that, if this supervenience thesis holds, the moral status of things doesn’t vary

from world to world: e.g., whether St. Francis is good or abortion is wrong or suffering is

bad doesn’t vary from world to world. Or, putting some complications back in, if the

moral status of things does vary, there had better be a difference in nonmoral properties

to go with the difference in moral properties. The heart of the supervenience thesis is that

there are no inexplicable variations (not even across worlds) in the moral status of

things—this may be a loaded way of putting it, but I do not think it is misleading.

Now the dilemma: does the supervenience thesis hold in those worlds where moral properties exist or not? If so, then these supervenience relations must be explained. If not, then we are left with inexplicable variations in the moral status of things. Either way, we are left with something deeply unsatisfying.

Take the first horn. Suppose the supervenience thesis does hold. Then there is a certain moral order holding throughout all worlds where moral properties exist. St.

Francis is (let us suppose) good in some possible world, and there is no individual in any possible world that is exactly like St. Francis in all nonmoral respects and yet evil. Let’s suppose abortion is wrong, and no procedure exactly like abortion in all nonmoral respects is morally permissible. Suffering, let’s suppose, is a bad thing, and nowhere is it a good thing. But we, we might ask, isn’t the other way around, with St. Francis evil,

the supervenience base here comprises every other property besides moral properties, and it is hard to imagine an object instantiating only moral properties. 132

abortion okay, and suffering a good thing? I contend that this moral order needs an explanation and should be not be simply tolerated: Mackie himself would say that it needs an explanation, anyone with any of the spirit of error-theory should agree, and in any case we are otherwise left with an unstable view.

First, we have already seen that Mackie himself has demanded an explanation for just such supervenience relations. This demand would not arise (or rather it would be easily satisfied) if the relations were analytic, but they are not. Mackie suggests that an explanation in terms of the creative acts of a god would succeed; we will examine this explanation in the next section.

Second, an error-theorist who steps back from all the possible worlds will notice that a certain moral order is in place in some of those worlds: St. Francis is good, abortion is wrong, suffering is bad. And this moral order is not counterbalanced by any other worlds in which St. Francis is evil, abortion is okay, and suffering is a good thing. So even though the actual world is neutral on moral matters, the collection of all possible worlds is not neutral. (It’s even pro-life!) This is, I think, clearly at odds with the spirit of error- theory.

Last, I suspect that no one—error-theorist or no—will be satisfied with a view according to which moral claims come out false due to the actual nonexistence of moral properties and yet there are facts as to which moral properties supervene on which nonmoral properties. For if there are such facts, then why aren’t they enough to vindicate the moral claims we make? If the supervenience relations governing those worlds with moral properties work out in such a way that suffering is a bad thing, isn’t there a real 133

sense in which the view that suffering is a bad thing is superior to the view that it is a

good thing? And if we can rank moral views in this way, don’t we have enough for a

workable, error-free, realist moral practice? Thus an error-theory with supervenience

governing possible moral properties is unstable: it threatens to collapse into realism.

Now take the second horn. Suppose the supervenience thesis does not hold. Then the

error-theorist has to swallow what I’ve called inexplicable variations in the moral status

of things: two things can have opposite moral statuses despite the fact that they are

exactly alike in every other respect. And this (as I hope is being brought out by my

loaded terminology) is highly implausible.

It’s not just that the supervenience thesis is plausible. It’s so plausible that it’s treated

as a matter of consensus: Mark van Roojen writes, “It is common ground among moral

theorists that moral properties supervene on non-moral properties. Two items cannot differ in their moral properties without differing in some non-moral property as well”.173

And it’s not just treated as a matter of consensus, it’s treated as a truth of quite privileged

status. Frank Jackson calls it “[t]he most salient and least controversial part of folk moral

theory”.174 R. M. Hare treats it as a constraint on proper use of the word ‘good’.175

Blackburn famously treats it as a matter of analytic necessity: “It seems to be a

conceptual matter that moral claims supervene on natural ones. Anyone failing to realize

173 Van Roojen (1999). 174 Jackson (1998: 118). 175 Hare (1952: 81). 134

this, or to obey the constraint, would indeed lack something constitutive of competence in

the moral practice”.176

There is only one defense I can think of on behalf of the error-theorist considering a rejection of the supervenience thesis. It is that error-theorists who deny actual moral properties but accept possible moral properties are already committed to rejecting the supervenience thesis. After all, of all the possible worlds with moral properties, there is presumably one just like the actual world in all nonmoral respects and yet bustling with moral properties. And if so, this is clearly a failure of supervenience (global supervenience and a fortiori strong local supervenience177): the worlds are exactly alike

nonmorally and yet morally different (since one world lacks moral properties altogether).

So any criticism of such an error-theorist is question-begging. It rests on a supervenience thesis that the error-theorist was already quite comfortable rejecting.

I have two sorts of reply to this defense. The first reply is to take this defense as just grounds for further criticism. It was bad enough that the error-theorist was driven by the dilemma I posed to deny the supervenience thesis. But now we see that the error-theorist was denying it from the start—and this just makes things worse! If it is success to drive error-theorists to absurdity, then it is an even greater success to point out that the absurdity was already there.

The second reply is less triumphantly self-satisfied. It is that the failure of supervenience the error-theorist was already committed to is less troubling than the

176 Blackburn (1993: 137). Note that expressivists don’t put supervenience in terms of properties, but in terms of constraints on thought and discourse. 177 I here rely on the plausible assumption that strong local supervenience entails global supervenience. 135

failure of supervenience showing up in the wake of the dilemma. The former failure allows for two individuals to be exactly alike in all nonmoral respects and yet for one of them to have moral properties and the other to lack moral properties altogether. Instead of good King Wenceslas and his evil twin, we have the good king and a twin king with no moral properties at all. This is troubling, perhaps, but an error-theorist might be used to seeing moral properties as an optional add-on, and perhaps we can explain why certain worlds would have and others would lack moral properties. But the latter failure allows for two individuals to be exactly alike in all nonmoral respects and yet different in their moral properties. Now we have evil twins on our hands—the same nonmoral properties somehow serve as the basis for directly opposed moral properties. This is the sort of supervenience failure that boggles the mind. Even if the error-theorist were comfortable with the former kind of supervenience failure, she should be loath to admit the latter kind.

5.5 Mackie’s proposed explanation of supervenience

As we have seen, in The Miracle of Theism Mackie suggests that the creative acts of a god might serve as a satisfactory explanation of supervenience relations. Since one of the horns of my dilemma concerns unexplained supervenience relations, it is worth examining what Mackie says, to see if he can escape the dilemma.

Mackie considers three objections to the argument for theism. One is that instead of seeing natural properties tied to reason-giving moral properties, we should see natural properties as themselves reason-giving. Mackie responds that we then need an explanation of why natural facts give the reasons they do. Another is that the very notion 136

of moral properties is incoherent (Mackie cites Hare). Mackie responds that he has

argued that they are not (see above). Finally, one might wonder why to posit a god as an

explanation. Mackie responds that serious mysteries call for explanations with serious power, that the alleged distribution of moral properties fits the purposes of a benevolent god, and that objective to-be-done-ness calls for “some analogue of human purpo- siveness”.178

Having dispensed of these three objections, Mackie writes:

There are, nevertheless, some difficulties for the proposed argument. If we put it in terms of supervenience, are these synthetic truths necessary or contingent? Do they hold in all possible worlds or only in some? Are there other worlds in which there are quite different truths of supervenience—so that radically different sorts of actions are right or wrong—or none at all? If the range of possible worlds is to cover all logical possibilities, there must be such variations. So a rational agent has not only to identify the natural situation in which he finds himself and reason about it; he has also to ascertain which of various possible worlds with regard to moral supervenience is the actual one—for example, whether the actual world is one in which pain is prima facie to be relieved, or one in which, other things being equal, pain is to be perpetuated.179

He continues by saying that “this problem only brings into the open the intuitionist moral

epistemology” implicit in moral realism. Moral knowledge is acquired not by “general

reasoning based on natural facts”, but by intuition: “the moral thinker has, as it were, to

respond to a value-laden atmosphere that surrounds him in the actual world”.180

How should we take Mackie’s proposal? At first blush it seems that Mackie is

denying supervenience. After all, he allows for radical variations in the moral status of

things from one world to the next: pain might be a bad thing in one world and a good

178 Mackie (1982: 116–7). 179 Mackie (1982: 117). 180 Mackie (1982: 117). 137

thing in another. It appears that he is explicitly taking the second horn of my dilemma, re-

jecting supervenience and embracing inexplicable variation.

But in fact he is not denying supervenience. More precisely, though he is denying

supervenience of the moral on the natural, he is not denying supervenience of the moral

on the nonmoral, and it is the latter that has occupied our discussion. After all, if we take

the world where pain is good and the one where it is bad, we’ll notice that this variation

in moral properties is backed up by a variation in nonmoral properties: namely, God has

performed different creative acts in the different worlds. These divine acts are included

within the supervenience base of the moral, and thus the variation admitted by Mackie is still acceptable by the lights of a global supervenience thesis. (Admittedly, he is denying

local supervenience theses: individuals might differ in their moral properties, not due to

any divergence in their nonmoral properties, but due to different divine acts operating in

the background. But we can easily shift to a global thesis for this part of the discussion.)

But if Mackie is accepting supervenience, then shouldn’t be impaled on the first horn

of my dilemma? Shouldn’t he be stuck with unexplained supervenience relations? If so, it

would be an ironic end, considering that this was supposed to be a clear example of how

to explain supervenience relations.

In fact, I think Mackie does meet with this ironic end. Mackie takes for granted that certain moral properties will supervene on certain creative acts of God. A world where

God creates this-a-way will always involve certain patterns of instantiation of moral properties; there is no world where God creates in just that way and yet there are different patterns. But we can always ask for an explanation of just these supervenience relations. 138

If there is a world with creative act A and goodness supervening on pain, and a world with creative act B and badness supervening on pain, then why isn’t there a world with A

and badness on pain, or B and goodness on pain? This problem is just a consequence of

the fact that God’s creative acts are part of the nonmoral, the supervenience base of the

moral. It is difficult to see how something included in that base can explain the moral-on-

nonmoral supervenience relations; it seems we can always ask why these base properties goes with those moral properties.

Matters are still more ironic for Mackie, for this point is a mere updating of the classic Platonist point found in the Euthyphro and Cudworth’s Treatise concerning

Eternal and Immutable Morality: just as God’s commands cannot create obligations without presupposing a more fundamental obligation to obey God, likewise God’s creative acts cannot create supervenience relations without presupposing a more fun- damental supervenience relation with God’s creative acts included in the base. Thus

Mackie’s attempt frame a defensible way for God to create morality fares no better than the discredited alternative, and for exactly the same reason.

So in the end, Mackie is stuck with unexplained supervenience relations. Admittedly,

Mackie’s solution does help with the latter two problems I raised for those taking the first horn of the dilemma. The collection of possible worlds is still neutral on specific moral

issues (with plenty of variation across possible worlds), and thus the threat of a workable

error-free realist moral practice goes away. But his attempt to explain supervenience

relations only presupposed further supervenience relations themselves in need of

explanation. 139

5.6 Error-theories in other domains

Morality is not the only domain where an error-theory has been pursued. Philosophers

have defended error-theory in the philosophy of mathematics and the metaphysics of

color, and atheism in the philosophy of offers a fair parallel. Moreover, it seems

a safe assumption that error-theory, or something quite close to it, would apply to less

philosophically controversial domains: e.g. witchcraft, phlogiston, astrology. Thus it

would be unfortunate if my argument against error-theory were so indiscriminate as to

apply equally well in all domains. Now, it would be far too difficult and lengthy a task to settle the prospects for error-theory and the fate of my argument in other controversial philosophical domains here. But I will attempt to show that my argument poses no threat to those easier cases where error-theory seems so plausible.

The foothold for my argument in the case of morality is the supervenience of the moral on the natural—or, as we have just seen, the supervenience of the moral on the nonmoral. I take this to be a conceptual constraint on moral evaluation, and the argument goes through with a good deal of force even if we treat it as only a central metaethical desideratum. Thus it ought to make a difference to the applicability of the argument whether a similar supervenience constraint can be counted on for e.g. witchcraft, phlogiston, and astrology. If so, then the argument is dangerously indiscriminate. If not, it has a safely limited applicability.

I think it is fairly clear that no such supervenience constraints can be counted on. If someone thinks different worlds with witchcraft could have different ‘laws of magick’ relating thaumaturgical to natural facts, so that reciting the words ‘Klaatu barada nikto’ 140

retrieves the Necronomicon in one world but revives dead loved ones in another, they are not guilty of conceptual incompetence when it comes to witchcraft. The same goes for astrology: a conceptually competent astrologer can allow for cross-world variations in the effects of celestial bodies on human affairs, so that one world’s “Mars effect” might bring athletic eminence and another world’s might bring excellence in scrapbooking. Nothing in the concepts of witchcraft or astrology serves to tie them necessarily to any particular natural states of the world.

The case of phlogiston is slightly different: arguably it would be conceptually incompetent to allow worlds to be exactly alike other than certain differences in matters phlogistic. But this is only because it is arguably part of phlogiston theory that it have a specified relation to other parts of the physical world. And then the particular supervenience constraint that governs phlogiston can be given an explanation not available in the case of morality: it follows from the very theory of phlogiston that fixes the concept, whereas no particular moral-natural supervenience constraint belongs to moral concepts. In short, phlogiston admits of something like the analytic truths that

Mackie (rightly) rejects in the case of morality.

5.7 Moral properties as impossible?

The argument thus far has turned on the supposition that error-theorists are committed to the metaphysical possibility of moral properties. But this supposition might be rejected.

Error-theorists, after all, are only committed to the descriptivist thesis that moral judgments are fact-stating judgments which purport to assert the existence of moral 141

properties. And one can accept this thesis without also accepting that moral properties are

metaphysically possible. So, to complete my argument, I need to consider the alternatives

available to error-theorists.

The most obvious alternative is that moral properties are conceptually possible but

not metaphysically possible. This space is opened up by the case of natural kinds: it is

conceptually possible that water is not H2O (for there is no conceptual mistake in

thinking this) but it is not metaphysically possible that water is not H2O (for water is in

fact the same thing as H2O). Likewise, perhaps it is conceptually possible that moral

properties exist, but not metaphysically possible that they exist. If error-theorists adopted this view, they could avoid my supervenience argument, because there would be no metaphysically possible worlds in which moral properties exist.

But this view has problems of its own. For one thing, in the case of water and H2O,

the conceptual possibility still seems to correspond to a metaphysical possibility. That is,

the conceptual possibility in which water is not H2O corresponds to the metaphysical

possibility that something other than H2O is in the speakers’ environment, playing the

watery-stuff role or just standing at the other end of a chain of reference. To be sure, this

metaphysical possibility is not one in which water is not H2O. That would be a

misdescription of a possibility better described as one in which something relevantly

similar to water is not H2O: fool’s water. But it is a metaphysical possibility

nonetheless.181

181 Here I am obviously drawing on Kripke (1980). 142

How does this point bear on error-theory? It means that, for the parallel between moral properties and natural kinds to hold good, the conceptual possibility in which moral properties exist must correspond to some metaphysical possibility which something relevantly similar to moral properties exist: fool’s moral properties. That is, error-theorists must recognize the metaphysical possibility of something much like moral properties, while maintaining that moral properties themselves are metaphysically impossible.

This, it seems, is a heavy burden to meet. In the case of natural kinds, we have some idea of how a superficially similar natural kind might serve to ground a possibility claim which is conceptually possible but not metaphysically possible. After all, two natural kinds can be superficially similar while differing in their underlying nature. But it is not clear how to apply this model to moral properties. What are the superficial similarities, and what are the underlying natures? What must genuine moral properties and fool’s moral properties have in common, and how must they differ?

But suppose that the burden could be met. Suppose, in other words, that there are metaphysically possible fool’s moral properties that are relevantly similar to genuine moral properties, so that moral properties are conceptually possible but not metaphysically possible. Now, one conceptual truth concerning moral properties is their supervenience upon natural properties. But this means that, if they are to be conceptually indistinguishable from genuine moral properties, fool’s moral properties must also supervene upon natural properties. And this marks the return of the above supervenience argument. Since these properties supervene upon natural properties, we should have these 143

properties in the actual world, in which case it is unclear why they do not qualify as

genuine moral properties. And if we waive this requirement and allow them to embody a

supervenience only in those worlds in which they exist, there is still a

supervenience pattern that needs to be explained.

