realism, naturalism, and moral semantics

Social Philosophy and Policyhttp://journals.cambridge.org/SOY

Additional services for Social Philosophy and Policy:

Email alerts: Click hereSubscriptions: Click hereCommercial reprints: Click hereTerms of use : Click here

Realism, Naturalism, and Moral Semantics

David O. Brink

Social Philosophy and Policy / Volume 18 / Issue 02 / June 2001, pp 154 176DOI: 10.1017/S0265052500002946, Published online: 13 January 2009

Link to this article: http://journals.cambridge.org/abstract_S0265052500002946

How to cite this article:David O. Brink (2001). Realism, Naturalism, and Moral Semantics. Social Philosophy and Policy,18, pp 154176 doi:10.1017/S0265052500002946

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/SOY, IP address: 128.42.202.150 on 08 Sep 2012

REALISM, NATURALISM, AND MORAL SEMANTICS*

BY DAVID O. BRINK

I. INTRODUCTION

The prospects for moral realism and ethical naturalism have been im-portant parts of recent debates within metaethics. As a first approxima-tion, moral realism is the claim that there are facts or truths about moralmatters that are objective in the sense that they obtain independently ofthe moral beliefs or attitudes of appraisers. Ethical naturalism is the claimthat moral properties of people, actions, and institutions are natural, ratherthan occult or supernatural, features of the world.1 Though these meta-ethical debates remain unsettled, several people, myself included, havetried to defend the plausibility of both moral realism and ethical natu-ralism.2 I, among others, have appealed to recent work in the philosophyof language—in particular, to so-called theories of "direct reference"—todefend ethical naturalism against a variety of semantic worries, includingG. E. Moore's "open question argument." In response to these arguments,critics have expressed doubts about the compatibility of moral realismand direct reference. In this essay, I explain these doubts, and then sketchthe beginnings of an answer—but understanding both the doubts and myanswer requires some intellectual background.

* For helpful discussion of issues addressed in this essay, I would like to thank RichardArneson, Thomas Bontly, David Copp, Michael Moore, Russ Shafer-Landau, Evan Tiffany,Mark Timmons, Steven Yalowitz, the editors of this volume, and an audience at the Uni-versity of British Columbia.

1 This way of understanding ethical naturalism presupposes realism or at least cognitiv-ism insofar as it presupposes that there are moral properties and that moral judgmentsascribing moral properties to person, actions, and institutions can be true or false (with somebeing true). This, I think, is a traditional way of understanding ethical naturalism. It mightbe contrasted with a broader understanding, according to which the ethical naturalist sim-ply tries to fit moral practices and judgments within a naturalistic worldview. Cf. GilbertHarman, "Is There a Single True Morality?" in David Copp and David Zimmerman, eds.,Morality, Reason, and Truth (Totowa, NJ: Rowman and Allanheld, 1984). Notice that, on thisbroad understanding, various forms of moral skepticism and noncognitivism might qualifyas forms of ethical naturalism.

2 David O. Brink, Moral Realism and the Foundations of Ethics (New York: CambridgeUniversity Press, 1989). Cf. Richard Boyd, "How to Be a Moral Realist," in Geoffrey Sayre-McCord, ed., Essays on Moral Realism (Ithaca, NY: Cornell University Press, 1988); PeterRailton, "Moral Realism," Philosophical Review 95, no. 2 (1986): 163-207; and Nicholas Stur-geon, "Moral Explanations," in Copp and Zimmerman, eds., Morality, Reason, and Truth.

154 © 2001 Social Philosophy & Policy Foundation. Printed in the USA.

REALISM, NATURALISM, AND MORAL SEMANTICS 155

II. THE OPEN QUESTION ARGUMENT

In Principia Ethica, G. E. Moore used the open question argument (OQA)to deny that moral properties, such as Tightness or goodness, are naturalor metaphysical (that is, supernatural) properties.3 Moral properties, onhis view, are nonnatural and sui generis. The OQA attempts to establishthis ontological or metaphysical thesis by semantic means.4 The OQAassumes that if moral properties are natural properties, then moral pred-icates can be defined in terms of natural predicates, which Moore appar-ently understood as nonmoral predicates drawn from the natural andsocial sciences (broadly construed). In arguing this way, Moore assumedsomething like the semantic test of properties, according to which predicatespick out the same property just in case they are synonymous. The OQAis supposed to show that no moral predicate is synonymous with anynatural or, more generally, nonmoral predicate. Consider any moral pred-icate 'M' and any nonmoral predicate 'N'. If 'M' and 'N' mean the samething, then it ought to be an analytic truth that N-things are M, just as itis an analytic truth that M-things are M. We can see, though, that there areno such analytic truths. It is not possible to doubt that M-things areM—"Is this M-thing M?" is always a closed question. However, it isalways possible to doubt that N-things are M—"Is this N-thing M?" isalways an open question. The fact that the first question is closed but thesecond is open shows that they are not epistemically equivalent—that is,they differ in what can be believed about them. Epistemic inequivalencewould establish semantic inequivalence if speakers were authoritativeabout the meaning of their words. Speakers would be authoritative aboutthe meanings of their words if a descriptional theory of meaning—accordingto which the meaning of a word or phrase is the set of descriptions orproperties that speakers conventionally associate with it—were true. For,on a descriptional theory, if two terms are synonymous (semanticallyequivalent), speakers competent with both terms must associate the sameproperties or descriptions with both terms and should be able to recog-nize that they do. The epistemic inequivalence of the sentences using 'M'and 'N' implies that speakers associate different descriptions with thoseterms and, hence, that those terms differ in meaning. The semantic test of

3 G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1903), chap. 1,esp. pp. 7-17.

4 A number of commentators have interpreted one strand in the OQA as reflecting Moore'sconcern with the normativity of ethics. On this internalist reading, the OQA claims thatmoral properties could not be natural or supernatural properties, because no natural orsupernatural property has the requisite internal or conceptual connection with practicalreason or motivation that moral properties have. See, for example, Stephen Darwall, AllanGibbard, and Peter Railton, "Toward Fin de Siecle Ethics: Some Trends," Philosophical Review101, no. 1 (1992): 115-89, esp. 117-18. This may well be one strand in the OQA, but much ofthe OQA makes no appeal to internalist assumptions. Though much of my discussioncarries over to the internalist reading, I focus on those parts of the OQA that do notpresuppose internalism.

156 DAVID O. BRINK

properties, then, implies that M and N are different properties. In thisway, the OQA undermines ethical naturalism: the failure of synonymyimplies that moral predicates are not naruralistically definable, and thesemantic test of properties implies that, as a result, moral properties arenot natural properties.

Moore and other intuitionists were cognitivists. According to cognitiv-ism, moral predicates are used to ascribe moral properties to actions,people, and institutions; moral judgments ascribing moral properties tosuch things can be true or false; and moral judgments express the ap-praiser's beliefs about the moral properties that such things possess. Be-cause Moore and other intuitionists rejected not just naturalistic ethics butany attempt to define moral terms in nonmoral terms (Moore rejectedmetaphysical ethics as well), they concluded that moral predicates pickout nonnatural sui generis properties. However, it is arguable that as ahistorical matter, noncognitivism rather than intuitionism was the realbeneficiary of the OQA. Noncognitivists deny that moral judgments ex-press the appraiser's beliefs about the moral properties of actions, insti-tutions, and people; instead, they claim that moral judgments express theappraiser's noncognitive attitudes and commitments, such as her desires.The noncognitivists agreed that the OQA undermines ethical naturalism,but found the metaphysical and epistemological commitments of intu-itionism mysterious and implausible. They reasoned that if ethics couldnot be fit into a naturalistic worldview, then it was better to give up thecognitivist presuppositions of ethical naturalism and intuitionism than toaccept intuitionism's metaphysical and epistemological commitments.