Error-theorists might then back off of the analogy to natural kinds, while still

maintaining the distinction between the conceptual and the metaphysical possibility of

moral properties. This would relieve them of the burden of identifying a metaphysically

possible kind of property that is relevantly similar to moral properties. Here a natural

enough analogy might be atheism about the God of classical theism.

Atheists about this God must do more than deny its actual existence. They must

contend that God is metaphysically impossible. This is due to the traditional conception

of God as a metaphysically necessary being. Due to this special status, either God exists

in all metaphysically possible worlds or in no metaphysically possible worlds. And this

means that, if God exists in any metaphysically possible worlds, then God actually exists

and atheism is false. This is the leading idea behind the modal ontological arguments of

the 20th century,182 and it effectively polarizes the theism-atheism debate. Instead of arguing over the actual world only, both sides are arguing over the entire range of metaphysically possible worlds.

But even though atheists must maintain that such a God is metaphysically impossible, they might still allow its conceptual possibility. For not all atheists contend that the God of classical theism is a conceptual absurdity. On the contrary, many allow that such a

182 See e.g. Plantinga (1974). 144

God makes good conceptual sense, but then appeal to Occam’s razor or the problem of

evil as justification for atheism.

The analogy with error-theory seems clear enough. Error-theorists can appeal to e.g. the argument from queerness as a justification for error-theory, deny that moral properties are metaphysically possible, all while allowing that moral properties are conceptually possible. And if error-theory is no more problematic than atheism, then surely that is

more than good enough for your average error-theorist.

But once again the analogy breaks down. The reason is that if something is conceptually possible but metaphysically impossible, without even any relevantly similar metaphysical possibilities to explain the conceptual possibility, then some explanation is needed. Now, in the case of God, an explanation can be found. For the very concept of the God of classical theism builds in a special modal status that determines it to exist in all metaphysically possible worlds or none. This is why atheists are driven to maintain the metaphysical impossibility of God’s existence. Moreover, pace Descartes, the concept of God draws on our familiarity with power and knowledge and wisdom and benevolence in the actual world. And this, whether through Aquinas’s “analogy” or

Locke’s “enlarging”, provides descriptive content capable of supporting the claim of conceptual possibility. But with moral properties, and normative properties generally, the difficulty of finding plausible descriptive content is by now evident. And without overcoming this difficulty, error-theorists cannot look to the case of God for relief.

Thus error-theorists who wish to avoid the supervenience argument of this chapter are left with the problem of impossible content. Without a way of understanding how moral 145

judgments could be beliefs whose descriptive content refers to metaphysically impossible properties, error-theorists who deny the possibility of moral properties will need to develop an alternative form of their view. For once they abandon descriptive content referring to moral properties, they effectively abandon error-theory and join expressivists in providing a nondescriptive account of moral judgment.

146

CHAPTER 6: FICTIONALISM

Fictionalism is the most prominent view that couples purely descriptive moral content with an attitude other than belief. In recent metaethical discussion, two forms of fictionalism have emerged. First, there is the revolutionary fictionalism of Richard Joyce: taking inspiration from error-theory, Joyce argues that the best move for a community that realizes the truth of error-theory might well be to carry on with a revised version of moral evaluation that involves something more like make-believe than belief.183 Second,

there is the hermeneutic fictionalism of Mark Kalderon: taking inspiration from

expressivism, Kalderon argues that a better interpretation of moral evaluation might agree

with expressivists that moral judgments are not beliefs but then insist that these non-be-

lief acceptances are of ordinary descriptive content, thereby avoiding the Frege-Geach

problem.184 Both views end up with a similar picture of moral evaluation, but with Joyce offering the picture as a practical proposal for revising our current practices, and

Kalderon offering the picture as an interpretive proposal for understanding our current

practices.

6.1 Joyce’s fictionalism

Since we have just examined error-theory, I will begin with Joyce’s revolutionary

fictionalism. Indeed, I will a press a similar dilemma about the purely descriptive content

that we are alleged to mistakenly believe and that we are advised to carry on make-

183 Joyce (2001). 184 Kalderon (2005). 147

believing: viz., does this content manage to refer to some possible ontology or not? If so,

then Joyce’s view inherits the burden of providing a semantic account capable of

explaining this (possibly successful but actually unsuccessful) reference. If not, then

Joyce must defend the descriptivist element of his theory against the problem of

impossible content. Joyce shows signs of attraction to both horns of the dilemma, and so

both horns call for direct engagement with his specific development of error-theory-cum-

fictionalism.

First horn: Joyce’s normative realism

The first horn—i.e., reference to possible ontology—shows itself in two ways: in Joyce’s

realist account of normative reasons, and in his acceptance of substantive “minimal

constraints” on moral content. At first blush, Joyce’s normative realism might be

somewhat surprising for an error-theorist, but in fact it lies at the heart of both his argument for error-theory and his fictionalist proposal. Recall that the first branch of

Mackie’s metaphysical queerness argument features an ambiguity between two different

kinds of queerness: whether moral properties are intrinsically motivating or categorically

binding. Joyce thinks that the debate over whether morality is committed to any

motivational queerness reaches an “impasse”,185 and in any case interprets Mackie as

leaning more on morality’s commitment to strong categorical reasons. And for Joyce, it is

this commitment that represents the Achilles heel of morality.186

185 Joyce (2001: 27). 186 Joyce (2001: 30–52). 148

As he develops an account of what exactly these strong categorical reasons are

supposed to be, Joyce finds two contrast classes. First, there are the institutional reasons

of e.g. etiquette or gladiatorial combat. These ‘reasons’ have something second-rate about

them, because they are not “binding”. We may legitimately speak of the institutional rea-

sons someone has, and thereby express our allegiance to the institution in question, but

we must acknowledge that this someone “may legitimately ignore” the would-be reason.

Whereas moral reasons are supposed to have an inescapable bindingness, institutional

reasons (even the categorical ones) lay no claim to true normative authority or

bindingness.187 Second, and more importantly, there are the reasons of practical

rationality, such as the reasons provided by a hypothetical imperative. These reasons are

supposed to be inescapably binding, and do lay claim to true normative authority, and

Joyce is quite happy to recognize their authority. The only question is whether practical

rationality can be shown to accommodate not only instrumental and prudential reasons, but also the categorical reasons of morality.188 And here Joyce’s error-theory shows up: there are no moral reasons, because there are no categorically binding reasons, only categorical non-binding reasons (e.g., etiquette reasons) or non-categorical binding

reasons (e.g., means-end reasons).

But how does Joyce’s normative realism bear on his fictionalism? It is that the fictionalist proposal is not a mere recommendation, but a recommendation backed up by true normative authority. It is not just being raised as an interesting possibility, but argued for as something that we genuinely ought to do. Now, it would of course be inconsistent

187 Joyce (2001: 34–41). 188 Joyce (2001: 35–39, 44–48). 149

with his error-theory for Joyce to insist upon a moral ought. Instead, then, it is “just a

straightforward, common-or-garden, practical ‘ought’” based on the benefits promised by

carrying on with a fictionalist revision of moral evaluation, as opposed to jettisoning the

practice in toto.189 Thus the normative realism about the reasons of practical rationality gives Joyce a safe place from which to support his fictionalism with arguments that can

deliver genuine oughts.

Now, as I have argued in my earlier critique of the Humean theory of reasons

defended by Mark Schroeder, there is good reason to be skeptical of the notion that

hypothetical imperatives are any safer or less mysterious than categorical imperatives.

Even if hypothetical imperatives do not implicitly contain categorical reasons, they still

raise the very same important metaethical questions about how normative concepts can

be understood to refer to robust normative facts. Supposing it true that anyone ought to

take the means to their ends, we need an explanation of this truth, and an error-theorist

needs a way of explaining it without positing the very otherworldly normative facts the

view is defined against.

And indeed, in a brief defense of Bernard Williams’ reasons internalism against the

critique of Jean Hampton,190 Joyce makes a move similar to Schroeder’s. The worry he

faces is that reasons internalists will be forced to accept the following conditional sort of

external reason: a reason for ϕing-if-a-fully-informed-and-correctly-deliberating-counter- part-of-yourself-would-want-you-to-ϕ. This alleged reason would apply to you regardless

of the contents of your subjective motivational set, and would therefore count as an

189 Joyce (2001: 177). 190 Joyce (2001: 115–23). 150

external reason.191 Likewise, Humean instrumentalists might be forced to accept a reason

for ϕing-if-you-want-X-and-ϕing-is-the-best-means-of-achieving-X, a reason which looks

like it would be categorically binding on all agents with all manner of personal desires.

Joyce answers this worry by distinguishing between different readings. For reasons

internalism, we can distinguish “If S+ wants S to ϕ, then S has a reason to ϕ” from “S has

a reason to ϕ-if-S+-wants-S-to-ϕ”. For Humean , we can distinguish

“You ought to (ϕ if you want X and ϕing is the best means of achieving X)” from “If (you

want X and ϕing is the best means of achieving X), then you ought to ϕ.” Since the first

readings provide internalists and instrumentalists with everything they want, all without

being of the categorical form “You ought to ϕ”, then they can breathe a sigh of relief.192

But this does not get Joyce out of the woods, for plenty of apparently categorical

moral imperatives can be put in if-then form. To keep it simple, we can stick with imperatives of the form “If __ want(s) X and your ϕing is the best means of achieving X, then you ought to ϕ”. Some Confucians might fill in the blank with ‘your father’, some nationalists might fill in the blank with ‘your homeland’, and some patriarchal mi-

sogynists might fill in the blank with ‘your husband’. And the same point holds if we

move from a simple Humean view to the sort of non-Humean (but still internalist) view

that Joyce defends: the blank in “If an idealized version of __ would want you to ϕ, then

you have a normative reason to ϕ” can be filled in with lots of different people. Thus,

even if an imperative rests its ‘ought’ upon the condition of certain circumstances

obtaining, even circumstances having to do with someone’s wants, this alone does not

191 Joyce (2001: 119). 192 Joyce (2001: 121–22). 151

keep it from being categorical. But more importantly for our purposes, this alone does not

keep it from needing a metaethical explanation.

I can find two points of Joyce’s that might be used to provide some sort of

explanation. The first is that Humeans can say it is simply a matter of definition. Joyce imagines someone pressing a Humean on implicit categorical commitments:

But the person continues with a second question: “Why does the mere fact that I desire something give me a reason for performing an action that will lead to its satisfaction?” Now at this point what the Humean should not do is trot out a more general imperative: “Because you ought to follow this principle: if you desire X, and ϕing is the best means of achieving it, then you ought to ϕ.” This would be futile, for the mere fact that the person needed to ask the second question shows that she is not going to accept this as an answer. Rather, the Humean thinks that the questioner does not properly understand what the word “reason” means. He needs to provide her with tuition, perhaps beginning by asking her what she understands a good reason to be.193

Now, recall that Joyce’s discussion of institutional reasons recognized a difference

between the legitimacy of reasons-talk and the binding and authoritative character of

strong reasons (or ‘real’ reasons, as he sometimes calls them): we can of course say that

you have a reason to use the small fork, but you have every right to ignore this

lightweight ‘reason’. But the above discussion is clearly about the latter question: i.e.,

whether, as Humeans contend, mere personal desires give binding reasons invested with

normative authority. And so while it may not be conversationally worthwhile for our

Humean to trot out that more general imperative, this does nothing to clear Humeans

from the burden of accounting for the binding authority of that imperative. Nor can this

burden be evaded by asking which normative reasons the interlocutor does accept, for the

interlocutor may deny binding and authoritative reasons altogether.

193 Joyce (2001: 112–13). 152

Given Humeanism’s strong normative commitments, it is not clear how an education on the word ‘reason’ could meet this burden: people applying the word ‘reason’ in

Humean circumstances does no more to justify any associated claims of normative authority than people applying it in Confucian/nationalist/misogynist circumstances would justify any associated claims of normative authority. And in any case, claims of direct synonymy call for direct Moorean tests. Could one coherently ponder the question of whether personal desires alone can provide binding reasons to take the means to those desires? Could someone take the view that personal desires alone do not provide binding reasons (perhaps because they think binding reasons cannot originate in our animal nature, or perhaps because they think our desires have been radically corrupted by original sin, or perhaps because they deny that there are any binding practical reasons at all)? And, finally, if one of these anti-Humeans got into a disagreement with a Humean about what reasons people have, would this be a genuine normative disagreement rather than a trivial disagreement in meaning? The answer to all these questions is apparently

Yes, with roughly the same measure of apparent correctness as with the corresponding questions in the moral case (i.e., addressed at those direct synonyms proposed for moral terms by simple reductive moral realists). This kind of explanation is thus not a promising one.

The other possible explanation appeals to a special status allegedly held by practical rationality: according to Joyce, it cannot be legitimately questioned. He defends this thesis in confronting Carnap’s famous quietism about external ontological questions. For

Carnap, we can ask legitimate theoretical questions about what things there are only from 153

within a “linguistic framework”: e.g., asking whether there are any even primes from

within the framework of mathematics. But when we try to step outside the framework

and ask higher-order philosophical questions about the ontological commitments of the

framework as a whole, we are either asking practical questions about the value of the

framework or instead asking pseudo-questions that do not make sense. Thus there may be

sense to be made of morally engaged questions about what moral facts there are, but there

is no higher-order metaethical sense to be made of moral facts. This would be bad news for Joyce, inasmuch as he (like any error-theorist) wishes to provide an external critique

of the ontological commitments of the moral framework as a whole.194

Joyce’s solution is to rely on practical rationality as the broad framework within

which internal questions can be pressed against morality as whole. To make sense of the

alleged moral facts to be critiqued, we can draw on our understanding of practical

rationality, and thus “the non-practical question—‘Are there really such things as moral

obligations?’—is a sensible one”.195 He continues:

This is because there is no bizarre or alien notion of “existence” being appealed to— whatever kind of existence moral obligations have, it is a type which we are familiar with from other frameworks already accepted. Moral obligations exist, I have argued, only if reasons for action exist, and reasons for action are something which we are accustomed to quite independently of the moral framework. This wider non-moral notion of having a reason comes from practical rationality.196

Now, of course, this response is effective only if the reasons for action of practical

rationality are themselves unproblematic. And Joyce is happy to say they are: “Outside

morality we know very well what reasons are: most conspicuously, we have a normative

194 Joyce (2001: 45–51). 195 Joyce (2001: 47–48). 196 Joyce (2001: 48). 154

framework of giving and accepting reasons for acting known as ‘practical rationality’.”197

And this is where his claim about the special unquestionable status of practical rationality

enters the picture.

Practical rationality, as Joyce understands it, is all-encompassing when it comes to

reasons for action, delivering the all-things-considered thing to do: “Practical reasons are, by definition, those which guide our actions when everything has been taken into account”.198 And this means there is no danger of some competing normative system trumping the dictates of practical rationality when it comes to action: whatever normative system wins, wins only because practical rationality has authorized its victory. But since

practical rationality is all-encompassing, it follows that an acceptance of its authority is

implicated in any practical question: to ask “why should I care about that? why do that?

what is that to me?” is to ask for a reason for action, which means the question is posed

from the standpoint of practical rationality. And if all practical questions only testify to

the authority of practical rationality, then no practical question can intelligibly question

this authority. As Joyce writes: “Can we imagine someone questioning practical ra-

tionality: ‘Yes, I recognize that there is a practical reason for me to ϕ, but what is that to

me?—Why should I adopt that set of rules?’? This, it seems to me, is incoherent”.199

Now, one obvious way to challenge this argument is to find a normative system that

can compete with practical rationality without being neatly subsumed under it. Here the

most promising candidate is epistemic rationality. There are plenty of stock cases where

197 Joyce (2001: 53). 198 Joyce (2001: 50). 199 Joyce (2001: 49). 155

what would be most epistemically reasonable for someone to believe comes into conflict with what would make their life go better (e.g., marital infidelity cases), and it is highly controversial whether epistemic oughts can be understood in terms of instrumental rationality or prudence, much less whether they must be so understood. Consider the extreme case where someone places such a high premium on intellectual integrity that they would sacrifice all practical value in its name. In a classroom discussion of Pascal’s

Wager, a student of mine once said he would refuse to take steps to form an unjustified belief about something completely trivial (I think the example was whether the trillionth digit of pi was odd or even), even to avoid eternity in hell. I do not know whether the student was sincere, and I certainly do not share his priorities, but the case still suggests that there is something intelligible about the thought that sometimes considerations of practical rationality should be completely ignored.