III. NONNATURALISM

The moral realist has two main ways to avoid the noncognitivist legacyof the OQA. One option would be to defend nonnaturalism. Nonnatu-ralism would be problematic for many people if it posited properties thatwere in no way related to familiar natural properties. However, nonnat-uralism, at least as Moore understood it, does not imply this. To committhe "naturalistic fallacy," which the OQA is supposed to expose, is toconfuse two properties that are correlated with each other for each other.5

Hence, the OQA is supposed only to block the identification of moralproperties with natural properties—but the denial of property-identity iscompatible with recognizing various other relations between moral andnatural properties. For example, Moore believes that moral propertiesstrongly supervene on natural properties, in the sense that the naturalproperties of a situation fix or determine its moral properties, such thattwo situations cannot differ in their moral properties without differing intheir natural properties. In his reply to critics, he writes:

5 Moore, Principia Ethica, 10, 13, 14, 16.


I should never have thought of suggesting that goodness was 'non-natural/ unless I had supposed that it was 'derivative' in the sensethat, whenever a thing is good (in the sense in question) its goodness. . . 'depends on the presence of certain non-ethical characteristics'possessed by the thing in question: I have always supposed that itdid so 'depend,' in the sense that, if a thing is good (in my sense),then that it is so follows from the fact that it possesses certain naturalintrinsic properties, which are such that from the fact that it is goodit does not follow conversely that it has those properties.6

One explanation of why moral properties supervene on natural ones isthat they are constituted by, but are not identical with, complex config-urations of natural properties. On this view, moral properties stand tonatural properties much as a statue stands to the bronze out of which itis constituted. If so, there is one sense in which moral properties arenothing over and above the natural properties on which they supervene:the natural properties of a situation fix or determine its moral properties;the moral properties of a situation do not have to be added separatelyto the natural properties of that situation. In another sense, though, themoral properties are something over and above the natural properties onwhich they supervene. The same moral properties could have been real-ized by somewhat different configurations of natural properties; this modaldifference implies that the properties are different properties, and groundsthe constitution claim rather than the identity claim. This constitution-relation is a common one. For instance, one could conceive of a lump ofbronze with a particular shape as a statue, in the sense that the statue isconstituted by, but is not identical with, the bronze. This is preciselybecause the statue could survive changes in the particular shape of thebronze or could have been realized by a somewhat different bronze shape.Insofar as this view about the relation between the moral and the naturalappeals to the 'is' of constitution rather than the 'is' of identity, it evenallows us to say that moral properties are natural properties. Moreover,this constitution-relation is thought by some to characterize the relationbetween various higher-level and lower-level categories and sciences.Indeed, it is in just such terms that nonreductive ethical naturalism issometimes formulated.7 But then Moore's nonnaturalism may be no lessplausible than nonreductive ethical naturalism.

IV. ETHICAL NATURALISM AND DIRECT REFERENCE

The second main response to the OQA is to reject it altogether. Even ifnonnaturalism is a coherent—perhaps even plausible—position, Moore's

6 G. E. Moore, "Reply to My Critics," in Paul Arthur Schilpp, ed., The Philosophy of G. E.Moore (La Salle, IL: Open Court, 1942), 588.

7 See, for example, Brink, Moral Realism and the Foundations of Ethics, chaps. 6-7, esp.pp. 156-67, 172-80, 193-97.

158 DAVID O. BRINK

argument for it is a bad one. In particular, the OQA is not a good argu-ment against the ethical naturalist who identifies moral properties withnatural properties. The OQA depends upon the semantic thesis, accordingto which moral predicates cannot be defined in naturalistic terms, and theaforementioned semantic test of properties, according to which termspick out the same property if and only if they are synonymous. Bothpremises are questionable.

We might question the semantic thesis. The OQA rightly recognizes theepistemic inequivalence of moral and natural predicates, but it is not clearthat epistemic inequivalence implies semantic inequivalence. As I said,the inference from epistemic to semantic inequivalence might be reason-able if a descriptional theory of meaning were true, because on this view,speakers could be expected to recognize whether they associate the samedescriptions with different words and, hence, whether those words aresynonymous. This view has some plausibility as an account of nominal ordictionary definition, but little plausibility as an account of real definition.A descriptional view of meaning will seem problematic if we combine itwith the traditional view according to which meaning determines refer-ence, for then the descriptional view implies that a term refers to all andonly those things that satisfy descriptions conventionally associated withthe term. This view of reference is problematic for names and natural-kind terms for a number of reasons. First, it does not allow us to usenames and general terms to refer to individuals or properties about whomor which we could associate erroneous descriptions. Intuitively, it seemsthat speakers can use names and general terms to refer to things aboutwhich they associate erroneous descriptions; for example, we mightwrongly associate the description 'teacher of Plato' with the name 'Aris-totle'. To explain how it is that we might say falsely of Aristotle that hewas Plato's teacher, the name 'Aristotle' must not refer via the satisfactionof an associated description (which might include 'student of Plato'). Theproblem here for the descriptional theory is that it does not distinguishproperly between what speakers' words refer to and what speakers believe.

This problem manifests itself in another way: the descriptional theoryhas difficulty representing disagreement. Genuine disagreement requiresunivocal meaning or reference and incompatible beliefs about the natureof the things to which one's words refer. If the descriptional theory is true,then meaning just consists in the descriptions conventionally associatedwith terms. On this account, it becomes impossible to represent a familiarkind of disagreement. Whereas the descriptional theory can representdifferent speakers disagreeing about whether something in the worldsatisfies a given description associated with a given word or phrase, itcannot represent disagreement about which descriptions to associate witha given word or phrase.

Consider synchronic disagreement. At any one stage in an intellectualinquiry, there may well be a number of descriptions conventionally as-


sociated with the use of a general term. Suppose most speakers associatefeatures X, Y, and Z with general term 'G'. It ought to be possible for aheretical inquirer to express disagreement with the prevailing view. Aspeaker ought to be able to say that the very thing that most speakers use'G' to refer to is not X, Y, and Z, but rather is A, B, and C (where 'X, Y,and Z' and 'A, B, and C have different extensions). However, this is ruledout by the traditional descriptional theory, for on this view the meaningand reference of 'G' is given by the description—X, Y, and Z—that isconventionally associated with 'G'. The heretic's claim would thus beanalytically false. But certainly not all heretical claims are false—muchless analytically false—as the progressive nature of various inquiriesshows us.8

This leads naturally to the case of diachronic disagreement. As I men-tioned, at any one stage in an intellectual inquiry, there may be a set ofdescriptions conventionally associated with the use of a general term.As inquiry progresses, the descriptions associated with these terms arelikely to change. We would normally like to say that there is diachronicdisagreement and progress, but both judgments presuppose univocalmeaning and reference. According to the traditional descriptional theory,however, there is no such continuity. Under this theory, the two lin-guistic communities in the example above associate nonequivalent de-scriptions with their terms, and hence mean and refer to different thingswhen using them. They no more disagree than do two interlocutorswhen one says "The bank [- financial institution] is a good place foryour money" and the other says "The bank [= side of the river] is nota good place for my money." On this view, there can only be dia-chronic intellectual change; there can be no genuine intellectual disagree-ment, continuity, or progress.9

These problems arise from the descriptional theory because it allowsreference to be determined by senses or descriptions. A natural alterna-tive to this kind of mediated reference is a theory of direct reference that doesnot make reference depend upon speakers' associated descriptions orbeliefs. One account of direct reference, the causal theory of reference, grewout of suggestions made by Keith Donnellan and was developed by Saul

8 On an individualistic version of the descriptional theory, 'G' means different things inthe mouths of the orthodox and the heterodox. On this view, the heterodox is not makinganalytically false claims, but neither is he disagreeing with the orthodox—he has simplychanged the subject. Thus, this version of the descriptional theory is also incapable ofallowing for genuine disagreement.