But I think it is more important to object to the relevance of the argument. Even if practical rationality cannot be challenged by a practical question without its authority being implicitly accepted by the questioner, this does nothing to vindicate its authority.

And even if we have a normative framework known as ‘practical rationality’, this does nothing to shed light on the concept of a normative reason. I’ll develop this objection as a dilemma between thick accounts of practical rationality and a thin account.

On the first horn, Joyce might understand practical rationality as including substantive claims about our normative reasons: e.g., that we have instrumental reasons or prudential reasons. But then his claim about its special unquestionable status comes out false. There is nothing unintelligible about asking “Yes, I recognize that there is an 156

instrumental reason or a prudential reason for me to ϕ, but what is that to me?—Why should I adopt that set of rules?” This is because there is nothing unintelligible about thinking that sometimes your own desires or your own well-being do not matter. Indeed, this sort of approach would seem to be a case of theft over honest toil: a moral realist might as well include substantive claims about the moral reasons we have. And so this horn offers no way of making sense of the particular kinds of reasons he takes to be unproblematic.

On the second horn, Joyce might understand practical rationality as nothing more than a convenient catch-all category for whatever reasons for action remain once all reasons for action have been considered, with no substantive claims included. This would vindicate the idea that practical rationality has a special unquestionable status. But it does not offer him a steady framework from which to criticize morality. That is, he cannot use it to criticize morality, because the framework is so thoroughly ecumenical that nothing in it could ever favor hypothetical imperatives over categorical imperatives, egoistic reasons over altruistic reasons, etc. His overall strategy of treating instrumental/prudential

reasons differently from moral reasons, with the former deemed an unproblematic

starting point of practical rationality and the latter deemed an unrealistic pipe dream,

could never even get started. Moreover, he would not even have a steady framework,

because so thin an account tells us nothing informative about normative reasons. To be

sure, we have some platitudinous material, such as that all practical questions asking for

practical reasons are posed from the perspective of practical rationality. But there is

nothing capable of explaining why any particular normative claim or theory of practical 157

reasons would be true or false, nothing that would lend support to realism about practical

rationality, or even of illuminating the allegedly descriptive content of normative claims.

Thus the wider domain of practical rationality cannot help us understand the allegedly

descriptive content of moral claims. Contra Joyce, we are left with no idea what reasons

are, or any understanding of how normative thought might make reference to some robust

ontology. If the problem with morality is that there is no making sense of categorically

binding reasons, then it is unclear why hypothetically binding reasons are any better off,

and an appeal to practical rationality as an unquestionable domain of what to do makes

things no clearer. But if there actually are hypothetically binding reasons, then it would

be hard to maintain without serious additional argument that categorically binding

reasons are impossible: in which case, the argument against error-theory from the

previous chapter begins all over again.

First horn: the ‘story of morality’

So much for the first way in which reference to possible ontology shows itself. The

second way comes up when Joyce is discussing the descriptive content that is ‘accepted’

in a fictionalist community, and noting the analogies and disanalogies with fictional

acceptance of an ordinary fictional story. One difference is that “the story of morality”

doesn’t have an author who created it as a fiction, and another is that the story of morality

doesn’t consist of a well-defined body of propositions. Nevertheless it does have some content:

I have in mind that [this content] need consist primarily of a few general existential claims such as the following: “There are obligations and prohibitions: things that we 158

ought to do, and things we ought not do, regardless of whether it suits us or furthers our interests,” “People have basic rights, regardless of whether any institution recognizes those rights,” “People have character traits—such as courage or dishonesty—which explain their actions and by which actions may be appraised,” “Wrong-doers deserve punishment,” etc. We should also include some particular claims, such as “Torturing babies to pass the time is always wrong,” which will provide some minimal constraints on the content of the obligations, rights, etc. The important thing to note is that at the “core” of morality is a cluster of key notions like obligation, right, virtue, etc., and to accept morality is to accept this . It does not follow that two moral fictionalists can come to agree over a particular case—whether a second-trimester abortion is permissible, for example— simply by consulting the “story of morality,” in the way that two Holmes fans may consult the canonical texts in order to settle a dispute about Watson’s war wound. The fiction of morality is in many ways “open,” but I suggest that by containing relations among its posits (between obligations and rights, for example), and some general constraints (such as “Painful consequences are relevant to right action”), it contains rough internal rules of disputation.200

Now one thing to note immediately is that the descriptive content of a fictionalist’s moral judgment is supposed to be the same as the descriptive content of a realist’s moral judgment. The difference is in the attitude: the (wised-up) fictionalist takes up an attitude somewhat like make-believe, and the (sadly mistaken) realist an attitude of genuine belief.

Next, note that this content includes substantive, and indeed controversial, moral claims. Certain Kantians would deny that painful consequences are ever relevant to right action, and Kant himself would at least deny that they are always relevant to right action.

We can guess what Bentham would say about basic pre-institutional rights, or about punishment as something deserved by wrong-doers as such. But if these claims belong to the very content of the story of morality, then we have to conclude that neither Kant nor

Bentham were engaging in moral evaluation. And this is fairly implausible.

200 Joyce (2001: 194–95). 159

But then the promise of informative descriptive content, capable of explaining reference to possible ontology, vanishes. The more information is built into the descriptive content of moral concepts, the more likely it is that we can explain why they manage to refer to this ontology rather than that. And while something as ecumenical as

“there are obligations” is unlikely to slim down the list of eligible referents by much, stronger and more substantive moral claims might be just the ticket. But with stronger moral claims come implausible constraints on moral evaluation: anything strong enough to explain reference ends up excluding important moral thinkers like Kant and Bentham.

And so the ‘story of morality’ is not well-suited to solve the semantic problem.

Second horn: impossible content

The other horn for Joyce’s revolutionary fictionalism is that the purely descriptive content of moral judgments does not manage to refer to any possible ontology. And this is a view that Joyce apparently has some attraction to, or at least wishes to make room for. Thus he considers what would follow if moral evaluation really were committed to some strong form of motivational internalism which deems amoralists a priori impossible. And he finds that the falsehood of such a strong internalism would, due to its modal status, not only keep moral judgments from being actually true, but even possibly true. But this impossibility would not cause trouble for the error-theorist’s interpretation of moral evaluation as centering on purely descriptive content. Error-theory can get by with impossible content.201

201 Joyce (2001: 27–28). 160

But this is a special case. Here the descriptive content is stipulated to carry a commitment to a modalized descriptive claim about motivation and moral judgment. And unfortunately, it is not that easy to tease out such commitments from the practice of moral evaluation, so as to reveal descriptive content with clear and identifiable truth-conditions.

An error-theorist can say that the descriptive content of moral evaluation is impossible, but this sheds no light on the descriptive content in the absence of further work: e.g., saying something about the constituent elements of the content and how they combine to generate an impossibility.

With this in mind, consider another possibility. Perhaps morality is not “committed to some false thesis”, but is instead “so nebulous and fragmented that its judgments do not fulfill the minimal criteria for being accorded truth value”. Joyce points to Wittgenstein’s claim that moral evaluation is swimming with similes on the surface but revealed to be nonsense once the similes are dropped, and also to the images conjured up by Bentham and Mackie (e.g., “a man lying down with a heavy body pressed upon him”, an “invisible cord”) in an attempt to capture the ‘feel’ of obligation. Now, images and similes do not seem to be a clear case of descriptive content, and indeed they are often resorted to when trying to indicate distinctive moods and emotions that resist descriptive analysis. Yet

Joyce seems satisfied that such an account of moral evaluation should be classified as an error-theory, showing us that perhaps “there need be nothing philosophically troubling about the idea of our believing and asserting impossible things”.202

202 Joyce (2001: 28–29). 161

But what error could there be in experiencing the distinctive ‘feel’ of obligation? And what error could there be in associating this feel with certain similes and images? To be sure, one might dislike the feel of obligation and seek to discourage others from giving into it. But this would be a far cry from a metaethical error theory. Joyce has acknowl- edged that such a view could not assign truth-values to moral judgments. But then he is in dire need of a conception of descriptive content and the error it might fall into that can apply to this kind of case, overturning the apparently plausible interpretation of ‘feel’ as a matter of nondescriptive attitudes.

These two cases illustrate the general problem of impossible descriptive content. One might provide an account of morality’s descriptive content that provides it with clear and identifiable truth-conditions that lead to a contradiction or some other impossibility. But while this would be plausibly descriptive, and plausibly impossible, it is difficult to find any such content in the terms and concepts of moral evaluation. Alternatively, one might find some nebulous ‘feel’ that is plausibly essential to moral evaluation; but then it is difficult to sustain the claim that this ‘feel’ amounts to impossible descriptive content.

Either way, the error-theorist is faced with a heavy burden to discharge.

Joyce’s later examples of impossible content in a fictionalist practice do not overcome this problem. While “a child may pretend that the candlestick walks and talks

(while remaining a candlestick)”, a mere physical impossibility neither avoids the need for a semantic account to handle other possible worlds nor plausibly belongs to our moral concepts. Likewise, while “the Holmes fan may pretend that the stories are true, despite containing contradictory references to the position of Watson’s war wound”, it would be 162

difficult to show that moral evaluation carries such directly contradictory elements within

its commitments. Certainly, “one might pretend that water is XYZ though it is necessarily

H2O”, but no extant account of the semantics of natural kinds has proved a good fit for morality. And though Socrates might perplex his interlocutors about common terms, as

Joyce reminds us, this alone does not show that any impossible content is involved—even if their concessions lead them to contradictions, these concessions do not seem to belong to the content of the disputed terms.203

Thus Joyce’s fictionalism does not seem capable of providing an account of the

descriptive content allegedly essential to moral judgments. If we follow his suggestions

about the possible ontology referred to by this content, reasons for action, we end up

faced with same metaethical problems all over again. Nor could an appeal to the ‘story of

morality’ serve to pin down any particular ontology as a referent for moral terms and

concepts. Alternatively, if we follow his suggestions about impossible content, we end up

with the sort of hazy ‘content’ that lends itself more to expressivism than its descriptivist

alternatives.

6.2 Kalderon: an overview

Kalderon defends a hermeneutic fictionalism. That is, he argues that a fictionalist account

of moral judgment—where there is fictional content about moral properties that we

accept with something more like make-believe than belief—is a plausible interpretation

of actual practice. Unlike Joyce, who recommends a fictionalist practice as an attractive

203 Joyce (2001: 198). 163

alternative to the error-ridden practice of moral evaluation, Kalderon does not think our moral practice rests on erroneous presuppositions. For even though fictional content about moral properties has an important role to play in moral judgment, we do not make moral judgments by believing in this content. So, as we will now see, while Joyce touts fictionalism as a worthy companion to error-theory, Kalderon touts it as a promising improvement on traditional forms of noncognitivism.

Why noncognitivism?

Indeed, Kalderon argues for noncognitivism—roughly, the view that the key attitude in morality is not belief—before even introducing any distinctively fictionalist elements into his account. He first provides a merely negative argument against cognitivism: a truly cognitivist domain always imposes at least a lax rational obligation not to be intransigent when inquiring in a public way on behalf of others, and yet inquiry in the moral domain is always public and on behalf of others without imposing any rational obligation not to be intransigent. ‘Intransigence’ here is the state of being unmotivated to reconsider one’s own views even when faced with an ideal opponent (a reasonable, informed, and interested person who disagrees with you about the issue itself or even about which considerations count as reasons). Thus Kalderon’s argument is that morality is so tolerant of intransigence in certain cases of disagreement that it cannot plausibly be construed as a cognitivist domain.204

204 Kalderon (2005: 8–42). 164

Next, after telling us what the important attitude in morality isn’t, he provides a positive argument intended to shed some light on what it is. Focusing on the phenomenology of an undecided individual considering both sides of a moral issue,

Kalderon notes that different features of the circumstance in question tend to become salient and present a certain normative appearance depending on what normative perspective is being adopted at the moment: this feature pops up as an apparent reason now, but later that feature will pop up as an apparent reason. But these phenomenological tendencies suggest that we are dealing with something affective: for “a normative perspective structures a person’s moral consciousness in just the way a certain kind of affect, [Scanlon’s so-called] desire in the directed attention sense, structures a person’s consciousness.” And if, as Kalderon proposes, we resist the urge to posit some independently-specified attitude responsible for these tendencies (giving a “substantive account”), and instead simply identify the affect with this particular syndrome of phenomenological tendencies (embracing “minimalism”), then we can conclude that “the adoption of a normative perspective is just a matter of appropriately configuring one’s affective sensibility.” And this all but entails noncognitivism about the key moral attitude: i.e., about the attitude involved in accepting a moral sentence and in having a moral commitment.205

205 Kalderon (2005: 43–50). 165

Why not expressivism?

But Kalderon resists the move from noncognitivism to the standard expressivist form of noncognitivism. For, while noncognitivism is only a thesis about the moral attitude,

standard expressivist noncognitivism has additional commitments regarding moral

content: namely, that moral content is not representational content describing “the

existence and distribution of moral properties” (‘nonfactualism’) but instead nothing

more than the attitude itself (‘expressivism’). Kalderon wants to develop a

noncognitivism which departs from this standard version: a factualist noncognitivism which makes a point of incorporating realist-style representational moral content. So he develops a critique of standard noncognitivism, which comes in two quasi-historical stages: primitive expressivism and sophisticated expressivism.

Primitive expressivism calls for “an atomistic reduction of moral content to the expression of noncognitive attitudes”. And according to Kalderon, the view faces an

“unanswerable dilemma”: the Frege-Geach problem. Moral sentences embedded in unasserted contexts do not express the attitudes invoked by expressivist semantics. If expressivists confine their account to freestanding occurrences, it is incomplete. And if they add a supplementary account for embedded occurrences, the two accounts must somehow assign the same content, lest the combined account fall apart when it comes to argumentation, belief-ascriptions, answering questions, etc. And here the dilemma is that assigning the same content seems incompatible with primitive expressivist semantics, and yet assigning different content ruins everything. Kalderon adds that the problem stems from the reductive identification of content with attitude-expression, and that mere 166

supervenience of content on attitude-expression (of the sort compatible with moral realism) could avoid the problem entirely, with denoted properties serving as “the common element of meaning between freestanding and embedded occurrences of moral predicates”.206

Kalderon then rebuts an objection that points the way towards diagnosing the underlying problem. The objection is that everyone else has the same problem: it’s just that instead of identifying content with expressed noncognitive attitudes, they identify content with expressed beliefs. But this is actually untrue. It confuses two senses of

‘belief’: the propositional attitude and the propositional content believed. It also confuses two senses of ‘express’: the semantic relation between a sentence and a proposition, and the pragmatic relation between an utterance and an attitude—Kalderon offers ‘convey’ as an alternative verb for this relation. So everyone else does not have the same problem, for everyone else does not identify content with the pragmatically conveyed propositional attitude in the way primitive expressivists identify content with the conveyed noncognitive attitude.207

The underlying problem, Kalderon proposes, is the ‘pragmatic fallacy’: “mistak[ing] the contents of moral sentences with what their utterances normally convey”. Just because utterances of moral sentences have practical effects, that doesn’t mean we can identify the content of these uttered sentences with those effects.208

206 Kalderon (2005: 53–61). 207 Kalderon (2005: 61–64). 208 Kalderon (2005: 64–66). 167

What about sophisticated expressivism? Kalderon sees this as a move from semantic behaviorism to functional role semantics: a moral sentence’s content is identified, not so much with the linguistic act of expressing a noncognitive attitude, but with the functional role played by the acceptance of that sentence. But Kalderon keeps raising two main lines of objection to the development of functional role accounts by sophisticated expressivists.

First, if they are taken as accounts of semantics, they turn out to be perfectly compatible with representational semantics, and so provide no route to a nonfactualist semantics of nonrepresentational content. Second, the accounts can be taken equally well as accounts of pragmatics instead of semantics, which again leaves nonfactualism unsupported, barring the pragmatic fallacy.

He focuses on Gibbard, who gives two successive accounts. The preliminary account is similar to primitive expressivism: it identifies normative content with the expression of the acceptance of a system of norms, and is thus vulnerable to the Frege-Geach problem.