9 These arguments are framed in terms of the traditional version of the descriptionaltheory and do not directly address other versions of the descriptional theory, like Searle's,that make the meaning of a term consist, not in a single (though perhaps complex) descrip-tion, but rather in a cluster or family of descriptions, some but not all of which need to besatisfied in order for the term to refer. See John Searle, "Proper Names," Mind 67, no. 266(1958): 166-73. However, like Kripke, I think that the arguments in the text apply, with onlysmall modifications, to the cluster theories of descriptions. See Saul Kripke, Naming andNecessity (Cambridge, MA: Harvard University Press, 1980), 74-77.

160 DAVID O. BRINK

Kripke and Hilary Putnam.10 According to the causal theory, names andnatural-kind terms refer to things in the world via complex causal histo-ries. On this view, language-users introduce words (e.g., names and gen-eral terms) to pick out interesting features of their environment. In theidealized case, a group of speakers introduces a term such that what theirterm refers to is that bit of the environment (e.g., that object, property, orrelation) that explains what they found interesting and intended to pickout. Subsequent speakers borrow this term with the intention of referringto the same thing; their use of the term inherits reference to the samefeatures of the environment via a causal-historical chain extending fromtheir use of the term, through the original "dubbing ceremony," to therelevant aspects of the environment. For example, the term 'water' wasintroduced to pick out the colorless, odorless stuff found in lakes, rivers,etc. that is suitable for drinking, bathing, supporting life, etc. This liquidis actually made of H2O, and it is this chemical composition that allowsit to serve these various functional roles. According to the causal theory,the reference of the term 'water' is determined by this causal-historicalchain; past, subsequent, and present use of 'water' refers to H2O, even asspeakers' beliefs about water have been very different. Because peoplehave not always realized that the chemical composition of water is H2Oor even that water has a chemical composition, the causal theory allowsus to explain how speakers can use terms meaningfully while being quiteignorant, perhaps mistaken, about their extension, and how speakers canspeak about the same thing while disagreeing significantly in their beliefsabout the subject matter. On this view, determining what the term 'water'refers to involves reliance on scientific theorizing about the chemicalstructure that explains the liquid's properties. We appeal or defer to ex-perts, not because their beliefs determine what our terms refer to, butbecause those beliefs provide us with the best available evidence aboutthe real nature of the referents of our terms.

Friends of the causal theory who want to preserve the traditional con-nection between meaning and reference will accept a referential view aboutmeaning, according to which the meaning of names and natural-kindterms just is (or at least involves) their reference.11 On this view, ascrip-tions of meaning to natural-kind terms involve explicit or implicit theo-retical commitments, and speakers cannot be assumed to be authoritativeabout the meaning of their words. This referential view about meaning

10 See Keith Donnellan, "Reference and Definite Descriptions," reprinted in Stephen P.Schwartz, ed., Naming, Necessity, and Natural Kinds (Ithaca, NY: Cornell University Press,1977); Kripke, Naming and Necessity; and Hilary Putnam, "The Meaning of 'Meaning',"reprinted in his Mind, Language, and Reality, vol. 2 of his Philosophical Papers (New York:Cambridge University Press, 1975).

11 In "The Meaning of 'Meaning'," Putnam conceives of meaning as a "vector" consistingof (a) syntactic markers; (b) semantic markers; (c) stereotypes, which do not determineextension; and (d) modal extension. Thus, the meaning vector for 'water' would be (a) massnoun; (b) natural kind, liquid; (c) transparent, odorless, colorless, potable; and (d) H2O.


leads to an account of real, as opposed to nominal, definition. The mean-ing of our terms is fixed by the nature of the world and can be knownonly through substantive investigation. On such a view of meaning, termsmight have the same meaning yet differ in connotation or cognitive sig-nificance. If so, there is no reason to suppose that epistemic equivalenceis a condition of semantic equivalence. 'M' and 'N' might have the samemeaning or real definition even if 'Are N-things M?' is an open question.In this way, a referential account of meaning might lead us to question thesemantic thesis.

Alternatively, we might question the semantic test of properties. If weaccept a descriptional theory of meaning, or any other view of meaningon which competent speakers must be authoritative about the meaning oftheir words, then it is harder to separate semantic and epistemic equiv-alence, but it becomes very implausible that semantic or epistemic equiv-alence is a test of property-identity. We entertain and accept many syntheticidentity claims whose truth is not ensured by the meanings of the wordsin which those claims are expressed. In doing so, we suppose that 'F' and'G' might express the same property without being synonymous. Some-times these claims involve intertheoretic identification; in these cases, wesuppose that general terms from different disciplines express the sameproperty, as when we suppose that water = H2O or that heat = meankinetic molecular energy. Terms and categories from different disciplinestypically have different epistemic and semantic values, which explainswhy intertheoretic identifications were discoveries and might still be newsto some competent speakers. It follows from this that neither epistemicnor semantic equivalence is required for property-identity. These andother claims of metaphysical equivalence are synthetic claims to be as-sessed by the same dialectical standards under which we assess all theo-retical claims.

The implications for the OQA of these general worries about the se-mantic thesis and the semantic test of properties ought to be pretty straight-forward. Ethical naturalism implies that moral properties are naturalproperties—that is, properties that can be picked out by means of pred-icates from the natural and social sciences (broadly understood). On onereading, this is to be understood as a constitution claim rather than anidentity claim. It is not clear that the OQA is meant to challenge this sortof nonreductive ethical naturalism.12 The main target of the OQA is theclaim that moral and natural properties are identical, that is, that moralpredicates and nonmoral predicates from the natural and social sciencescan express the same property. The OQA appeals to the semantic thesisand the semantic test of properties to undermine this version of ethical

12 However, the intuitionist's reasons for accepting the semantic test of properties—thatall necessity is analytic—may also undermine the claim that moral properties, though notidentical with natural properties, nonetheless strongly supervene on, and are constituted by,natural properties.

162 DAVID O. BRINK

naturalism. However, the ethical naturalist can plausibly reject either thesemantic thesis or the semantic test of properties.