Gibbard’s subsequent account has three components. First, a formal account modeled on possible worlds semantics and put in terms of world-norm pairs. This is not yet nonfactualism, for it is clearly compatible with representational semantics. Second, a mapping of the first component onto the functional role of the acceptance of a normative sentence: Gibbard appeals to a Davidsonian radical interpreter and the practical element of functional role captured by the notion of ‘normative governance’. This still isn’t nonfactualism: it “uniformly represents the contents of normative sentences in freestanding and embedded occurrences”, which is nice, but the solution to the Frege-

Geach problem will not be complete until the account becomes nonfactualist. The third 168

component picks up the slack, giving a dispensability argument for nonfactualism: since

“the functional role of normative predicates can be explained without supposing that they

denote any distinctively normative properties”, we should treat these predicates as

nonrepresentational.209

Kalderon applies his two main objections to Gibbard’s dispensability argument. First,

the argument would prove too much: functional role semantics needn’t always lead to

nonrepresentational semantics, despite the fact that it always specifies meaning

independently of denoted properties, for functional role might place sufficient constraints

on content to determine a denotation. Indeed, Gibbard’s own account determines a

(perhaps massively complex) associated property.210 Gibbard would cite the open

question argument in resisting any move from an associated property to a denotation, but

such an argument falls to familiar water-H2O considerations. Second, dispensability

considerations fit perfectly well with a noncognitivist pragmatics instead of a nonrepresentational semantics. Just as a van Fraassen–style constructive empiricist can

embrace dispensability considerations when developing an account of scientific theo-

rizing while insisting that scientific theories do purport to represent unobservable

structures, likewise a normative fictionalist could do the same with an account of

normative evaluation and the representation of putative normative facts.211

Kalderon then applies his objections to sophisticated expressivism in general,

considering three possible arguments that could take us from a functional role account to

209 Kalderon (2005: 66–77). 210 Gibbard (2003) makes use of just this idea. 211 Kalderon (2005: 77–82). 169

nonfactualism. The first argument reasons that since functional role determines meaning, then there must be nothing more to meaning beyond functional role. But this conflates supervenience with identity.212 The second is the now-familiar ‘argument from semantic

dispensability’. Kalderon repeats his ‘proves too much’ objection, with the

counterexample of logic: even if abstract logical objects are explanatorily unnecessary

when it comes to giving a functional role semantics for logical vocabulary, it might still

denote logical objects.213 The third is an argument from practical function: “[a]ction is

implicated in the functional role of moral predicates” in a certain way which “show[s]

that moral discourse is nonrepresentational”. Now, it would beg the question against

moral factualism to assume that no truly representational predicate can have “relations to

action” in its “content-determining functional role”. And it would also be false: belief

states are individuated in part by their relation to action. At most, then, this argument

shows something about attitude, not about content: it confuses noncognitivism with

nonfactualism, pragmatics with semantics, ignoring noncognitive uses of representations,

thereby committing the pragmatic fallacy.214

Now, all that Kalderon has claimed to show is that nonfactualism is unsupported: the

sophisticated expressivist functional role account of moral semantics fails to secure the

nonfactualist thesis that moral content is nonrepresentational. But he also seems

convinced that nonfactualism is implausible and thus to be avoided if possible: “We

appear to be in the uncomfortable position of choosing between a plausible semantics

212 Kalderon (2005: 83–84). 213 Kalderon (2005: 84–86). 214 Kalderon (2005: 86–93). 170

wedded to an implausible cognitivism and an implausible semantics wedded to a

plausible noncognitivism”.215 And it is at this point of crisis that fictionalism is supposed to save the day. By wedding noncognitivism about moral attitudes to factualism about moral content, we can have our cake and eat it too.

Developing hermeneutic moral fictionalism

Kalderon provides further characterization of moral fictionalism both by canvassing the options available for understanding fictionalist ‘acceptance’ and quasi-assertion and also

by offering an explanation of how and how not to understand the noncognitive attitude

his account incorporates. First, though fictionalists may wish to eschew ordinary belief

and assertion when it comes to the surface content they deem fictional, they are happy to

recognize something close: acceptance and quasi-assertion. I might not believe that

Achilles was touchy, but I in some sense accept it.216 One approach to understanding

acceptance and quasi-assertion is ‘metalinguistic’: though we do not believe in or assert

the fictional content itself, we really do believe in or assert something about the fictional

content. Thus in accepting or quasi-asserting that Achilles was touchy, I really do believe

or assert that “Achilles was touchy” is true according to the Iliad. Or in accepting or

quasi-asserting that Santa is jolly, I might really be believing or asserting that “Santa is

jolly” is something fun and socially valuable to pretend.217 Another approach is

‘objectual’: in accepting or quasi-asserting the fictional content, we believe in or assert

215 Kalderon (2005: 93–94). 216 Kalderon (2005: 110–12, 119–21). 217 Kalderon (2005: 121–22). 171

something about real-world conditions that are connected with the fictional content. In accepting or quasi-asserting that the browser dislikes special characters, I am really believing or asserting that the browser cannot correctly render special characters unless they are indirectly represented with HTML entities, a real-world condition that makes the fictional content convenient. Certain proverbs about the physical world arguably embody similar fictions: “nature abhors a vacuum”, “nature does nothing in vain”, “water seeks its own level”.218 Thus both the metalinguistic and objectual approaches to fictionalism

understand acceptance and quasi-assertion of fictional content in terms of belief and assertion of some distinct “real content”. Ultimately, then, these accounts are cognitivist and assertoric.

But the final option eschews real content and understands acceptance and quasi- assertion of fictional content in fully noncognitive and nonassertoric terms. We might use metaphor not to express any beliefs or make any assertions but merely to draw attention to certain possible comparisons. We might use mathematical sentences as a mere device for computation and proof. And indeed we might use representational content purporting to denote moral properties as a way of expressing noncognitive attitudes.219

Kalderon favors a mixed approach, which understands acceptance and quasi-assertion as a combination of cognitive and noncognitive, assertoric and nonassertoric elements.

For example, in van Fraassen’s ‘constructive empiricist’ version of fictionalism,

“scientific acceptance [of a scientific theory] is an amalgam of belief and intention”: we both believe (or assert) real content, claiming of the scientific theory that it is empirically

218 Kalderon (2005: 123–26). 219 Kalderon (2005: 126–28). 172

adequate, and also have a noncognitive intention to “deploy that theory in the conduct of

science”. Likewise, Kalderon wants to characterize moral acceptance and quasi-assertion as “an amalgam of cognitive and noncognitive attitudes”. On the cognitive side, there are thoughts about what is being evaluated: i.e., real content “plausibly limited to

representing morally salient features of the circumstance”, and the “thoughts or

perceptions” that feature this real content. On the noncognitive side, there is something

affective: i.e., “an appropriate affective response” wherein the morally salient features

“present a certain complex normative appearance” so that we have a “phenomeno-

logically vivid sense of the moral reasons apparently available”. And since Kalderon

favors minimalism about this key moral attitude, he is willing to combine the cognitive

and noncognitive into a single kind of state: instead of seeing the cognitive and

noncognitive attitudes as “distinct”, he characterizes moral acceptance in terms of the

entire cluster of psychological tendencies.220

Now, one peculiarity of this account is that there is no apparent role to be played by

the fictional content purporting to denote moral properties. To be sure, the real content

plays a role: we have thoughts or perceptions about the features of the situation we find

morally salient. And there is also work to be done by our affective sensibility: its

configuration helps determine the normative appearance these features present. But

where is the fictional content? Kalderon’s final remarks on the key moral attitude provide

something like an answer.

220 Kalderon (2005: 129). 173

Kalderon is officially pessimistic about a providing “an accurate noncircular

specification of the attitude centrally involved in moral acceptance”, but he softens our

disappointment by offering explanations of why circularity cannot be avoided. The first,

and seemingly least important in Kalderon’s estimation, is a simple matter of the

“contingent expressive limitations” of human language. The second explanation

highlights the possibility of moral disagreement stemming from different “moral

sensibilities”, and suggests that perhaps different moral communities might work with

different moral attitudes, so that “no general, substantive account of moral acceptance”

will be available. It is the third explanation that finds a place for the moral properties de-

noted by the fictional content featured in Kalderon’s account. To shed light on the

“normative appearance” involved in the relevant attitude, perhaps there is no better way

than invoking what it is an appearance of: viz., “the normative properties apparently

instantiated by the object of the affect”. Indeed, Kalderon is willing to suggest that there

may be no way for noncognitivists to account for noncognitive attitudes without

embracing “representational semantics”: viz., fictional content purporting to denote moral properties.221

We are left with something like the following account of moral evaluation. To think

that torture is wrong is to have a moral commitment understood in terms of accepting the

sentence “torture is wrong”. And this state of acceptance is specified in terms of a cluster

of psychological tendencies, naturally divided into two elements: (i) thoughts or

perceptions about morally salient features of the situation (these being represented by the

221 Kalderon (2005: 130–35). 174

“real content” of one’s acceptance), and (ii) the presentation of a normative appearance by those features (these being understood in terms of one’s affective sensibility, and in terms of fictional content about moral properties). Thus I think (e.g.) of the agony of the victim and the ruthlessness of the torturer, and these features appear to me as wrong- making features or reasons that morally count against torture, this because of my affective sensibility and because of fictional content representing torture as instantiating the property of moral wrongness due to these features.

6.3 Kalderon: a critique

My critique of Kalderon’s version of fictionalism will be in line with my overall critique of descriptivist views: he will have great difficulty developing a semantic account capable of explaining why the descriptive content featuring in moral judgments refers to the ontology it does. But in pressing this critique, which I hope has significant force of its own, I will also appeal to dialectical considerations relating to the particular commitments of Kalderon and his efforts to gain a leg up on expressivism. This seems especially appropriate, given that both of our criticisms center on semantic issues.

As before noted, Kalderon’s case against expressivism is in reality more of an undermining of a case for expressivism. His main point is that the functional-role semantics employed by sophisticated expressivists like Gibbard to overcome the Frege-

Geach problem does not lead us to the conclusion that moral content is nonrepresentational, and is thus perfectly compatible with the sort of ‘factualist’ interpretation of moral content Kalderon espouses. And by itself, this point is neutral 175

between expressivism and fictionalism, favoring neither view over the other. If I am right

that introducing descriptive content only burdens a metaethical theory with the problem

of explaining reference, then this would break the tie and sink fictionalism. Perhaps

Gibbard could not argue directly from his functional-role semantics to expressivism, but he could argue against the introduction of descriptive content on independent grounds.

Moreover, the problem of explaining reference will still cause trouble even if we

grant Kalderon’s repeated suggestion that there are impressive “difficulties facing an

expressivist nonfactualism”. For even if expressivism involves “an implausible semantics

wedded to a plausible noncognitivism”,222 expressivists can always retort that fictional-

ism involves impressive semantic implausibilities of its own. And then expressivists can

use Occam’s razor to shave off the additional semantic commitments of fictionalism,

leaving expressivism the dominant view within the realm of noncognitivism.

But of course this all assumes that the problem of explaining reference does indeed

arise for fictionalists. We have seen several ways in which the problem might arise. For

realists, naturalists and nonnaturalists alike, there is a problem of explaining the

referential relation between moral concepts and actually instantiated moral ontology. For

error-theorists, there is a problem of explaining how moral concepts referentially relate to

possibly instantiated moral ontology. Or, for error-theorists who take the radical step of

denying the very possibility of moral ontology, there is the problem of explaining how

moral concepts could count as descriptive concepts that purport to refer to moral

ontology if in fact their purported referent is impossible. Depending on what fictionalists

222 Kalderon (2005: 146). 176

think about the actuality or possibility of moral ontology, any of these problems could

arise.

Now, Kalderon himself would probably allow for at least the possibility of moral

ontology. He is clear that his view is intended to be compatible with actually instantiated

moral ontology. It is just that, if moral properties do in fact exist, our moral judgments do

not consist in beliefs about those properties. Indeed, even if my moral judgment quite

accurately represents “the existence and distribution of moral properties”,223 the

representational content is not believed by me. And this alone raises serious questions: if

I were to believe that content, would that belief constitute a moral judgment, and if so

doesn’t Kalderon’s view inherit all the semantic worries of a realist view? But we can set that point aside, for the problem about descriptive content making reference to moral properties should not turn on precisely what attitude we take towards that content.

But we can find in Kalderon’s critique of expressivism a suggestion as how to account for reference: perhaps the functional-role semantics of Gibbard could itself suffice to provide moral concepts with referents. In accepting a complete system of norms, I associate either a single property (e.g., utility-maximization) or a heterogeneous disjunction of different properties with an action being permissible. But then this (single or disjunctive) property will be a plausible candidate for what the relevant normative predicate denotes. Gibbard offers open question considerations to defeat this suggestion, but Kalderon points out that these considerations succumb to the familiar case of water and H2O. Summing up his case, he writes:

223 Kalderon (2005: 151). 177

Suppose that the functional role of normative predicates does indeed deter mine their content. The assigned functional role might place sufficient constraints on a normative predicate’s content to determine a denotation. Indeed, Gibbard’s account, suitably developed, is a representational moral functional role semantics.224

But this suggestion is not terribly promising, for it has no obvious way of properly accommodating the phenomenon of moral disagreement. To be sure, my moral standards may pick out some single or disjunctive property as what makes actions right. But your moral standards may pick out a different property, and an account of moral semantics which interpreted us as simply referring to different properties would end up precluding genuine moral disagreement. What I say is true if it corresponds to my properties, and what you say is true if it corresponds to your properties, so that we are not even discussing the same topic. Of course, the threat of this unsavory relativism would dissipate if the functional-role semantics could determine the same referent for every- body, regardless of their own moral standards. But this would go far beyond anything that could be expected from a Gibbard-style functional role, for it only determines properties by drawing on an individual’s complete system of norms.

Kalderon could perhaps defend the suggestion by providing a nonrepresentational account of moral disagreement. Much as Stevenson construed moral disagreement as less

“disagreement in belief” and more “disagreement in interest”,225 Kalderon could draw on

the affective elements of his theory to account for disagreement. But this would run into a

couple of serious problems: first, Kalderon proposes to understand these affective

elements in terms of the normative properties represented, which means a move to the

224 Kalderon (2005: 81). 225 Stevenson (1937). 178

affective side of things does not leave behind the normative properties that caused the problem in the first place. But, second, even if he is willing to understand the affective

elements independently of the normative properties represented, this could only go

partway towards accommodating moral disagreement. For, no matter how opposed in affective response you and I may be, if we are referring to completely different normative properties, then we do not disagree over whether the object of evaluation instantiates

those properties. And of course any conception of normative properties in which

normative disagreement has nothing to do with those properties is hard to fathom.

And if Kalderon’s fictional content will not make sense of moral disagreement, his

“real content” will fare no better. As he notes, the real content of a moral judgment is

“plausibly limited to representing morally salient features of the circumstance”.226 But consider a simple situation where you and I agree on the morally salient features of the circumstance, but disagree on whether these features are right-making or wrong-making.

Presumably our resulting moral judgments will share the same real content (about the features), and differ in their fictional content (about the moral properties). But then the real content cannot serve to explain the way in which we disagree.

Kalderon’s appeal to functional-role semantics, then, does not provide fictionalism with a plausible approach to explaining reference. The theory is thus left with the same problem as other descriptivist theories. Leaving aside the problems of any appeal to functional-role semantics in particular, I have three general objections to Kalderon’s theory. Each objection directly draws on the semantic problem.

226 Kalderon (2005: 129). 179

First, without a semantic account that can determine the very same moral properties for very different moral evaluators with very different moral standards, Kalderon is left with the same problem we saw earlier threatening David Copp: his view will collapse into expressivism with relativist trimmings. After all, if the different evaluators have fictional content referring to different properties, then what they have in common is found on the noncognitive and affective side of things. This would not handle evaluators lacking in affect, and they would be expelled from under the moral umbrella. And that means that it is the noncognitive and affective side of moral evaluation, as opposed to the fictional content purporting to denote moral properties, that makes moral terms and concepts count as moral. The fictional content will do nothing more than provide relativist truth-conditions.

I suspect that Kalderon would respond to this objection by reminding us of his commitment to minimalism: we cannot separate the noncognitive-cum-affective from the cognitive, for both are nothing more than a single cluster of psychological tendencies.