The ethical naturalist might be a referentialist about the meaning ofmoral terms. If so, she will think that the meaning of a moral predicate is(or involves) the property it picks out. On this view, speakers cannot beassumed to be authoritative about the meaning of moral terms; what theymean is a matter of substantive moral theory that we articulate anddefend by familiar dialectical methods. The (reductive) ethical naturalistsupposes that moral terms can be given this sort of real—as opposed tonominal—definition, in naturalistic terms. The epistemic inequivalence ofmoral definienda and naturalistic definiens is no obstacle to the sort ofreal definition to which this sort of ethical naturalist might be committed.For instance, an ethical naturalist might put forward hedonistic utilitar-ianism as a real definition of moral duty, according to which an agent'sduty is to perform the action, among the alternatives available, that wouldmaximize overall pleasure. This real definition is not undermined by theepistemic inequivalence of the phrases 'maximizes pleasure' and 'is one'sduty'. Rather, it is to be assessed (and I would suppose ultimately re-jected) dialectically in terms of its fit with our other moral beliefs.

Alternatively, the ethical naturalist might accept, at least for the sake ofargument, the sort of descriptional theory of meaning on which the se-mantic thesis seems to rest. She will then infer semantic inequivalencefrom epistemic inequivalence and so accept the semantic thesis, but thenshe can and should reject the semantic test of properties. Moral andnonmoral terms can pick out the same property even if those terms arenot synonymous. These will be synthetic property-identity claims, whichare to be justified by a dialectical moral inquiry. For example, hedonisticutilitarianism is not false just because 'is one's duty' and 'maximizespleasure' are not synonymous. This particular property-identity claim isa synthetic claim, to be assessed on substantive moral grounds.

Hence, we might say, speaking somewhat loosely, that the probativeforce of the OQA is confined to analytical ethical naturalism and does notextend to synthetic or metaphysical ethical naturalism.13 Insofar as syn-thetic or metaphysical ethical naturalism appeals to the theory of directreference, it promises to explain how we can use moral terms meaning-fully and to refer even when speakers have erroneous moral beliefs, and

13 It is common to distinguish analytic truths, as those statements made true by virtue ofthe meanings of the words in which the statement is expressed, and synthetic truths, as thosestatements whose truth depends upon the way the world is and not simply the meaningsof the words in which the statement is expressed. If analytic truths are simply statementsmade true by virtue of the meanings of the words in which the statement is expressed, thenreal definitions might express analytic truths, and so at least one way of defending ethicalnaturalism could be understood as a defense of analytical naturalism. It is common, how-ever, to think that analytic truths ought to be comparatively obvious truths about whichcompetent speakers ought to be authoritative. Insofar as this is part of analyticity, even thereferentialist about meaning ought to reject analytical naturalism.


how speakers can use moral language univocally and succeed in dis-agreeing even when they have very different moral beliefs. This is be-cause on this view a moral appraiser's words refer to features of people,actions, and institutions via her participation in an extended causal-historical chain linking past and present speakers of her language tomoral features of the world—not via speakers' associated descriptions orbeliefs. Moral appraisers can thus use moral language, such as "right"and "wrong," meaningfully and to refer, provided that their use is part ofthis causal-historical chain—even if they have few beliefs or largely falsebeliefs about which actions are right and which are wrong, or about whatmakes right acts right and wrong ones wrong. Moreover, participation inthe same causal-historical chain will ensure univocal meaning or refer-ence for the use of common moral terminology, such as "right" and"wrong," by different speakers at a time and by speakers at differenttimes, regardless of differences in their moral beliefs and their criteria forapplying these terms. This allows us to represent how both synchronicand diachronic disagreement are possible between speakers with verydifferent moral codes. It also allows us to represent changes in moralattitude that have other relevant progressive aspects—for instance, wheninformed and reflective changes in moral attitude are unidirectional—asforms of moral progress.14

V. DOUBTS ABOUT REALISM, NATURALISM, AND DIRECT REFERENCE

I have come across two main kinds of worry about the moral realist'suse of the theory of direct reference to defend synthetic ethical natural-ism. These worries share a common conception of how the ethical natu-ralist makes use of the theory of direct reference, consideration ot whichsuggests an alternative conception that promises to resolve both worries.I will explain the two worries first and then discuss possible resolutions.

I begin with the worry that is easiest to state. Though it has echoes inpublished discussions,15 I have come across it most clearly in conversa-tion.16 Stated baldly, the worry is that substantive moral reasoning andtheorizing become obsolete if we let the ethical naturalist respond to theOQA by appeal to the causal theory of reference. The causal theory ofreference appears to make the reference of speakers' use of moral pred-icates a matter of empirical—specifically, historical—fact insofar as sub-sequent use of these predicates inherits reference, via speakers' intentionsto use the predicates with the same reference as earlier speakers, from theoriginal use of those predicates to name interesting features of the envi-

14 See Brink, Moral Realism and the Foundations of Ethics, 204-9; cf. Michael Slote, "TheRationality of Aesthetic Value Judgments," Journal of Philosophy 68, no. 22 (1971): 821-39.

15 See Eric H. Gampel, "A Defense of the Autonomy of Ethics: Why Value Is Not LikeWater," Canadian Journal of Philosophy 26, no. 2 (1996): 191-209.

16 Especially with John G. Bennett.

164 DAVID O. BRINK

ronment of those original speakers. If so, it would seem that the causaltheory implies that disputes about the meaning or reference of moralterms ought to be resolved not, as one would think, by moral reasoning,but by means of a historical inquiry about which features of actions,people, and institutions moral appraisers intended to pick out when thosemoral terms were introduced. If so, appeal to the causal theory of refer-ence is not a good response to the OQA.

The first worry, then, assumes that determination of which propertiescausally regulate, in the relevant way, speakers' use of moral terms is apurely historical matter. The second worry concerns the possibility thatsome speakers have their use of moral terms causally regulated by dif-ferent properties of the people, actions, and institutions that their moraljudgments concern. This worry is set out most elaborately by Mark Tim-mons and Terence Horgan in a series of articles.17 The upshot of theirargument is that they see a tension between moral realism and ethicalnaturalism insofar as the defense of ethical naturalism against the OQArequires appeal to theories of direct reference, such as the causal theory,which they see as incompatible with moral realism. On their view, youcan be a moral realist or an ethical naturalist, but not both.

Timmons and Horgan begin by noting that recent defenses of moralrealism have tended to eschew the apparent mysteries of intuitionism infavor of some form of ethical naturalism, and have defended ethicalnaturalism against Moore's OQA by appealing to claims about meaningand reference that are part of the direct-reference tradition. They focus onRichard Boyd's account of the causal theory of reference, according towhich

Roughly, and for nondegenerate cases, a term t refers to a kind (prop-erty, relation, etc.) k just in case there exist causal mechanisms whosetendency is to bring it about, over time, that what is predicated of theterm t will be approximately true of /c.... Such mechanisms willtypically include the existence of procedures which are approxi-mately accurate for recognizing members or instances of k (at leastfor easy cases) and which relevantly govern the use of t, the socialtransmission of certain relevantly approximately true beliefs regard-ing k, formulated as claims about t ... , a pattern of deference toexperts on k with respect to the use of t, etc. . . . When relations of thissort obtain, we may think of the properties of k as regulating the useof t (via such causal relations), and we may think of what is said

17 See Terence Horgan and Mark Timmons, "New Wave Moral Realism Meets Moral TwinEarth," Journal of Philosophical Research 16 (1990-91): 447-65; Terence Horgan and MarkTimmons, "Troubles for New Wave Moral Semantics: The Open Question Argument Re-vived," Philosophical Papers 21, no. 3 (1992): 153-75; and Terence Horgan and Mark Timmons,"Troubles on Moral Twin Earth: Moral Queerness Revived," Synthese 92, no. 2 (1992): 221-60. The first two of these articles are most directly relevant to my present concerns.