Alternatively, he might embrace his suggestion that moral affect cannot be understood except in terms of apparent normative properties. If he could make the case for holding fictional content and affective response to be inseparably bound at the hip, then the possibility of evaluators lacking in affect could be ignored.

But as before, this is a tall order, and neither of the two points just mentioned seems capable of filling it. Simply because a mental state is understood as nothing more than a cluster of psychological tendencies, it by no means follow that the noncognitive elements cannot separated from the cognitive. For that, we would need some independent reason to 180

think that the content in question is incapable of being represented without a certain

affect. And the suggestion of understanding moral affect in terms of apparent normative

properties only piles mystery upon mystery: what could these mysterious properties be,

whose very representation somehow constitutes affect—i.e., a configuration of one’s

affective sensibility? In either case, the possibility of different properties for different

evaluators makes things worse, for now there are lots of possible properties with this

mysterious connection to affect.

Second, Kalderon is left with a conception of normative properties that violates some

highly plausible constraints on what a normative property would have to be. With

different normative properties for different evaluators, as mentioned before, normative

disagreement turns out to have nothing to do with normative properties. Our ‘wrong’

enters into fictional content purporting to denote this property, their ‘wrong’ enters into fictional content purporting to denote that property, and so a disagreement over what is wrong must be understood in terms of something like practical opposition. But if a property is irrelevant to normative disagreement, it is hard to see how it could deserve the name ‘normative property’. For it seems to be a constraint on normative properties that they have something important to do with normative disagreement.

Or consider the constraint that normative properties are relevant to deciding what normative judgments to adopt and what kind of life to live. If our ‘wrong’ goes with this property, and their ‘wrong’ goes with that property, then knowing this provides us with no normative guidance. We might wonder whether to adopt the fictional content of the other community, or stay with our own. We might wonder whether to live our lives in 181

accordance with their fictional content, or our own. But these normative properties would

say nothing on the matter, and would thus hardly deserve the name ‘normative property’.

Lastly, Kalderon would need an expressivist-style account of cross-community

agreement on embedded normative judgments, thereby losing any Frege-Geach advantage he hoped to secure over expressivism. Recall that the main advantage of introducing descriptive fictional content was to provide a “common element of meaning between freestanding and embedded occurrences of moral predicates”: namely, the denoted properties. But what if different communities denote different properties? The commonality is lost, and this serves to jeopardize the advantage.

To see this point, consider two disagreeing communities who exemplify the old adage

“one man’s modus ponens is another man’s modus tollens”. They agree on the conditional ‘If lying is wrong, then getting your little brother to lie is wrong’, but disagree about whether both acts are wrong or neither. Now, if one community’s ‘wrong’ refers to F, and the other’s ‘wrong’ refers to G, then we can understand how each side manages to reason through their respective modus, with F as the common element for the first community and G for the second. But how can we understand the fact that both communities agree on the conditional? We cannot make use of the denoted properties, for ex hypothesi they are different for each community. So we will have to understand their agreement in terms of the attitudes they adopt to lying and getting little brothers to lie.

But if this is successful, then there is a way of successfully understanding normative judgments in embedded contexts in terms of attitudes alone, without bringing in 182

descriptive content purporting to denote properties. And that means expressivists will

have a solution to the Frege-Geach problem.

In the end, Joyce’s fictionalism does not manage to provide an satisfactory semantic account for the descriptive fictional content that distinguishes his view from expressivism. Even if he is right that expressivism is burdened with semantic implausibilities (thanks to the Frege-Geach problem), his own view faces semantic burdens enough of its own—and indeed, there seems to be no way to discharge these burdens without saving expressivism from its burdens. Thus hermeneutic fictionalism is not only bereft of any semantic advantage over expressivism, but in a worse semantic position overall.

183

CHAPTER 7: BLACKBURN’S SUPERVENIENCE ARGUMENT

Simon Blackburn’s supervenience argument—focusing on the mysterious “ban on mixed

worlds”—attracted a great deal of metaethical discussion in the 1980s and 1990s.227

Moral realists have recognized the argument as an important challenge to their view, and have proposed a wide variety of possible responses.228 And yet I think it is fair to say that there is no commonly recognized account of how exactly the argument works and where exactly its weaknesses might lie.229 Different discussions interpret the argument in significantly different ways,230 and there is no standard set of objections used to frame critical examination of the argument. As a result, it is difficult for an outsider—or even an insider—to find their bearings when it comes to Blackburn’s argument.

In this paper, I hope to provide a defense of the argument that clarifies both the argument and the objections it must overcome. Many of the extant objections, I will argue, fail to engage the argument in its true form. And to counter the more compelling objections, it will be necessary to bring in additional argumentation that Blackburn

227 The argument was first presented in Blackburn (1971/1993). Important discussions in the 1980s and 1990s include Klagge (1984); Blackburn (1984); Blackburn (1985/1993); Elliot (1987); Dreier (1993); Shafer-Landau (1994); and Zangwill (1995). 228 Responses to the argument can be found in Klagge (1984); McFetridge (1985); Elliott (1987); Dreier (1993); Shafer-Landau (1994); Zangwill (1995); Bloomfield (2001); Shafer-Landau (2003); Majors (2009). 229 Dreier notes that he is “not entirely confident that [he has] correctly represented Blackburn’s argument as [Blackburn] intended it” (1993: 16). Part of the problem surely lies in the fact that the notion of supervenience was still unsettled when Blackburn was developing his argument, and in his 1971 use of the term “logical necessity”, which is ambiguous between and conceptual necessity. 230 For example, Klagge (1984) and McFetridge (1985) focus on accepting weak supervenience but denying strong supervenience, posing a dilemma between readings employing metaphysical necessity and readings employing logical necessity. But Dreier’s “Supervenience Argument” focuses on accepting strong metaphysical supervenience and denying strong analytic supervenience. Similarly, Shafer-Landau often returns to the question of whether all metaphysical necessities or impossibilities must also be conceptual necessities or impossibilities: see his (2003: 86) and his (2005: esp. 327). In contrast, Zangwill’s reading, much like my own, focuses on accepting a supervenience as a conceptual constraint while denying that any particular natural-to-moral necessities count as conceptual constraints. 184

himself does not clearly provide. One important upshot of the discussion is that moral

realism is not the only target of the argument: it raises trouble for a variety of metaethical

views—from error-theory to constructivism—due to their common commitment to

realist-friendly construals of moral evaluation. Rightly understood, Blackburn’s argument

not only survives the prominent responses, but turns out to be more powerful than his

critics and Blackburn himself have recognized.

7.1 Preliminary points

I shall begin with a brief outline of the common understanding of Blackburn’s argument:

for despite the disorientation over details, considerable consensus emerges at a

sufficiently high level of abstraction. All discussions of Blackburn’s argument agree that

it is an argument against moral realism. Morality is committed to two theses—lack of en- tailment and supervenience—which generate a mystery when combined, the so-called

“ban on mixed worlds”.231 And realism about morality, on this construal of the argument,

has no explanation capable of dissolving this mystery. Realists, then, can respond that

morality is not committed to both theses (once the theses are rightly understood), or that

their combination generates no mystery (as shown by similar cases from elsewhere in

philosophy), or that realism does in fact have a satisfactory explanation for the mystery.

And indeed, most realist responses fall somewhere in this three-part classification:

disavowing the theses, denying the mystery, or offering an explanation.

231 This terminology is introduced in Blackburn (1984). 185

In my view, this common understanding of the argument does not provide an entirely

accurate picture of the argument’s ambitions. First, Blackburn’s argument is fairly mod-

est in that it is a comparative argument. That is, it does not aspire to conclusively dis-

prove realism by showing that realists have no possible explanation of the mysterious ban

on mixed worlds. Rather, the argument presents us with an anti-realist explanation and

deems this explanation satisfying and illuminating in a way that realists have yet to match. This would merely leave realists at an explanatory disadvantage when it comes to the ban on mixed worlds.232 Second, as I will argue in section 5, the argument is fairly ambitious in that it is not only anti-realist but generally anti-descriptivist. To be sure,

Blackburn has focused on challenging realists with the argument. But the anti-realist explanation he touts is tied to his expressivism, which gives the argument itself a wider scope: it can readily be carried over to challenge other descriptivist views. To take one example, error-theorists agree with realists about the descriptivist nature of moral evaluation and the ontological commitments that come with it—they only deny the ontol- ogy itself. But this means, as we will see, that the argument challenges error-theory and realism alike, promising Blackburn’s expressivism an advantage over both views.

But if the argument is about the metaphysical mystery of a ban on mixed worlds, then how could it pose a challenge to views locked in direct metaphysical opposition: i.e., how could it jeopardize both error-theory and realism? The all-important answer is that the

232 This point can counter Zangwill’s “partners in crime” response in his (1995: 252–56). Blackburn’s argument need not be not guilty of overreacting to the presence of synthetic a priori principles, recklessly rushing to an anti-realist conclusion. Instead, anti-realism gains plausibility only because (and if) it has a satisfactory explanation and realism does not. If e.g. color anti-realism had a similar advantage over color realism when it came to explaining the synthetic a priori principles of that domain, then it would likewise be reasonable to shift a measure of loyalty from realism to anti-realism about color. 186

argument is not about a metaphysical mystery, but a conceptual mystery. Thus the two theses that generate the mystery are to be understood in terms of conceptual competence233 rather than metaphysical necessity and possibility. Blackburn himself has

insisted on this point,234 but I will argue that careful attention to it does away with a great many objections in the literature. And indeed, once we appreciate how the two theses are

not merely tied to conceptual competence, but deeply embedded and indispensable in the

very practice of moral evaluation, other objections end up losing their force as well (as I

will argue in section 4). The success of the argument thus turns on a proper understanding

of the modal terminology occurring in the two theses.

Thus I am insisting on four points for understanding the ambitions and the success of

the argument: (1) The argument is comparative. (2) The argument is generally anti-

descriptivist. (3) The argument’s modal terminology is used to talk about matters of

conceptual competence. (4) This conceptual competence is supposed to be exceptionally

deeply embedded in moral evaluation.

7.2 How the argument works

I will now present a version of the argument. First, the thesis assigned the label ‘lack of

entailment’. It is highly plausible that there are cases where the natural properties of a

thing do not conceptually entail its moral properties, as witnessed by the enduring

233 The necessities of conceptual competence are supposed to lie somewhere between strict logical necessities (e.g. “p and q” and “q and p”) and metaphysical necessities (e.g. “this is water” and “this is H2O”). It is not entirely clear whether conceptual necessity is equivalent to analytic necessity (e.g., “I am a bachelor” and “I am an unmarried male”), but this is irrelevant for Blackburn’s argument. 234 Especially in his (1985a/1993). This point is recognized by Zangwill (1995) and Bovens and Drai (1999). Dreier comes close to recognizing the point but then sets it aside when evaluating the argument (1993: 15). 187

popularity of Hume’s is-ought gap and Moore’s open question argument.235 Thus a case

of sadistic pleasure with its natural properties stipulated may be deemed good for its

pleasure by me and bad for its sadism by you, without either of us being guilty of concep- tual incompetence. To speak in terms of conceptual possibility, as Blackburn does, one could say that both of our judgments correspond to conceptual possibilities, since neither is ruled out by conceptual constraints.236

Second, supervenience. It is also highly plausible that only the conceptually incompetent could think that two cases differ morally but not naturally, as witnessed by the widespread popularity of seeing moral-natural supervenience as a conceptual truth.237

Thus two cases of sadistic pleasure stipulated to be naturally indistinguishable may both

be deemed good by me, and both be deemed bad by you, but no one could deem one

good and the other bad without being conceptually incompetent. Again, one could say

that my judgment and your judgment correspond to conceptual possibilities, but that the

mixed judgment comes out conceptually impossible. Hence the phrase ‘ban on mixed

worlds’ might well be replaced with ‘ban on mixed judgments’.238

235 In addition to Moore (1903/1993), one should consult his (1922/1993). The ‘lack of entailment’ theses directly corresponds to Moore’s thesis that the properties belonging to the category ‘intrinsic value’ are never among a thing’s intrinsic properties. See Blackburn (1985a: 130). 236 Of course, the lack of entailment thesis would be rejected by analytic naturalists, including analytic moral functionalists like Frank Jackson and Philip Pettit. See Jackson and Pettit (1996) and Jackson (1998). 237 Thus Jackson writes that “[t]he most salient and least controversial part of folk morality is that moral properties supervene on descriptive properties, that the ethical way things are supervenes on the descriptive way things are” (1998: 118). Also Smith: “everyone agrees that it is an a priori truth that the moral su- pervenes on the natural” (1994: 22). Nicholas Sturgeon writes that moral supervenience “has until quite recently had the status of a commonplace among academic philosophers” (2009: 53). 238 Thus Zangwill doubts that talk of possible worlds can make sense of Blackburn’s argument (1995: 247– 50). 188

Many commentators on moral supervenience would insist that this ban on mixture— or crusade for consistency—be extended across all metaphysically possible worlds.239

And I am inclined to agree: whether the two cases of sadistic pleasure are in the same metaphysically possible world or in different metaphysically possible worlds, only the conceptually incompetent could deem one good and one bad. This insistence on extending the supervenience pattern across all metaphysically possible worlds, so that no two individuals from any possible worlds can differ morally but not naturally, goes by the name ‘strong metaphysical supervenience’. And thus we should not just say that supervenience is a conceptual truth and leave it at that; we can embrace the hybrid con- ceptual-metaphysical thesis that strong metaphysical supervenience is itself a conceptual truth.240 Thus I may judge that all cases of sadistic pleasure in all metaphysically possible worlds are good, and you that they are bad, but only the conceptually incompetent could hold that some are good in some worlds and some are bad in other worlds.241

Given these two theses, then, what is the mystery? It lies in the fact that scenarios which are apparently possible—or at least easily describable—are ruled out as

239 Klagge (1984: 376); Elliott (1987: 135–36); McFetridge (1985: 248–51); Dreier (1993: 15); Zangwill (1995: 248). 240 Zangwill makes the same suggestion in his (1995: 257–58). 241 One might have special concerns about the fact that the supervenience thesis is being forwarded not as a mere metaphysical thesis, but as an inviolable conceptual constraint. In doing so, the argument is sticking its neck out and making itself more vulnerable. First, one might ask whether someone conceptually competent might not accept (the metaphysical possibility of) a version of divine command theory on which God might assign different moral statuses to otherwise identical cases: cf. Klagge (1984: 374–75). Second, one might ask whether someone conceptually competent might not accept (the metaphysical possibility of) egoism or a ‘might makes right’ principle, on which two otherwise identical cases (of e.g. murder or theft) are to be evaluated differently simply because of how it impacts the evaluator or the interest of the stronger: see Bloomfield (2001: 50–52). I do not think these objections are difficult to get around: in both cases moral differences still require differences in supernatural facts or in (presumably natural) facts about the identity of individuals. It is just that doing so ends up introducing complications into the standard (and highly convenient) talk of the moral supervening on the natural. 189

conceptually impossible. Given two naturally indistinguishable cases, then supposing that there is nothing conceptually impossible about them both being good, and likewise noth- ing conceptually impossible about them both being bad, then why would it be conceptually impossible for one to be good and the other bad? Why would such a judg- ment be a sign of conceptual incompetence? Whence the conceptual constraint that

strikes it down?

Here Blackburn offers an anti-realist explanation: moral evaluation is a matter of

social coordination and decision-making on the basis of the features of our environment.

It would hardly be practical to insist that everyone share the same moral standards, since

we will inevitably differ and we need a ecumenical practice capable of accommodating

these differences as a form of in-house disagreement. But it is practically necessary that everyone conform their evaluations to a supervenience constraint: without it, we lapse into practical and discursive chaos, where no one has any evaluative standards to speak of. So, simply due to human needs, we end up with a practice incorporating both the lack of entailment thesis and the supervenience thesis.242

But for the realist, this practice is primarily a descriptive one, aimed at faithfully

representing the presence or absence of certain normatively significant facts in the world.

Now, it is understandable enough that such a practice would incorporate a lack of

entailment thesis: we could hardly be so confident in any particular theory about these

facts as to build it into the very concepts of the practice. But then why would the practice

242 Blackburn (1971/1993: 122); Blackburn (1984: 185–86); Blackburn (1985a/1993: 143–44). See also Mabrito (2005: esp. 302–04). In this paper I am making the assumption that the expressivist explanation is a good one. 190

incorporate a supervenience thesis? After all, there is no obvious antecedent reason to expect that these facts will conform to any pattern whatsoever, much less to single out the supervenience pattern and enshrine it as a conceptual constraint from the outset of the practice. Hence it is fairly puzzling that morality should be restricted by such a constraint if it is, as realists say, a descriptive practice.