using t as providing us with socially coordinated epistemic access to k;t refers to k (in nondegenerate cases) just in case the socially coordi-nated use of t provides significant epistemic access to k, and not toother kinds (properties, etc.).18

Timmons and Horgan link a naturalistic approach to ethics, such as myown, with Boyd's version of the causal theory, thereby associating ethicalnaturalism with the following causal semantic thesis:

Each moral term t rigidly designates the natural property N thatuniquely causally regulates the use of t by humans.19

They then ask us to consider a Twin Earth story of the sort Putnam usedto motivate his referential account of meaning.20 In Putnam's originalversion, Earth and Twin Earth are otherwise indistinguishable planets,but the stuff on Earth that causally regulates the use of 'water' is H2Owhereas the stuff on Twin Earth that causally regulates the use of 'water'is XYZ (where XYZ ¥= H2O). Putnam used Twin Earth scenarios to arguethat meaning and reference are a function of causal interaction with thespeaker's environment, and concluded that natural-kind terms, such as'water', would mean different things in different environments (on Earthand Twin Earth, for example). Timmons and Horgan ask us to imaginethat Earth and Moral Twin Earth are otherwise indistinguishable, but thatthe same moral term 'M' is causally regulated by different natural prop-erties on the two planets. We are to imagine that people, actions, andinstitutions on the two planets are qualitatively indistinguishable exceptfor the standards for moral assessment that appraisers on each planetemploy. For simplicity, they ask us to imagine that consequentialist con-siderations uniquely causally regulate use of moral language on Earth,whereas deontological considerations uniquely regulate the use of morallanguage on Moral Twin Earth.21

18 Boyd, "How to Be a Moral Realist," 195. There is a puzzle here insofar as Boyd appearsto commit the causal theory to successful reference, for I take fallibilism and the possibilityof reference failure to be a defining feature of realism. However, perhaps Boyd would regardreference failure as a "degenerate" case.

19 Horgan and Timmons, "Moral Realism Meets Moral Twin Earth," 455.20 See Putnam, "The Meaning of 'Meaning'."21 However, there is an important difference between Putnam's Twin Earth and Moral

Twin Earth. The difference between Earth and Twin Earth appears to be merely compo-sitional—the stuff on Earth is H2O, whereas the stuff on Twin Earth is XYZ—and not to havewider functional significance. However, the difference between Earth and Moral Twin Earthappears not to be merely compositional. Presumably, different people, actions, and institu-tions will satisfy consequentialist and deontological standards. If people have the samecommitments to morality on Earth and Moral Twin Earth, the differing standards will causeeach planet's people to assess people, actions, and institutions differently; over the long run,this should affect the course of individual and social histories on Earth and Moral TwinEarth. Though the members of both planetary pairs—Earth and Twin Earth and Earth and

166 DAVID O. BRINK

In these circumstances, would Earthlings and Moral Twin Earthlingsuse moral language with a common meaning and reference—and hencehave disagreements in moral beliefs—or would moral language on Earthand Moral Twin Earth have different meaning and reference? The moralthat Putnam draws from his Twin Earth scenarios would seem to suggestthat the meaning and reference of moral language are different on Earthand Moral Twin Earth. This is also what the causal semantic thesis wouldseem to imply, inasmuch as the same 'M' is causally regulated by differentN-properties on Earth and Moral Twin Earth. However, this would ap-pear to imply interplanetary semantic relativism. Timmons and Horganfind this consequence unattractive, presumably because they want tointerpret Earthlings and Moral Twin Earthlings as being involved in agenuine moral disagreement, which presupposes univocal meaning andreference. Indeed, they think that the moral realist must deny this sort ofinterplanetary relativism.

This claim may seem a bit puzzling. Whereas we often contrast realistand relativist claims, it may seem peculiar to require the moral realist todeny this sort of semantic relativism. After all, Putnam appeals to a se-mantic relativist interpretation of his Twin Earth story as part of an argu-ment for direct reference, and direct reference is supposed to be part of—orat least compatible with—metaphysical realism. Putnam sees no tensionbetween metaphysical realism and his brand of semantic relativism. Whyshould we see a tension between moral realism and this sort of semanticrelativism? The answer, I think, lies in the kind of semantic relativism withwhich we are dealing. One reason metaphysical realism seems compatiblewith the kind of semantic relativism implicit in Putnam's version of directreference is that both realism and direct reference want to distinguish be-tween speakers' beliefs and the subject matter of those beliefs; this is whatallows the realist to distinguish between how things are and how they ap-pear to inquirers. Putnam's semantic relativism respects this condition; themeaning of the term 'water' is different on Earth and Twin Earth, but thisis because of differences in physical composition on Earth and Twin Earth,not because of different beliefs of Earthlings and Twin Earthlings. How-ever, the semantic relativism in Timmons and Horgan's case would makemultiple reference a consequence of different moral beliefs on Earth andMoral Twin Earth, because speakers have different moral beliefs just in-sofar as they regulate their use of moral language by different N-properties.If so, it appears that the semantic relativist interpretation of Moral TwinEarth is inconsistent with moral realism.

Moral Twin Earth—are, as I said, otherwise indistinguishable, this caveat includes manymore differences in the second pair than in the first. As it seemed important to Putnam'soriginal arguments that differences between Earth and Twin Earth be minimized, the moreextensive differences between Earth and Moral Twin Earth may complicate Timmons andHorgan's argument. In what follows, I ignore these complications, if only for the sake ofargument.


Moreover, interplanetary relativism might seem problematic for themoral realist if it is just the first step toward intraplanetary relativism.Why shouldn't the use of moral language by different cultures and bydifferent groups within a given culture here on Earth be causally regu-lated by different N-properties? Indeed, isn't this just what moral dis-agreement involves—that appraisers do have their use of moral languagecausally regulated by different N-properties? But then it looks as if allmoral disagreements would involve multiple meaning and reference.Hence, the causal semantic theory to which the ethical naturalist is com-mitted implies a rampant intraplanetary relativism, which it would behard to square with moral realism.

It would be ironic if direct reference commits us to moral relativism,inasmuch as it was supposed to be a virtue of the theory of direct refer-ence that it allowed us to represent common meaning and reference amonginquirers with very different beliefs about a given subject matter. Be thatas it may, the worry is that the Moral Twin Earth scenario exposes atension between ethical naturalism and moral realism; the defense ofethical naturalism involves semantic commitments to direct reference thatare inconsistent with moral realism. If so, we can be realists or naturalistsabout ethics, but not both; in particular, we cannot defend moral realismagainst Moore's OQA, as some (including me) have recently supposed,by understanding ethical naturalism in terms of the theory of directreference.

We can now see how the first and second worries about the moralrealist's use of the theory of direct reference to defend synthetic ethicalnaturalism depend on a common conception of the theory of direct ref-erence. They both understand the theory of direct reference in terms ofthe causal semantic thesis that moral terms refer to whatever N-propertiescausally regulate their use. If so, ascertaining the meaning and referenceof moral terms would appear to be a historical inquiry rather than a moralinquiry. Moreover, we should not expect appraisers' use of moral termsalways or even usually to be causally regulated by the same N-properties—the causal semantic thesis thus appears to support relativism, not realism.