This presentation of the argument can serve as an illustration of the four points of the last section. The argument is comparative, turning as it does on the superiority of the anti- realist explanation on offer. It is generally anti-descriptivist, for anyone who agrees with realists about the nature of moral evaluation will face the same challenge. This challenge is one of accounting for the conceptual constraints of a practice, rather than anything particularly metaphysical. And, though this last point will gain clarity in section 4, the conceptual constraints in question are part and parcel of moral evaluation: deeply embedded in the practice rather than picked up along the way.

7.3 Countering objections with conceptual competence

I claimed earlier that many objections come to grief through insufficient attention to the central role of conceptual competence in Blackburn’s argument. In this section, I attempt to demonstrate this claim, considering three objections. The first two take the disavowing the theses approach, and the last takes the offering an explanation approach. 191

The first objection can be found in many discussions,243 and it turns on the distinction

between weak and strong supervenience. In brief, the objection is that Blackburn cannot

run his argument with mere weak supervenience without inheriting the general instability of such a thesis, but cannot advance to strong supervenience without destroying his own argument.244 On one horn of the dilemma, weak supervenience (on which no individuals

from the same possible world can differ morally but not naturally) is by itself simply too

weak: (i) it understates the dependence we want out of supervenience, by allowing that a

subvening property could possibly (in some other world) be instantiated without the

property which supervenes on it in the world under consideration;245 and (ii) it

mischaracterizes morality in particular, by allowing something so trivial as mere

‘location’, i.e. being in one world rather than another, to make a moral difference.246 And

on the other horn, strong supervenience (on which no individuals from any possible worlds can differ morally but not naturally) would ruin Blackburn’s argument, for strong supervenience and lack of entailment are polar opposites: we cannot accept the one with- out denying the other, since lack of entailment allows for and strong supervenience denies the possibility that two naturally alike individuals could somehow end up differing morally. But this means Blackburn’s argument is in trouble: realists can simply insist on

243 Klagge (1984: 376); Elliott (1987: 135–36); McFetridge (1985: 248–51); Shoemaker (1987); Zangwill (1995: 248); Shafer-Landau (1994: 148–49); Sobel (2001: esp. 371–73); Shafer-Landau (2003: 85–86); Shafer-Landau (2005: 326–27). 244 Blackburn runs his argument with weak supervenience only: thus individuals can differ morally but not naturally, as long as they are in different possible worlds (1985a/1993: 131–35). But, as he explicitly states, his talk of possible worlds is to be understood as expressing matters of conceptual competence (1985a/1993: 137–38). 245 Kim, J. (1984: esp. 159–61). Kim, J. (1990: esp. 12–13). Blackburn himself makes the point in his (1985a). 246 Elliott (1987). 192

strong supervenience and deny the lack of entailment thesis: a form of what I called disavowing the theses.

But in my view this objection works only if we misunderstand the argument, focusing not on conceptual competence, but on metaphysically possible worlds. The objection is correct to point out that an astute realist should accept strong metaphysical supervenience and thereby deny lack of metaphysical entailment. Why, after all, should a realist think that natural duplicates could ever differ morally, even if they do reside in different metaphysically possible worlds? But if we return to conceptual competence, the objection unravels: to accept strong conceptual supervenience and thereby deny lack of conceptual entailment is to embrace something fairly implausible, for there does not seem to be any- thing conceptually incompetent going on when I judge the case of sadistic pleasure good and you judge it bad. And there is no obvious instability in rejecting strong conceptual entailment and running the argument with mere weak conceptual supervenience: (i) there is no demand for conceptual dependence of the sort that rules out conceptually competent moral disagreement; and (ii) it does not mischaracterize morality to allow that mere conceptual ‘location’, i.e. being evaluated under one conceptually competent moral outlook rather than another, could result in different moral statuses being assigned to an individual. So the two theses are still in good shape, as long as we understand them in terms of conceptual competence.

But when Blackburn offered a response similar to the one just presented, emphasizing the distinction between conceptual and metaphysical matters,247 he only invited a follow-

247 Blackburn (1985a/1993: 137–38). 193

up objection from critics.248 There is no way to keep strong supervenience confined to the harmless realm of the metaphysical, the objection goes, for strong supervenience seems

to be a matter of conceptual competence as well: for, if flouting weak supervenience is a

misuse of moral concepts, surely flouting strong supervenience is as well. That is, no

competent evaluator could think that two natural duplicates could differ morally merely

due to being in one world rather than another.249 But if strong supervenience is required

by a constraint of conceptual competence, then it might seem to follow that strong

conceptual supervenience is after all true. Once again, and this time in the realm of the

conceptual, the realist should simply accept strong supervenience and deny lack of en-

tailment.

But this would be a serious mistake. There is an all-important difference between

strong supervenience as a conceptual truth and strong conceptual supervenience. The

latter would enshrine some moral standard in particular as a conceptual constraint, so

that no one could deviate from its verdicts without being branded conceptually incompe-

tent. But the former merely insists that some moral standard or other be applied consis-

tently across all possible worlds, so that conceptual incompetence obtains only when

someone is guilty of inconsistently applying their own moral standard, evaluating natural

duplicates differently. Thus if we accept strong metaphysical supervenience as a con-

ceptual truth, as I proposed in section 2, the two theses are still in good shape and the

248 Elliott (1987). This follow-up objection is discussed in Zangwill (1995). 249 Thus Elliott writes that “Judging that [an actual state of affairs] is good but not that [a naturalistically indiscernible possible state of affairs] is good would just as much breach the conceptual constraint on moralizing as would failure to abide by the ban on mixtures within the actual world. It is ‘constitutive of competence’ with the moral vocabulary that mixtures be eschewed across the structure of possible worlds” (1987: 135). 194

realist has no easy way out. Lack of conceptual entailment and strong metaphysical

supervenience as a conceptual truth are both highly plausible and perfectly compatible

with each other, since there are no particular conceptual necessities being imposed but

only a general and ecumenical consistency requirement on the evaluation of any and all

metaphysically possible worlds. The argument is still intact.

Brad Majors has a second objection drawn from the distinction between global and

local supervenience.250 Blackburn’s argument has always been framed in terms of local

supervenience, in which the moral and natural properties of individuals are compared

with each other. But Majors argues that since “morality is highly context-dependent”,

there is nothing impossible about two individuals differing morally even though they

have exactly the same natural properties—for this moral difference might be owing to the

different contexts the individuals inhabit. Thus the argument cannot fairly saddle realists

with local supervenience, and must instead go global: the two theses must now be global

supervenience, in which the moral and natural properties of entire worlds are compared with each other, and a globalized lack of entailment thesis, in which the naturalistic character of an entire world fails to entail its moral character.251

But if we go global, Majors contends, the lack of entailment thesis loses its

plausibility: “the thesis that fixing the naturalistic character of the entire world fixes also

its moral character… is something no realist should be prepared to deny.” Now, the lack

of entailment thesis is supposed to draw its plausibility from the is-ought gap marked by

“Hume’s Law”, and Majors acknowledges that a localized version of Hume’s Law is

250 Majors, B. (2009: esp. 43–49). 251 Majors (2009: 36–40, 45). 195

plausible enough. But Hume’s Law crumbles when it is globalized: “no one who has

upheld this alleged law has meant to deny that naturalistic propositions describing entire

worlds fail to imply moral propositions; such a denial would be manifestly incompatible with any plausible version of moral supervenience, whether it be local or global.”252 Thus

Blackburn’s argument fails big and fails small: if its theses are localized, then it clashes

with moral context-sensitivity, and if its theses are globalized, then its lack of entailment

thesis is dead in the water. Again, realists can take the disavowing the theses approach.

But I am convinced that globalization is harmless and that a globalized Hume’s Law

is in fact plausible, and once again I think the point turns on the distinction between the

conceptual and the metaphysical. Presented with the thesis that the naturalistic character

of an entire world fails to metaphysically necessitate its moral character, Majors would be

right to deem this thesis both exceptionally implausible and “incompatible with any

plausible version of moral supervenience”. But a parallel thesis working with conceptual

competence—the thesis that you and I might make different moral evaluations about a global naturalistic description without either of us misusing moral concepts—is neither.

Plenty of Hume’s Law fans—from Moorean non-naturalists to expressivists—are virtually guaranteed to accept this lack of entailment thesis. Moreover, there are plausible versions of moral supervenience—e.g., global metaphysical supervenience—that are perfectly compatible with the thesis. And most importantly, those who have found the law plausible about individuals should also find it plausible about entire worlds, for it is difficult to see how the is-ought gap could ever be bridged by mere ‘is’-addition, even if

252 Majors (2009: 45–47). 196

we add together every ‘is’ in the world. Thus Majors’ objection reduces to the earlier

objection that realists should not accept the lack of entailment thesis when put in me-

taphysical terms. And it calls for the familiar reply that Blackburn’s argument is not

about the metaphysical but the conceptual.

But the objection is nevertheless an important one. For in reminding us that morality

is context-sensitive—perhaps so context-sensitive that only global supervenience will

do—it might force us to make a minor revision of Blackburn’s argument. Arguably the

only supervenience thesis that can be treated as a conceptual truth is a global su- pervenience thesis, for those who think morality is maximally context-sensitive (however implausible we may find their view) may not be making any conceptual mistake, and yet those who would evaluate two naturally indistinguishable worlds differently are definitely making a serious conceptual mistake. This would mean that the mystery underlying the argument should be stated as follows: given two naturally indistinguishable worlds, then supposing that there is nothing conceptually incompetent in evaluating them both one way, and likewise nothing conceptually incompetent in evaluating them both another way, then whence the conceptual incompetence in evaluating one world one way and the other world the other way? The ban on mixed worlds is now more of a ban on mixture in one’s evaluations of entire worlds, but the basic argument remains the same.253

The third and final objection can be found in the work of Russ Shafer-Landau. It

features a number of diverse elements, but perhaps the leading idea is that realists can

explain supervenience in terms of the metaphysical relation of constitution: what I called

253 For the sake of convenience alone, I will drop this point for the remainder of the paper, and continue to speak in terms of individuals rather than entire worlds. 197

offering an explanation. For on Shafer-Landau’s view, each instantiation of a moral prop- erty is exhaustively constituted by the instantiation of some concatenation of natural properties.254

Now, to work our way through his response, it is necessary to examine Shafer-

Landau’s formulation of the supervenience thesis:255

(S1): if a concatenation of [natural] properties once underlies a moral one, then it

must (in that world) always do so.

He puts this to use in an ‘innocence by association’ appeal to psychophysical supervenience: if you find no mystery in the claim that the mental supervenes on the physical, you should likewise find no mystery in moral supervenience.256 But this

denying the mystery approach faces an immediate problem: whereas moral supervenience

is plausibly construed as a conceptual truth, there is no conceptual constraint violated by

dualists who deny psychophysical supervenience, and (as Blackburn is fond of pointing

out257) this entirely spoils the parallel. To be sure, there is a way of reading S1 so that

psychophysical supervenience comes out as a conceptual truth: we can just read the

supervenience relation into the ‘underlies’ relation. But then S1-style claims are utterly

trivial conceptual truths, for we can assert them of virtually anything (e.g. flavor

254 Shafer-Landau (2003: 73–78, 86–88). 255 Formulations featuring an ‘underlies’ relation are modeled on Blackburn’s formulation in his (1985a/1993). 256 Blackburn (1985a/1993: 151–52); Shafer-Landau (2003: 87). 257 For example, in his (1985a/1993: 139). Mabrito makes this point in connection with Shafer-Landau in his (2005: 301). 198

supervenes on price, baldness supervenes on height) simply by pointing out that their

antecedent is false. 258

Shafer-Landau also appeals to innocence by association in arguing that no further

explanation is needed for such conceptual truths. And while the triviality problem is still

around, an appeal to Leibniz’s Law might do the trick. For if, as Shafer-Landau claims, the supervenience thesis is “but a slight variation of the law of identity of indiscernibles”—after all, “supervenience tells us that those things indiscernible at the level of base properties must be identical in moral assessment”—then perhaps it really

doesn’t need any further explanation.259 But this comparison cuts no ice: the superve-

nience thesis in Blackburn’s argument is far stronger than the identity of indiscernibles.

For supervenience, unlike Leibniz’s Law, is not restricted to individuals which share all

their properties; it also applies to individuals which might well share only some of their

properties (natural properties), and insists that they share other properties as well (moral

properties). And indeed, the supervenience of the moral on the natural is structually

identical to any number of obviously false supervenience theses (e.g., heat supervenes on

longitude, salary supervenes on blood-type). Since none of these are “slight variation[s]”

of Leibniz’s Law, it is hard to see why moral supervenience should be.

Perhaps, then, the analogy with Leibniz’s Law can be salvaged with some help from

constitution. For if Shafer-Landau’s constitution thesis is true, then we have

supervenience as a direct consequence: as he puts it: “necessarily, if any grouping of base

258 Mabrito recognizes that this would be “vacuous” (2005: 298–301). Triviality complaints about Blackburn’s similar formulation can be found in Dreier (1993: 17); Zangwill (1995: 260 n. 3); and Majors (2009: 43). 259 Shafer-Landau (1994: 147–48; 2003: 88; 2005: 328). 199

properties B* underlies (because it constitutes) property S, then (in that world) anything else that is also B* must be S.”260 And this constitution relation brings supervenience closer to Leibniz’s Law: if the base properties of the supervenience thesis constitute all the remaining higher-level properties, then any two individuals which share their base properties will turn out to share all their properties. Base-indiscernibles will be indiscernibles simpliciter. And then supervenience, like Leibniz’s Law, might be so fundamental as to need no further explanation. Or, if supervenience does stand in need of a further explanation, then constitution can help with that: “[n]on-naturalists (and non- reductive ethical naturalists) will cite this relation [constitution] as the basis for explaining the supervenience of the moral on the descriptive”.261 So an appeal to constitution should either give us a further explanation of supervenience, or else help us see why no further explanation is needed.

But here Shafer-Landau’s response meets with a serious obstacle: the constitution thesis is not itself a conceptual truth, for Moorean non-naturalists and expressivists deny that the moral is constituted by the natural without misusing moral concepts. But then it is hard to see how an appeal to constitution can help with supervenience as a conceptual truth. It certainly cannot give us a further explanation: even if constitution explains supervenience itself, it won’t even begin to explain why supervenience is a conceptual truth. Nor can it help us see why no further explanation is needed: it may well be a fundamental inexplicable truth reminiscent of Leibniz’s Law that if the moral is consti-

260 Shafer-Landau (2003: 88). As is argued by Ridge (2007), this claim relies on some potentially controversial assumptions about constitution. 261 Shafer-Landau (2003: 86). 200

tuted by the natural then it supervenes upon the descriptive, but we cannot therefore say

the same for the constitution thesis itself or the supervenience thesis itself. I would

conjecture that Shafer-Landau has lost sight of the fact that what Blackburn wants

explained is not so much supervenience as its status as a conceptual truth. For while an

appeal to constitution might explain the former, it could not explain the latter. We are

thus left wondering why supervenience acts as a conceptual constraint on moral

evaluation.

7.4 Natural kinds

In my view, the most challenging objection to Blackburn’s argument is the one dwelt on

by Blackburn himself262 and pressed by James Dreier.263 It draws a parallel between

morality and natural kinds, finds that the two theses and the alleged mystery can be found

in both cases alike, and concludes that moral realism is no more mysterious than realism about natural kinds. And this conclusion would clearly overturn the argument, in effect taking a denying the mystery approach.

To see the parallel between morality and natural kinds, note that something like the

two theses of supervenience and lack of entailment must hold for natural kinds. The

natural kind water clearly supervenes on the chemical compound H2O: since the two are

outright identical with each other, it is a fortiori impossible for two individuals (or entire

worlds) to be alike with respect to H2O and yet different with respect to water. Moreover,

262 Blackburn (1985a/1993: 141–44). 263 Dreier (1993: 19–21). 201

there is no conceptual entailment: failure to recognize that a sample of H2O is water may be a sign of poor education, but not of conceptual incompetence.