VI. CAUSAL REGULATION

It would be premature to accept these two worries about the moralrealist's use of direct reference to defend ethical naturalism, for the com-mon conception of direct reference underlying these two worries de-serves rethinking. On Boyd's version of the causal theory, which Timmonsand Horgan associate with ethical naturalism, a speaker's term refers tothose features of the world that causally regulate her use of the term,including her use of it to communicate with others. One important issuefor this view concerns the proper understanding of how features of the

168 DAVID O. BRINK

world causally regulate the use of moral-kind terms. Both of the worriesdiscussed in the previous section are most plausible when regulation isunderstood in a narrowly extensional way. This extensional interpreta-tion understands regulation in terms of the features of the world thatcausally regulate people's actual use of moral terms. Determining whichN-properties regulate a speaker's actual use of moral-kind terms lookslike a historical inquiry rather than a moral inquiry, and the existence ofsignificant moral disagreement suggests that actual usage is regulated bydifferent N-properties in different speakers. However, causal regulationcan and should be understood in counter/actual terms. On this view, termsrefer to properties that regulate not just actual usage, but also counter-factual or hypothetical usage—in particular, the way speakers wouldapply terms upon due reflection in imagined situations and thoughtexperiments.

Any interpretation of regulation must also understand it in a way thatis sensitive to various demands of consistency. To apply moral terms invirtue of N-properties is just to have a set of beliefs, however implicitly,about which N-properties are morally relevant. Presumably, even actualusage does not track any one set of morally relevant properties consis-tently. We therefore need an account of error. Any ascription of local errorwill ascribe to the speaker principles, however implicit, about whichN-properties are morally relevant. In this way, we can see local errors asperformance errors against the background of an underlying competence.We might decide which principle to ascribe to a speaker by seeing whichone minimizes the ascription of error to her. As we seek consistency in theapplication of moral terms across imagined as well as actual situations,there is greater room for inconsistency and so a greater need to under-stand regulation in a way that will allow for principled correction to aspeaker's usage. Demands of consistency arise not only from patterns inactual and counterfactual use of a given term, but also from the use ofother terms, both moral and nonmoral. This is because the principles thatimplicitly regulate a speaker's use of any one moral term interact with anumber of other moral and nonmoral views that she holds.

This leads to a picture of regulation that is dialectical. Our actual use ofmoral predicates is imperfectly guided by our (perhaps implicit) accep-tance of moral principles that identify morally relevant factors. We iden-tify principles by looking for patterns in our actual and counterfactualjudgments that employ those predicates; we test these principles by draw-ing out their implications for real or imagined cases and comparing theseimplications with our own existing or reflective moral assessments ofthose cases. If a principle has counterintuitive implications, this countsagainst it. If the counterintuitive implications of the principle are fairlycommon, this is reason to abandon or modify the principle. However, ifa counterintuitive implication is isolated and the principle explains ourviews—especially our common moral views—better than alternative prin-


ciples, then this is reason to revise the particular judgment or precept thatconflicts with the principle. Ideally, we make trade-offs among our prin-ciples, considered moral judgments, and other views in response to con-flicts, making adjustments here at one point and there at another, ascoherence seems to require, until our ethical views are in dialecticalequilibrium.22

On this view, a natural property N causally regulates a speaker's use ofmoral term 'M' just in case his use of 'M' would be dependent on hisbelief that something is N, were his beliefs in dialectical equilibrium. Bothworries about the use of the theory of direct reference to defend syntheticethical naturalism—the worries that it makes moral reasoning obsoleteand that it implies a form of semantic relativism that is incompatible withmoral realism—are less compelling when we shift from the extensional tothe counterfactual or dialectical understanding of causal regulation.

Determining the meaning and reference of moral terms might look likehistorical reasoning rather than moral reasoning if we are interested in theproperties that extensionally regulate moral terms. Insofar as the relevantnotion of regulation is counterfactual and dialectical, though, the processof ascertaining the meaning or reference of one's moral terms will involvejust the sort of thought experiments and analogical reasoning that ischaracteristic of moral reasoning.

There is enough difference in people's actual beliefs about morallyrelevant factors to make it plausible, at least in the case of some speakers,that different properties extensionally regulate their use of the same moralterms. However, the problem of multiple reference and relativism is lesscompelling on the counterfactual conception of regulation. The issue hereis really just the familiar one about whether extant moral disagreementundermines prospects for dialectical convergence. Although dialecticalreasoning takes people's pretheoretical moral beliefs as input, these be-liefs will be revised in the process of reaching dialectical equilibrium.Indeed, there is no guarantee that any of the beliefs with which onebegins the dialectical process will be preserved unmodified. A properdialectical equilibrium should be broad, representing a dialectical accom-modation not simply among our moral beliefs, but among our moralbeliefs and various philosophical and empirical beliefs.23 Given the de-pendence of many of our moral beliefs on complex empirical and philo-sophical issues, there is every reason to expect any broad dialectical fitbetween various (moral and nonmoral) beliefs to be revisionary. Becausea broad dialectical fit is an intellectual aim that we can at best hope to

22 This account of dialectical methods and dialectical equilibrium is, of course, similar toRawls's account of reflective equilibrium. See John Rawls, A Theory of Justice (Cambridge,MA: Harvard University Press, 1971), 19-21, 46-51, 578-81. See also David O. Brink, "Com-mon Sense and First Principles in Sidgwick's Methods," Social Philosophy and Policy 11, no. 1(1994): 179-201.

23 See Brink, "Common Sense and First Principles in Sidgwick's Methods," 184-87, 200.

170 DAVID O. BRINK

approximate, the existence of prereflective and even reflective moral dis-agreement is no sign that moral disagreement is in principle unresolvable.Indeed, the claim that moral disagreements are in principle unresolvableby dialectical methods is just one claim about what the results of a sys-tematic dialectical inquiry among different interlocutors would be, andenjoys no privileged a priori position in relation to its nonskeptical com-petitors. This means that whereas there is in the nature of things noguarantee of common reference, this provides no reason to doubt that ourterms do have common reference or that dialectical methods are our routeto moral knowledge.24

If this is right, the two aforementioned worries about the moral realist'suse of the theory of direct reference to defend synthetic ethical naturalismrest on an inadequate understanding of the way in which features of theworld might regulate the use of language. If the moral realist understandsthe meaning and reference of moral-kind terms, as Boyd does, in terms ofcausal regulation, then regulation should be understood in counterfactualand dialectical terms. When regulation is understood in this way, theworries about the obsolescence of moral inquiry and about multiple ref-erence become less compelling.

VII. REFERENTIAL INTENTIONS AND THE MORAL POINT OF VIEW

Perhaps the dialectical understanding of causal regulation gives us adecent reply to the two worries about the moral realist's use of the theoryof direct reference to defend synthetic ethical naturalism, if we assume, asBoyd and Timmons and Horgan do, that the ethical naturalist must em-brace an account of direct reference that makes the meaning and referenceof moral-kind terms a function of the properties that causally regulatetheir use. I am not sure, however, that we should understand directreference this way. We get a different semantic picture when we recognizeand emphasize the role, within the theory of direct reference, of speakers'intentions in determining meaning and reference.25 After explaining theseclaims, I will briefly contrast this conception of direct reference with onethat invokes the causal regulation thesis.