Admittedly, Blackburn can (and does) point to an important disanalogy which promises to break the parallel. Whereas moral supervenience functions as a conceptual constraint, the supervenience of a natural kind on its physicochemical constitution does not go quite that far. If you think two physically and chemically identical samples might differ in respect of their water-status, you are making quite a mistake, but this is no sign of conceptual incompetence. People who are fully competent with the concept of water may have no idea of its relation to physicochemical constitution, perhaps thinking some additional factor makes the difference between water and something else, perhaps deny- ing that its constitution even counts as a factor in the first place.264

But the disanalogy doesn’t quite stick. For (as Blackburn himself goes on to note) different communities might have different standards for competence with natural kind concepts, and we can imagine a community whose standards do correspond to the two theses. According to such a community’s standards, we are not required to have specialist knowledge (e.g., that water is H2O, that baking soda is NaHCO3), thus securing the lack

of entailment thesis; but we are required to have framework knowledge (e.g., that water supervenes upon some or other physicochemical constitution), so that the conceptual constraint built into the supervenience thesis is also secured. Thanks to these modified standards, morality and natural kinds are back on an equal footing.265

264 Blackburn (1985a/1993: 142). 265 Blackburn (1985a/1993: 143). Blackburn credits this objection to Michael Smith. 202

Dreier uses a different method toward the same end. He simply stipulates a natural kind term for which the two theses hold:

Surely we could introduce a term that had this feature, that one would have to know that it named a physical kind, but not necessarily which kind (in the sense of being able to specify it in a canonical way, as a material scientist would) to be a competent user. Let us call “cosmite” whatever element predominates in the constitution of Halley’s Comet. We let “cosmite” be a rigid designator whose reference is fixed by the description “the element that predominates in Halley’s Comet.” Now, you and I know that being made of cosmite supervenes on something’s physical properties, but we don’t know (at least I don’t!) which properties, we don’t know any relevant reduction sentences.266

Either way, then, through modified standards or through sheer stipulation, it seems we can find natural kind concepts for which the two theses hold, so that the parallel is restored. Both cases involve a ban on mixed worlds, which means (once again) that moral realism seems no more mysterious than realism about natural kinds. Blackburn’s argument is in need of a further line of response.

Now, in developing a response that might save the argument, I will be drawing on some fairly sketchy comments from Blackburn that have attracted the attention of Dreier.

The basic idea is that although the two theses do generate a mystery wherever they occur—both for morality and for natural kinds—realists about natural kinds are in a far better position to handle the mystery, for they have an easy and satisfying explanation that cannot be borrowed by moral realists. Here is a major chunk of what Blackburn says:

The [ban on mixed worlds in the case of natural kinds] is explicable purely in terms of beliefs of ours—in particular, a belief that we suppose competent people to share. We believe… that no two things could be identical physically without also forming the same stuff or kind and we believe that all competent people will agree. [The ban] has now been explained purely by the structure of beliefs that can coexist with competence… There is indeed no further inference to a metaphysical conclusion

266 Dreier (1993: 20). 203

about the status of wateriness, because the explanation which, in the other cases, that inference helps to provide, is here provided without it. To put it another way, we could say that in the moral case as well, when we deal with analytically possible worlds, we are dealing with beliefs we have about competence: in this case the belief the competent person will not flout supervenience. But this belief is only explained by the further, anti-realist nature of moralizing. If moralizing were depicting further, moral aspects of reality, there would be no explanation of the conceptual constraint, and hence of our belief about the shape of a competent morality.267

Now, it is not obvious what the proposed easy and satisfying explanation is supposed to be, beyond the claim that it works “purely in terms of beliefs of ours”. And Dreier takes the explanation to bottom out in something like sheer arbitrary convention: we have decided to adopt certain standards of conceptual competence in the case of natural kinds, and the ban on mixed worlds is a result of this decision and nothing else. But then we need to know why moral realists could not say the same thing, i.e. that the ban on mixed worlds in the case of morality is simply a matter of arbitrary conventions we have decided to adopt. Dreier can only offer the suggestion that perhaps Blackburn sees so humble an explanation as somehow incompatible with realism, or “giv[ing] the game away to the antirealist”. But as Dreier points out, that is just not tenable:

Taken literally, [Blackburn’s comments] impl[y] that no realist could ever explain any analytic connections among concepts. Imagine a challenge to a realist about bachelorhood and marriage, asking her to explain the necessary connection between being a bachelor and being unmarried. Why is it that anyone who claims, “Juan is a bachelor”, but denies “Juan is unmarried,” thereby shows that she is confused about the predicates in those sentences? The realist tries to explain in the obvious way: the meaning of “bachelor” logically implies that anyone who satisfies it also satisfies the predicate “unmarried”, given the meaning of “unmarried”. It is a fact about our language, about the way we use words that explains the necessary connection between the two sentences. But Blackburn replies, “Aha, you have given the game away to us antirealists about bachelorhood. For any explanation that appeals to our linguistic practices, rather than to metaphysical facts, commits one to antirealism.”

267 Blackburn (1985a/1993: 143). 204

That would be silly—of course bachelorhood is a real property, and of course the realist is entitled to appeal to semantic facts to explain analytic connections.268

So interpreted, then, Blackburn’s comments do nothing to save his argument. If this style of conventionalist explanation works equally well for morality and for natural kinds, then

(by switching to an offering an explanation approach) moral realists can rest easy.

I think Blackburn’s comments should be taken in a different direction. Rather than seeing the easy and satisfactory explanation as bottoming out in sheer arbitrary convention, we should see it as bottoming out in empirical discovery. Presumably those of us interested in natural kinds would not have adopted a supervenience constraint just for kicks: instead, we collected a good deal of empirical information in the process of investigating natural kinds and their physicochemical constitution, and it was this information that led us to adopt such a constraint. Thus Blackburn writes that “the best explanation of why competent people recognise the supervenience of kinds upon physical or chemical structure is that we live in a culture in which science has found this out”269.

So in the case of natural kinds, if we have adopted a set of conceptual constraints that requires framework knowledge (e.g. supervenience) while leaving out specialist knowledge, this is to be explained on the basis of empirical discoveries made in the course of investigating things and classifying them by natural kind.

But if this is the easy and satisfactory explanation for the ban on mixed worlds, why can’t moral realists say the same thing? In short, my answer is that moral supervenience cannot be plausibly construed as the product of empirical discoveries made in the course

268 Dreier (1993: 20–21). 269 Blackburn (1985a/1993: 144). 205

of investigating things and evaluating them morally. Consider the sequence of events in the case of natural kinds: our practice of classifying things by natural kind comes first, unrestricted by any supervenience constraint, and it is only later, after this practice has given us a piece of general empirical knowledge (viz., that natural kinds supervene on constitutions), that a supervenience constraint finds its way into our concepts and begins regulating our attributions of competence. This is plausible enough. But now consider the same sequence in the case of morality: our practice of moral evaluation would come first, unrestricted by any supervenience constraint, and it is only after this practice has given us a piece of general knowledge (viz., that the moral supervenes on the natural), that a supervenience constraint finds its way into our concepts and begins regulating our attributions of competence. Such a sequence involves serious implausibilities. First, it would have moral evaluation taking place prior to, and in the absence of, any supervenience constraint. Second, it would construe the supervenience constraint as the product of discoveries made in medias res, as if moral evaluators began to pick up on a supervenience pattern, found no counterexamples to it, and eventually elevated it to the status of conceptual constraint.

Thus the difference between natural kinds and morality is that the supervenience constraint is far more deeply embedded in the case of morality. We can make sense of a practice of classifying natural kinds which is unbound by any supervenience constraint, but we cannot make sense of moral evaluation unbound by any supervenience constraint.

We can make sense of kind-constitution supervenience as a discovery made in the course 206

of classifying natural kinds, but we cannot make sense of moral-natural supervenience as

a discovery made in the course of moral evaluation.

This point has implications beyond the current objection. For in my view, it is this—

the fact that supervenience seems to constrain moral evaluation from the very outset— that ultimately gives Blackburn’s argument its force.270 On the realist construal of moral

evaluation, we are aiming to describe certain facts in the world. And it is not at all clear

why we would have such sweeping expectations about moral facts as to embrace moral

supervenience on day one, even as a mere tentative conjecture. More puzzling would be

its adoption as an indubitable conviction of common sense: in any world-guided practice,

it makes sense to allow that things might turn out to differ from our original expectations.

But most puzzling of all would be its adoption as a full-fledged conceptual constraint

binding the very practice of moral evaluation: this takes overweening confidence to a

new level. Far better, it would seem, to explain moral evaluation as a practice whose con-

straints were adopted without regard for any facts to be described, but instead with regard

to facts to be evaluated nondescriptively.

270 I believe the point also makes sense of some of Blackburn’s remarks elsewhere: “[A realist] has the conception of an actual A-state of affairs, which might or might not distribute in a particular way across the B-states. Supervenience then becomes a mysterious fact, and one which he will have no explanation of (or no right to rely upon)” (1984: 185); “[On realism], each proposition may be construed as asserting that a certain state of affairs exists… And… the definite consequence is that the truth of propositions should be subject to just the logical constraints that govern the existence of states of affairs… Supervenience becomes, for the realist, an opaque, isolated, logical fact for which no explanation can be proffered (1971/1993: 119). 207

7.5 Other targets

If I am right that Blackburn’s argument is more anti-descriptivist than anti-realist, then the argument should target other descriptivist metaethical theories apart from realism. To illustrate this point, I will provide a brief consideration of error-theory and some forms of constructivism, including response-dependence views.

By now it should not be too difficult to see how error-theory would fall under the scope of the argument. For even if error-theorists do not accept the metaphysical commitments of realists, they agree with realists about the nature of moral evaluation, construing it as a descriptive practice aimed at faithfully representing moral facts.271 And

the big problem for realists—the problem of explaining why such a practice would

incorporate supervenience as a deeply embedded conceptual constraint in the absence of

conceptual entailment—is present regardless of whether the moral facts presupposed by

the practice actually exist. Error-theorists and realists agree on everything but

metaphysics; Blackburn’s supervenience argument does not concern itself with me-

taphysics; so the argument applies to realists and error-theorists alike.

But because of their differing metaphysical commitments, error-theorists and realists

have different resources available for their attempts to provide an explanation. Thus for

realists, supervenience is not only a conceptual constraint on coherent moral evaluation,

but also a metaphysical truth about real moral facts. This means that one natural move for

realists hoping to provide an explanation is to understand the conceptual constraint as

somehow resulting from the metaphysical truth: first we pick up on the supervenience

271 Mackie (1977: esp. 30–35). 208

pattern in reality and then this pattern somehow finds its way into our standards for

conceptual competence. Now, as I have argued, this kind of explanation does not seem to

do justice to the deeply embedded character of the supervenience constraint. But what is

important here is that no such explanation is available for error-theorists: they could not likewise help themselves to metaphysical truths about moral supervenience, not without admitting an ontology of moral facts and thereby giving up on their error-theory. In short: realists can try starting with moral metaphysics and then moving to human concepts, but this path is closed to error-theorists.

Error-theorists must therefore attempt to provide the mysterious conceptual constraints with a more human origin. In Mackie’s discussion of “patterns of objectification”, we can see the beginnings of such an explanation.272 Humans have a

tendency to project our moral attitudes and our desires onto things, investing the resulting

ersatz qualities with an air of “absolute authority” taken from the external social

pressures and codes of secular and religious law that have been internalized in the form

of the original attitudes and desires. Now, as Mackie points out, these attitudes and

desires are responsive to the features of the thing under examination, and there are strong

social motives for a sort of universalization:

We need morality to regulate interpersonal relations, to control some of the ways in which people behave towards one another, often in opposition to contrary inclinations. We therefore want our moral judgements to be authoritative for other agents as well as for ourselves: objective validity would give them the authority re- quired.273

272 Mackie (1977: 42–46). 273 Mackie (1977: 43). 209

But if the resulting ersatz qualities are treated as depending on underlying features, and in a way that must not vary across individuals and situations, we are well on our way to a bona fide supervenience constraint. Error-theorists have a promising explanation on their hands.

Unfortunately, this explanation sounds a little too familiar. It works with the exact same social needs that drive the expressivist explanation offered by Blackburn. The only difference is that the attitudes and desires are not merely expressed in realist-sounding language, but also coupled with bona fide realist beliefs. But you might ask: what’s wrong with offering an explanation similar to that offered by expressivists? Why should that matter, so long as the explanation is indeed a satisfying one?

The problem is that even if the explanation is satisfying at explaining why a supervenience constraint would emerge in the early stages of social pressures and practical attitudes, it fails to explain how and why the constraint carries over into the era of objectified qualities and realist beliefs. To see why, suppose that the nondescriptive factors at work in the early stages of Mackie’s account do conform to a supervenience constraint. That is, we are required to have the same moral attitudes and the same desires toward the same kind of situation, naturalistically described, on pains of social pressure.

Now once we begin to project our attitudes and desires onto things and acquire realist beliefs, this supervenience constraint becomes highly questionable. We constrain our moral beliefs in such a way as to rule out apparently possible distributions of moral properties as impossible, but this constraint governs our moral beliefs only because there is a corresponding constraint governing the nondescriptive attitudes that this descriptive 210

practice originates in. That is, a constraint on belief that might potentially cut us off from moral truth is acquiesced in simply because of a prior constraint on nondescriptive attitudes.

This casts doubt on the error-theorist interpretation of moral evaluation as a primarily descriptive affair. If the purpose of moral evaluation is to describe moral facts as they actually are, independently of us and our practices, then why is this purpose held hostage to a constraint on nondescriptive attitudes which results from human needs and nothing more? Or, in other words, if the realist beliefs we find associated with moral evaluation are invariably restricted by socially useful constraints on nondescriptive attitudes, then shouldn’t we move the realist beliefs to the periphery and place the nondescriptive attitudes at the center of our account of moral evaluation? Mackie’s account as we have been understanding it helps itself to a pre-existing practice of social regulation through moral attitude, a practice which already features a supervenience constraint, and then subordinates the realist beliefs which follow to this practice. But this only fuels the expressivist interpretation of moral evaluation, on which any realist beliefs are excess baggage and nondescriptive attitudes form the heart and of morality.274

Of course, I have been supposing that the supervenience constraint can be attributed to the nondescriptive factors at work in the early stages of the account. And what if these factors do not conform to a supervenience constraint? On this horn of the dilemma, there is a different problem: it becomes mysterious how these constraints could be introduced into the practice by the descriptive factors that come later. After all, these objectified

274 Cf. Blackburn (1985b). 211

qualities and realist beliefs are not supposed to float free of the nondescriptive factors that precede them; on the contrary, they are supposed to be outgrowths and projections of these factors. So without an additional explanation of how and why the descriptive factors might bring these constraints into the practice, when there is no apparent reason why the moral facts we hope to describe would conform themselves to any such constraints, the ban on mixed worlds remains as mysterious as ever.

Thus while error-theorists can avail themselves of a down-to-earth expressivist-style explanation of the ban on mixed worlds, they do so at their own peril. For, the more explanatory work is done by nondescriptive factors, and the more lasting control these factors exercise on the resulting descriptive beliefs, the less plausible it is to insist that they are only secondary and subordinate to an essential descriptive function of moral evaluation. Or, if error-theorists wish to credit descriptive factors with having introduced the constraints in question, they are in need of some additional explanation, and are thus on all fours with realists.

Of course, error-theory is not the only descriptivist alternative to realism—there is also the constructivist idea that morality aims to describe certain key psychological features of certain key persons. But constructivism comes in a wide variety of flavors, from the austere objectivity of ideal-observer theories to the unabashed subjectivity of moral relativism. And thus constructivist theories are not all equally vulnerable to

Blackburn’s argument. When constructivism aims to describe something fixed and independent of the social practices developed to meet human needs, it has a hard time accounting for the supervenience constraint. But when it joins expressivists in tying 212

moral evaluation to our social practices, constructivism can escape the sweep of the

argument.