One way in which referential intentions function in the theory of directreference is to provide continuity of reference among a community ofinquirers. It is sometimes said that participation in the causal-historicalchain linking words to the world allows later speakers to "borrow" ref-erence from earlier ones. What allows later speakers to do this—whatmakes them a link in the same chain—is their use of their predecessors'

24 For a fuller discussion of the resources within a dialectical method for dealing withm o r a l d i s a g r e e m e n t , see Brink, Moral Realism and the Foundations of Ethics, 197-209.

25 Donnellan, Kripke, and Putnam all recognize the importance of referential intentions infixing reference; their importance is also stressed in Alan Sidelle, Necessity, Essence, andIndividuation: A Defense of Conventionalism (Ithaca, NY: Cornell University Press, 1989).


words with the intention of talking about the same thing that they did. Itis these shared referential intentions among synchronic and diachroniccommunities of inquirers that establish common meaning and reference.If so, shared referential intentions among communities of moral inquiryshould also establish common meaning and reference. The intentions ofmoral inquirers to say and think things about the same features of people,actions, and institutions as each other will block ascription of differentmeanings or reference to fellow participants in a common moral inquiry,and will lead us to interpret such differences as there are as differences inbelief about the extension of moral terms.

This may block the slide from interplanetary relativism to intraplane-tary relativism, precisely because there is, at least typically, intellectualand linguistic interaction among moral inquirers here on Earth. However,where there is no such interaction—by hypothesis, there is none in theinterplanetary case, and there might fail to be any in some intraplanetarycases—there apparently can be no such shared referential intentions. Inthese cases, we cannot appeal to such intentions to secure commonreference.

Does this commit us to a more limited relativism? If so, we could askwhether this form of relativism is inconsistent with moral realism. I thinkthat we can avoid answering this question, though, because the role ofreferential intentions within the theory of direct reference provides amore thoroughgoing response to the relativist worry. The relativist positstwo or more lines of moral inquiry in which the same words have dif-ferent meaning and reference. However, if these lines of inquiry aredistinct—that is, they do not interact—and are counterfactually and dia-lectically regulated by different properties, what justifies us in interpret-ing each line of inquiry as a line of moral inquiry? It is true that bothcommunities of inquirers use the same words—'right', 'wrong', 'virtu-ous', 'vicious', etc. —but these might be mere homonyms. If we are par-ticipants in one community, why should we assume that assertions usingthese words, if made by participants in a different community, should beinterpreted as moral assertions?

One answer requires understanding the role of speakers' intentions infixing the reference of a term at the time of its introduction. Consideragain the nonmoral case. What makes it the case that the introduction ofa term picks out one particular feature of the speaker's environmentrather than another? It is the intentions of the speaker. To borrow from anexample used by Donnellan, I use the description "the man holding themartini" to pick person A out of the foreground of a crowd at a party, ofwhom I correctly say something else, for instance, that he is wearing anice suit.26 I can successfully pick out A even if we discover that his glasscontains water rather than a martini. What makes my description pick out

26 Donnellan, "Reference and Definite Descriptions," 48.

172 DAVID O. BRINK

A, rather than nobody or B, who was hidden from view but was drinkinga martini, is my intention, which you recognize, to single out the fellowin the foreground who is drinking from a martini glass and wearing asuit. Similarly, those who introduced the term 'water' intended to refer tothe structure, whatever it is, that explains the perceptible and functionalfeatures of the colorless, odorless stuff—found in lakes, rivers, etc.—thatis suitable for drinking, bathing, and supporting life. It is this intentionthat fixes the reference of 'water'. As it turns out, it is the chemicalmicrostructure H2O that answers this explanatory description. Moreover,the role of referential intentions explains why ascertaining the meaningand reference of 'water' is a scientific inquiry, not a historical inquiry: themeaning and reference of 'water' depends upon the answer to the scien-tific or explanatory question about what features of the colorless, odorlessliquid explain its observable and functional properties.27

To apply this account to the moral case, we need some parallel descrip-tive specification of the referential intentions of moral inquirers that wouldjustify us in interpreting a community of inquirers as engaged in moralinquiry. This requires a descriptive formulation of the moral point ofview, but it must be a description that is sufficiently abstract, so that awide variety of views (some quite unorthodox) might be thought to sat-isfy this description. Moreover, what best satisfies this description mustbe a matter of substantive moral theory.28 We might put these points interms of the useful distinction between concepts and conceptions.29 Often,we want to distinguish between an abstract concept and different sub-stantive conceptions of that concept. For example, we distinguish be-tween the concept of distributive justice, on the one hand, and utilitarian,egalitarian, and libertarian conceptions of distributive justice, on the otherhand. To explain how different conceptions are all conceptions of thesame concept, we need some abstract characterization of a given subjectmatter that allows us to see different conceptions as rival accounts of thatsubject matter. In this instance, we might explain the concept of distrib-

27 Different speakers, or even the same speakers in different contexts, can employ differ-ent intentions with respect to the same word, though to do so they cannot have the intentionto refer to whatever the other speakers are talking about. For instance, some speakers mightuse 'fish' to pick out a biological kind, in which case their use of 'fish' would not refer towhales. Other speakers or conversational contexts might not share this intention directly, orindirectly by way of an intention to refer to whatever the speakers with biological intentionswere talking about. For instance, those who live by the sea might use 'fish' to pick outmarine life, in which case their use of 'fish' would refer to whales. This kind of relativity ofmeaning or reference with respect to intention is perfectly compatible with the theory ofdirect reference.

28 These claims suggest that direct reference and descriptional theories need not be anti-thetical, provided that the descriptional theory gives the meaning of natural-kind terms insufficiently abstract descriptions, the satisfaction of which is a potentially controversialsubstantive matter.

29 Cf. H . L. A . H a r t , The Concept of Law, 2d ed . (Oxford: C l a r e n d o n Press , 1994), 159-60;R a w l s , A Theory of Justice, 5, 10; a n d R o n a l d D w o r k i n , Law's Empire ( C a m b r i d g e , M A :Harvard University Press, 1986), 70-72, 90.


utive justice as concerning the appropriate distribution of benefits andresponsibilities among members of a society; utilitarian, egalitarian, andlibertarian conceptions offer different substantive distributive principles.Analogously, to ascertain whether inquirers have referential intentionsthat would justify us in regarding them as moral inquirers, we needsome abstract account of the concept of morality common to differentconceptions of morality. Which conception is the best conception of theconcept of morality—and hence determines the extension of moral terms—will be a substantive theoretical matter to be determined by dialecticalmethods.30

We can entertain different descriptive specifications of the concept ofmorality. Present purposes do not require settling on one specification inpreference to all others; all we need is a plausible illustration of how sucha specification can function to establish common meaning or reference.One possibility is that we adopt the moral point of view toward anythingwhen we assess it as meriting praise or blame. However, this account ofthe concept of morality threatens to be too broad, forcing us to representall evaluation as moral evaluation. Another familiar idea appeals to thecontractualist idea that assuming the moral point of view involves as-sessing one's own conduct and that of others in terms of standards thatadmit of interpersonal justification.31 Hume articulates this idea in An En-quiry Concerning the Principles of Morals:

When a man denominates another his enemy, his rival, his antagonist,his adversary, he is understood to speak the language of self-love, andto express sentiments, peculiar to himself, and arising from his par-ticular circumstances and situation. But when he bestows on anyman the epithets of vicious or odious or depraved, he then speaksanother language, and expresses sentiments, in which he expects allhis audience are to concur with him. He must here, therefore, departfrom his private and particular situation, and must choose a point ofview, common to him with others; he must move some universalprinciple of the human frame, and touch a string to which all man-kind have an accord and symphony.32

30 There are dangers inherent in making any one description, however abstract, consti-tutive of the concept of morality, insofar as one ought to be quite liberal about what mightcount as a conception (however unorthodox) of morality. Another possible understanding ofthe concept of morality, which I will not explore further here, is that we count something asa moral code (however unorthodox) just in case its organizing principles make explanatorycontact with some of our pretheoretical beliefs about morality.