As an example of a constructivist theory that is more objective and less tied to our

practices, consider Kantian constructivism. According to such a view, there is a

distinctive perspective of practical decision-making that all rational beings can adopt, this

perspective provides us with a uniform system of verdicts regimenting various intentional

actions as forbidden or not, and moral evaluation purports to accurately describe the

verdicts of this perspective.275 This view is in the same boat as realism. We need some

antecedent reason for expecting the verdicts of Kantian practical reason to conform to a

supervenience constraint, so that we might understand why such a constraint has

governed moral evaluation from the very outset of the practice. After all, there is no

obvious impossibility in the thought that a faculty of our psychology—be it empirical or

transcendental—might turn out to issue verdicts that run afoul of such a constraint. And it

is not as if the nature of such a faculty could be altered by the social practices we choose

to adopt: if we aim to describe it as it actually is, and describe its verdicts as they actually

are, then it mysterious that we would rule out apparent possibilities as conceptually

impossible from the outset.

But now consider a constructivist theory that less objective and more tied to our

practices, consider Humean constructivism. According to such a view, there are flexible

social norms that humans have developed to meet their needs, these norms provide us

with a fuzzy system of verdicts assessing different character traits and actions as virtuous

275 Of course, I am offering a simplification of a very difficult view. See Korsgaard (2009); Rawls (1980); Velleman (2000). 213

or not, and moral evaluation purports to accurately describe the assessments of these

norms.276 This view is in the same boat as expressivism. We have a very good antecedent

reason for expecting these norms and their assessments to conform to a supervenience

constraint: namely, the norms were developed to meet social needs that are best met with

a supervenience constraint. And this explains why a practice whose function is to

describe the norms would have adopted a supervenience constraint for itself from the

outset. It is not as if the nature of these norms could be totally independent of the practice

of moral evaluation: after all, the practice and the norms are subject to the exact same

social pressures, so that changes in the one can be reasonably expected to correspond to

changes in the other.

Of course, I do not mean to exaggerate the similarities between expressivism and

Humean constructivism, or any other similar form of constructivism. Expressivism

characterizes moral evaluation as primarily a matter of nondescriptive attitudes, whereas

constructivists of a sentimentalist stripe characterize moral evaluation as primarily a matter of descriptive beliefs about nondescriptive attitudes. Thus constructivists can provide moral judgments with orthodox truth-conditions, and expressivists must rest content with the sort of unorthodox truth-conditions associated with deflationary theories of truth. I only mean to suggest that both kinds of views are well-placed to explain the fact that moral evaluation is governed by a deeply-embedded supervenience constraint.

Response-dependence views can be plausibly classified as a kind of constructivism, and indeed they face similar problems. Take, for example, Firth’s well-known ideal

276 This simplification is inspired by Hume himself, and the ongoing work of Sharon Street, Valerie Tiberius, and Dale Dorsey. 214

observer analysis of moral judgments: “Any ideal observer would react to x [the

evaluated individual] in such and such a way under such and such conditions”.277 What

must be explained is why we would expect the reactions of any ideal observer to conform

to a supervenience constraint. If the answer is that we have explicitly built in such a

constraint into our conception of an ideal observer, then we have simply helped ourselves

to the constraint without explaining why it was adopted in the first place. And if the

supervenience constraint is supposed to follow from other conceptually required features

of the ideal observer, then we need further explanation of how these features are

connected to supervenience and how they got their status as conceptual truths.278 Similar

points apply to Michael Smith’s rationalist response-dependence view:279 if the view is

not to simply help itself to the supervenience constraint (e.g. taking it as an a priori

platitude without explaining how it got this status280), then we need an explanation for

why the desires of an idealized “fully rational” self must conform to such a constraint. As

before, going Humean might do the trick: if our normative conception of what counts as

‘ideal’ or ‘rational’ is itself the product of human social needs, then we can explain why a

supervenience constraint would have been adopted. But this would put response- dependence views on a softer foundation than their proponents might like: moral

277 Firth (1952: 321). 278 Cf. Firth on the status of “consistency” (1952: 341–44). 279 Smith (1994). 280 Smith (1994: 40). For similar reasons, the analytic moral functionalism of Jackson and Pettit seems to be vulnerable to Blackburn’s argument. For while they accept particular natural-to-moral conceptual entail- ments and thus deny the lack of entailment thesis, this is only because they have helped themselves to the supervenience thesis when characterizing the (mature) folk morality Ramsey sentence that grounds the en- tailments. Thus it seems that Jackson is only begging the question when he writes: “Many, but Simon Blackburn in particular, have properly demanded an explanation of the supervenience of the ethical on the descriptive. The answer, it seems to me, is given by the a priori nature of the supervenience: it tells us that it is part of our very understanding of ethical vocabulary that we use it to mark distinctions among the descriptive ways things are” (1998: 125). 215

evaluation may be given an absolutist or rationalist treatment of sorts, but the ultimate

normative underpinnings turn out to be a matter of mere social coordination and non-

descriptive attitudes.

7.6 Conclusion

Blackburn’s argument focuses our attention on the origin and the continued

underpinnings of the conceptual constraints that shape our practice of moral evaluation.

What realists (and those sympathetic to realist construals of moral evaluation) need is an

account of these constraints that makes them more than a quirk of human psychology, a

need of human society, or a well-supported empirical generalization. The constraints must have the right to dictate our metaphysics once and for all: the right to rule out apparent possibilities prior to and independent of any moral views we might come to have in the course of evaluating the world. And while this task may not be an impossible one, it does not look to be an easy one. Thus Blackburn’s argument earns its place as an important challenge to moral realism, and indeed as an important line of support for views which, like expressivism and Humean constructivism, put nondescriptive attitudes at the center of moral evaluation.

216

WORKS CITED

Adams, R. M. 1999. Finite and Infinite Goods: A Framework for Ethics. Oxford: Oxford University Press.

Ayer , A. J. 1936. Language, Truth, and Logic. London: Gollancz.

Blackburn, S. 1971. “Moral Realism.” In J. Casey (ed.) Morality and Moral Reasoning. London: Methuen. 101–24. Reprinted in his Essays in Quasi-Realism. New York: Oxford University Press. 1993. 111–29.

Blackburn, S. 1984. Spreading the Word. Oxford: Clarendon Press.

Blackburn, S. 1985a. “Supervenience Revisited.” In I. Hacking (ed.) Exercises in Analysis. Cambridge: Cambridge University Press. 47–67. Reprinted with an addendum in his Essays in Quasi-Realism. New York: Oxford University Press. 1993. 130–145.

Blackburn, S. 1985b. “Errors and the Phenomenology of Value.” In T. Honderich (ed.) Morality and Objectivity: A Tribute to J. L. Mackie. London: Routledge & Kegan Paul. 1–22. Reprinted in his Essays in Quasi-Realism. New York: Oxford University Press. 1993. 149–65.

Blackburn, S. 1993. Essays in Quasi-Realism. New York: Oxford University Press.

Blackburn, S. 1998. Ruling Passions. Oxford: Clarendon Press.

Blackburn, S. 2005. “Quasi-Realism No Fictionalism.” In M. Kalderon (ed.) Fictionalism in Metaphysics. Oxford: Oxford University Press. 322–37.

Block, N. 1986. “Advertisement for a Semantics for Psychology.” In P. A. French, T. E. Uehling and H. K. Wettstein (eds.), Midwest Studies in Philosophy, X. Minneapolis: University of Minnesota Press. 615–678.

Bloomfield, P. 2001. Moral Reality. Oxford: Oxford University Press. 217

Bovens, L. and D. Drai. 1999. “Supervenience and Moral Realism.” Philosophia. 27. 241–45.

Boyd, R. 1988. “How To Be a Moral Realist.” In Geoffrey Sayre-McCord (ed.), Essays on Moral Realism. Ithaca, NY: Cornell University Press. 181–228.

Brink, D. 1989. Moral Realism and the Foundations of Ethics. Cambridge: Cambridge University Press.

Copp, D. 2007. Morality in a Natural World. Cambridge: Cambridge University Press.

Churchland, P. M. 1984. “Subjective from a Materialist Point of View.” PSA: Proceedings of the Biennial Meeting of the Association, 2. 773–90.

Dancy, J. 2006. “Nonnaturalism.” In D. Copp (ed.) The Oxford Handbook of Ethical Theory. Oxford: Oxford University Press. 122–45.

Dreier, J. 1993, “The Supervenience Argument Against Moral Realism.” Southern Journal of Philosophy. 30. 13–38.

Durrant, R. G. 1970. “Identity of Properties and the Definition of ‘Good’.” Australasian Journal of Philosophy, 48. 360–61.

Dworkin, R. 2011. Justice for Hedgehogs. Cambridge, MA: Belknap Press of Harvard University Press.

Elliot, R. 1987. “Moral Realism and the Modal Argument.” Analysis. 47. 133–37.

Enoch, D. Forthcoming. Taking Morality Seriously: A Defense of Robust Realism. March 2010 draft.

Finlay, Stephen. 2007. “Four Faces of Moral Realism”, Philosophy Compass, 2. 218

Firth, R. 1952. “Ethical Absolutism and the Ideal Observer.” Philosophy and Phenomenological Research. 12. 317–45.

FitzPatrick, W. J. 2008. “Robust Ethical Realism, Non-Naturalism, and Normativity.” In R. Shafer-Landau (ed.) Oxford Studies in Metaethics, volume 3. Oxford: Clarendon Press. 159–205.

Fodor, J., and Lepore, E. 1992. Holism: A Shopper’s Guide. Cambridge: Blackwell.

Foot, P. 1958. “Moral Arguments.” Mind 67. 502–13.

Gibbard, A. 1990. Wise , Apt Feelings: A Theory of Normative Judgment. Cambridge, MA: Harvard University Press.

Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.

Hare, R. M. 1952. The Language of Morals. Oxford: Oxford University Press.

Hare, R. M. 1957. “Nothing Matters.” In his Applications of Moral Philosophy. New York: MacMillan. 1972. 32–47.

Horgan, T. and M. Timmons. 1990–91. “New Wave Moral Realism Meets Moral Twin Earth.” Journal of Philosophical Research. 16. 447–65.

Horgan, T. and M. Timmons. 2000. “Copping Out on Moral Twin Earth.” Synthese. 124. 139–52.

Horgan, T. and M. Timmons. 2006. “Cognitivist Expressivism.” In T. Horgan & M. Timmons (eds.) Metaethics after Moore. Oxford: Clarendon Press.

Huemer, M. 2005. Ethical Intuitionism. New York: Palgrave Macmillan.

Hume, D. 1739–40. A Treatise of Human Nature, edited by David Fate Norton and Mary J. Norton. New York and Oxford: Oxford University Press, 2000. 219

Jackson, F. & P. Pettit. 1996. “Moral Functionalism, Supervenience and Reductionism.” Philosophical Quarterly. 46. 82–6.

Jackson, F. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Clarendon Press.

Joyce, R. 2001. The Myth of Morality. Cambridge: Cambridge University Press.

Kalderon, M. E. 2005. Moral Fictionalism. Oxford: Oxford University Press.

Kim, J. 1984. “Concepts of Supervenience.” Philosophy and Phenomenological Research. 45. 153–76.

Kim, J. 1990. “Supervenience as a Philosophical Concept.” Metaphilosophy. 21. 1–27.

Klagge, J. C. 1984. “An Alleged Difficulty Concerning Moral Properties.” Mind. 93. 370–80.

Korsgaard, C. 2009. Self-Constitution: Agency, Identity, and Integrity. Oxford: Oxford University Press.

Kripke, S. 1980. Naming and Necessity. Cambridge: Harvard University Press.

Lenman, J. 2010. “Uggles and Muggles.” Philosophical Studies.

Lewis, D. 1980. “What Teaches”, Proceedings of the Russellian Society. 13. 29–57.

Mabrito, R. 2005. “Does Shafer-Landau Have a Problem with Supervenience?” Philosophical Studies. 126. 297–311.

Mackie, J. L. 1977. Ethics: Inventing Right and Wrong. New York: Penguin Books.

Mackie, J. L. 1982. The Miracle of Theism. Oxford: Clarendon Press. 220

Majors, B. 2009. “The Natural and the Normative.” In R. Shafer-Landau (ed.) Oxford Studies in Metaethics, volume 4. Oxford: Oxford University Press. 29–52.

McFetridge, I. G. 1985. “Supervenience, Realism, Necessity.” Philosophical Quarterly. 35. 245–58.

Moore, G. E. 1903. Principia Ethica, revised edition, edited by T. Baldwin. Cambridge: Cambridge University Press. 1993.

Moore, G. E. 1922. “The Conception of Intrinsic Value.” In his Philosophical Studies. London: Routledge & Kegan Paul. 253–75. Reprinted in Principia Ethica, 280–98.

Moore, G. E. 1942. “A Reply to my Critics.” In P. A. Schilpp (ed.), The Philosophy of G. E. Moore. Evanston, IL: Northwestern University Press. 535–677.

Moore, G. E. 1965. Ethics. New York: Oxford University Press.

Nagel, T. 1986. The View from Nowhere. New York: Oxford University Press.

Norton, D. F. 1982. : Common-Sense Moralist, Sceptical Metaphysician. Princeton: Princeton University Press.

Olson, J. 2009. “Reasons and the New Non-Naturalism.” In S. Robertson (ed.) Spheres of Reason: New Essays in the Philosophy of Normativity. Oxford: Oxford University Press.

Parfit, D. 2006. “Normativity.” In R. Shafer-Landau (ed.) Oxford Studies in Metaethics, volume 1. Oxford: Clarendon Press. 325–80.

Plantinga, A. 1974. The Nature of Necessity. Oxford: Clarendon Press.

Plantinga, A. 2010. “Naturalism, Theism, Obligation, and Supervenience.” Faith and Philosophy. 27. 247–72.

Putnam, H. 2004. Ethics without Ontology. Cambridge, MA: Harvard University Press. 221

Rawls, J. 1980. “Kantian Constructivism in Moral Theory.” The Journal of Philosophy. 77. 515–72.

Ridge, M. 2007. “Anti-Reductionism and Supervenience.” Journal of Moral Philosophy. 4. 330–48.

Russell, P. 2008. The Riddle of Hume’s Treatise: Skepticism, Naturalism, and Irreligion. New York and Oxford: Oxford University Press.

Scanlon, T. M. 2003. “Metaphysics and Morals.” Proceedings and Addresses of the American Philosophical Association. 77. 7–22.

Schroeder, M. 2007. Slaves of the Passions. New York: Oxford University Press.

Shafer-Landau, R. 1994. “Supervenience and Moral Realism.” Ratio. 7. 145–52.

Shafer-Landau, R. 2003. Moral Realism: A Defence. Oxford: Oxford University Press.

Shafer-Landau, R. 2005 “Replies to Critics.” Philosophical Studies. 126, 313–29.

Shoemaker, S. 1987. “Review of Simon Blackburn’s Spreading the Word.” Noûs. 21. 438–42.

Sinnott-Armstrong, Walter. 2006. Moral Skepticisms. New York: Oxford University Press.

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

Sobel, J. H. 2001. “Blackburn’s Problem: On Its Not Insignificant Residue.” Philosophy and Phenomenological Research. 62. 361–83.

Stevenson, C. L. 1937. “The Emotive Meaning of Ethical Terms.” Mind. 46. 14–31. 222

Sturgeon, N. 2009. “Doubts about the Supervenience of the Evaluative.” In R. Shafer- Landau (ed.) Oxford Studies in Metaethics, volume 4. Oxford: Oxford University Press.

Timmons, M. 1999. Morality without Foundations: A Defense of Moral Contextualism. New York: Oxford University Press.

van Roojen, M. 2009. “Moral Cognitivism vs. Non-Cognitivism.” Stanford Encyclopedia of Philosophy. Available here: http://plato.stanford.edu/entries/moral- cognitivism/Velleman, J. D. 2000. The Possibility of Practical Reason. New York: Oxford University Press.

Wedgwood, R. 2006. “The Meaning of ‘Ought’.” In Russ Shafer-Landau (ed.), Oxford Studies in Metaethics, I. Oxford: Clarendon Press. 127–60.

Wedgwood, R. 2007. The Nature of Normativity. Oxford: Clarendon Press.

Wedgwood, R. 2010. “The Nature of Normativity: A Reply to Holton, Railton, and Lenman.” Philosophical Studies.

Whiting, D. 2009. “Conceptual Role Semantics.” Internet Encyclopedia of Philosophy. Available here: http://www.iep.utm.edu/conc-rol/

Wierenga, E. 1983. “A Defensible Divine Command Theory.” Nous, 17. 387–407.

Zangwill, N. 1995. “Moral Supervenience.” Midwest Studies In Philosophy. 20. 240–62.