31 For a contemporary statement of contractualism, see T. M. Scanlon, "Utilitarianism andContractualism," in Amartya Sen and Bernard Williams, eds., Utilitarianism and Beyond(Cambridge: Cambridge University Press, 1982).

3 2 David H u m e , An Enquiry Concerning the Principles of Morals [1751], ed . P. H . N idd i t ch ,3rd ed. (Oxford: Clarendon Press, 1975), sec. 9, pt. 1.

174 DAVID O. BRINK

On this account of the concept of morality, what is distinctive of the moralpoint of view is that we assess people, actions, and institutions accordingto standards that others can and should accept.

This understanding of the concept of morality admits of many verydifferent conceptions. It admits of a Kantian interpretation in terms of theuniversal law formulation of the Categorical Imperative, as well as aHumean interpretation in terms of affective responses from the generalpoint of view. It does not settle important questions about how best tomodel interpersonal justification, such as what counts as reasonable agree-ment or rejection. Nor does it settle what principles or standards wouldbe adopted under the right conception of interpersonal justification—whether they would be egoist, utilitarian, deontological, or somethingquite different. Working out answers to these and other questions aboutinterpersonal justification involves forming and defending substantiveconceptions of morality.

Appealing to this account of the moral point of view in terms of inter-personal justification allows us to elaborate further the way a moral re-alist can understand and invoke the theory of direct reference. On thisview, we should understand perhaps all moral appraisers, and certainlythose who introduced moral categories and terms, as using those catego-ries and terms with the intention of picking out properties of people,actions, and institutions—whatever those properties are—that play animportant role in the interpersonal justification of people's characters,their actions, and their institutions. Subsequent appraisers inherit thisintention, if only because they use the same words as their predecessorsand have the intention of continuing an inquiry into the same subject.This picture allows us to explain the conditions under which it is reason-able to interpret the judgments of distinct communities of inquirers asmoral judgments. Even when appraisers from distinct communities uselanguage in ways that are counterfactually regulated by different prop-erties of their environments, we should interpret their language as morallanguage and the judgments that employ that language as moral judg-ments only if those judgments are based on standards that the appraisersendorse, if only implicitly, as interpersonally justifiable.

This account answers the puzzle for the relativist concerning the con-ditions under which we should interpret the judgments made on MoralTwin Earth as moral judgments. It does so, however, in a way that un-dermines the relativist interpretation of Moral Twin Earth. Recall thatrelativism appeared to be the commitment of the theory of direct refer-ence insofar as this theory was unable to identify a common meaning andreference about which appraisers from Earth and Moral Twin Earth helddifferent beliefs. But our account of the shared referential intention to pickout people, actions, and institutions that are interpersonally justifiable, invirtue of which the judgments of Earthlings and Moral Twin Earthlingsare both moral judgments, identifies just such a common meaning or


reference about which the two communities have a disagreement in be-lief. Their disagreement is one about which features of people, actions,and institutions make them interpersonally justifiable, with Earthlingsholding consequentialist views and Moral Twin Earthlings holding de-ontological views.33 Moral realism and the theory of direct reference,then, are compatible, and there is no reason to see a tension betweenethical naturalism and moral realism.

Moreover, this picture of the role of referential intentions within thetheory of direct reference also helps us answer worries about whether thetheory of direct reference implies the obsolescence of moral reasoning. Ifit is part of the referential intentions of moral appraisers to use morallanguage to pick out the properties—whatever they are—of people, ac-tions, and institutions that make them interpersonally justifiable, then theprocess of ascertaining the meaning or reference of moral terms just is aninquiry into the properties of people, actions, and institutions that makethem interpersonally justifiable. This inquiry can only be conducted bysubstantive moral reasoning.

Notice that this appeal to the role of referential intentions within thetheory of direct reference in order to defend a semantics for moral realismmakes no use of the causal regulation thesis, which Boyd embraces andTimmons and Horgan associate with ethical naturalism. On the presentview, reference is fixed by an original intention to adopt the moral pointof view—that is, to use moral language to pick out those properties,whatever they are, that make objects of assessment interpersonallyjustifiable—not by the properties that regulate, perhaps only implicitlyand counterfactually, speakers' use of moral terms. Whereas the accountof reference in terms of causal regulation must allow that the morallanguage of different speakers might be regulated by—and so refer to—different properties, the account of direct reference in terms of referentialintentions does not imply that there is multiple reference whenever lan-guage use is regulated by different properties. This brings out a way inwhich an account of reference in terms of causal regulation cannot dis-tinguish between a speaker's beliefs and her subject matter. Under thecausal regulation thesis, the properties that regulate, perhaps only dia-lectically, a speaker's use of moral terms just reflect the moral beliefs shedoes hold or perhaps would hold in dialectical equilibrium; this impliesthat the meaning and reference of a speaker's terms is fixed, perhaps ina very complicated way, by her beliefs. Hence, on this account, differencesin belief of the right sort between speakers do imply that they are talkingabout different things. Of course, the emergence and appeal of direct

33 Not surprisingly, this response to the relativist worry shares structural features withfamiliar responses to other relativist worries, in particular, the reply to relativist worriesabout moral disagreement that attempts to explain moral disagreement by finding shared orcommon principles that are applied in different empirical conditions, or at least in conjunc-tion with different empirical beliefs.

176 DAVID O. BRINK

reference was due in significant part to its promise to distinguish betweendifferences in belief and differences in subject matter and to explain howthere could be genuine disagreement of belief. Furthermore, it is thecommitment to fallibilism about our pretheoretical and reflective beliefs—characteristic of realism—that explains why this promise of direct refer-ence is especially appealing to realists. However, the promise of directreference, especially for the realist, is better realized by understandingreference in terms of referential intentions rather than in terms of causalregulation.34

VIII. CONCLUDING REMARKS

In conclusion, it may be worth noting that this account of moral se-mantics in terms of referential intentions to adopt the point of view ofinterpersonal justification is fiercely nonreductionist. To characterize themoral point of view in terms of interpersonal justification is to character-ize it in ineliminably normative terms. This makes it a substantive ques-tion, which I have not addressed here, whether moral terms do refer and,if so, which properties they pick out. Thus, I have not argued directly onbehalf of moral realism or ethical naturalism in this essay. In particular, Ihave not argued that there is a set of properties of people, actions, andinstitutions (presumably, in some sense, unique) in virtue of which char-acters, actions, and institutions are interpersonally justifiable, much lessthat those properties are natural properties. Those issues should not besettled by adopting a theory of reference. Rather, my primary aims havebeen to respond to doubts about the compatibility of moral realism anddirect reference, and to make moral semantics safe for moral realism. Themoral realist is free to use the theory of direct reference to defend ethicalnaturalism against the OQA.

Philosophy, University of California, San Diego

' If so, Boyd's semantic commitments turn out not to be robustly realistic.

realism, naturalism, and moral semantics

Documents