SOURCE: "Reductionism in Ethics and Science: A Contemporary Look at G. E. Moore's Open-Question Argument," in American Philosophical Quarterly, Vol. 25, No. 3, July, 1988, pp. 197-213.

The so-called "open question" argument is an argument against a general meta-ethical theory, or group of theories, known as ethical naturalism, according to which statements about what is morally right or wrong are reducible or equivalent in meaning to certain statements about empirical facts of nature. The standard objection to Moore's argument is that it is question-begging or circular. While this objection is currently accepted by ethicists, it has been recently objected that Moore's argument is in fact logically invalid, since it would yield inaccurate results if applied to established reductionist identities in science. Part I of the present essay examines, and rejects, the version of this objection given by Gilbert Harman. Part II examines, and again rejects, Putnam's version of the objection, which relies also on an analogy to science but goes on to apply Kripke's theory in philosophy of language. Part III attempts to show that the analogy between ethics and science, assumed by both Harman and Putnam, is incorrect. Finally, Part IV formulates and defends an interpretation of Moore's argument which is neither invalid nor circular. The conclusion is that the open-question argument provides relevant evidence against the usual forms of naturalist reductionism which Moore, as well as emotivists or moral non-cognitivists today, are concerned to refute.

I

Ethical naturalists have offered a variety of definitions to specify exactly which natural facts are being asserted by ethical statements. It has been suggested, for instance, that statements of the form "X is good" or "X is morally obligatory" just mean that "X is desired" or "is approved of (by the speaker, the speaker's society, or mankind generally, etc.). Such definitions have been criticized individually as having strange implications which render them implausible accounts of what ethical statements really do mean. The purpose of G. E. Moore's open-question argument, however, is to show that any naturalistic analysis of moral terms must be implausible, regardless of any more specific objections to the particular proposals. Such proposals will have the form:

"x is good or ought to be done" = df. "x (or doing x) has natural property P."

The insight of Moore's argument is that no matter what a naturalist takes P to be, one can recognize that a given x has P and yet still reasonably ask whether x is good or morally right—that is, this remains an "open question." Even if the answer is yes, or if being good/right is always correlated with having P, the fact that this question nevertheless makes sense shows that the correlated statements don't mean the same, as naturalists allege. Responding to Russell's proposal that "good" be defined as what we "desire to desire," Moore himself puts it this way in the locus classicus which premiers the open-questions argument:

. . . [W]hatever definition is offered, it may always be asked, with significance, of the complex so defined, whether it is itself good ... It may indeed be that what we desire to desire is always good; perhaps, even the converse may be true: but it is very doubtful that this is the case, and the mere fact that we understand what is meant by doubting it, shows clearly that we have two different notions before our minds.¹

Harman raises two objections to this argument. On the one hand, he suggests that the argument is circular or question-begging. This is the rather standard objection that if "good" or "right" really means what the naturalist says it does, then the question indicated above is not really "open," and to assert that it is amounts simply to assuming, rather than arguing, that naturalism is false.² We will return to this objection in Part IV. The second objection, which appears to be new, is that the open-question argument is logically invalid, since the same sort of argument could be used to "refute" established results of physical science:

More important, perhaps, is the fact that as it stands the open-question argument could be used on someone who was ignorant of the chemical composition of water to 'prove' to him that water is not H20. This person will agree that it is not an open question whether water is water but it is an open question, at least for him, whether water is H20. Since this argument would not show that water is not H20, the open-question argument cannot be used as it stands to show that for an act to be an act that ought to be done is not for it to have some natural characteristic C.³

This objection to the open-question argument (hereafter "OQ") is not, as it stands, very convincing. One's initial reaction to it is that Moore's argument against ethical naturalism is an attempt to refute a theory of meaning, and that a parallel argument in the water/H20 case could not yield the conclusion that water is not H20, as Harman indicates, but only the conclusion that the terms "water" and "H20" do not mean the same, or that, as Moore puts it in the quoted passage, there are "two different notions" involved here. On this view, the proposed application of Moore's argument would indicate only that it is an "open question" or a "significant question," whether water is H20, and Moore would surely say that it is such a question—and one which has been significantly answered by science. What Moore would presumably deny here is not the truth of "Water is H20" but rather the status of this truth as analytic, or as one of definition.

There are two supports for the above interpretation of OQ and for rejecting Harman's argument against it. The first is that, historically, the main tradition of ethical naturalism has in fact presented itself as a semantic theory which gives a reportive definition of the actual meanings of ethical terms in ordinary language.⁴ The second is that this at any rate is clearly the sort of theory that Moore himself, and others who have endorsed his argument, have intended to refute with OQ, and it is the only view consistent with Moore's own position in ethics. In holding the meta-ethical position that "good" is indefinable, Moore does not deny the truth of all statements about what activities or experiences are good; the claim is only that all such statements are "synthetic and never analytic," or can never give "the very meaning of the word."⁵ Part of Moore's position in meta-ethics is in fact the thesis that some such synthetic statements connecting "good" with empirical states of affairs are known to be true by a cognitive faculty of intuition. Specifically, as an act-utilitarian in normative ethics, Moore holds that right actions are those which maximize happiness, though again he rejects the attempts which he finds exemplified in Bentham and Mill, respectively, to treat these statements as definitions of what "right" and "good" mean.⁶ Moore's using OQ to refute such a definition is consistent, then, with his recognizing that statement as synthetically true. Analogously, in response to Harman's counterexample, Moore can deny that "water" and "H20" have the same meaning without denying that water is H20, since he clearly—and explicitly in the passage cited—is construing "meaning" in such a way that two expressions can have different meanings even though their extensions are synthetically the same.

It may appear that this defense of OQ against the objection of invalidity can be rebutted on the basis of recent challenges to the analytic-synthetic distinction itself, but this appearance is deceptive. Admittedly, Moore would have trouble establishing that the putative definitions of ethical naturalists are "synthetic and never analytic" if there is no such distinction; he cannot prove, by posing the "open question," that a statement like "X ought to be done if and only if x has natural property P" is not analytic, if there exist no criteria for analyticity to begin with. In the Preface to his text, Harman seems sympathetic to this view as he comments on what he takes to be its historical influence in ethics, suggesting that the arguments of Quine and others "undermined the distinction between meta-ethics and normative ethics by showing that there can be no real separation between questions of substance and questions of meaning."⁷ There are, however, several points in response to this line of criticism that may be made on behalf of Moore. First of all, it is of course not obvious that the arguments of Quine and others conclusively establish that there is no analytic-synthetic distinction. Even if one has no general counterarguments, such as those offered by Grice and Strawson, much less an articulated account of the criteria for distinguishing analytic from synthetic statements, one may still reasonably take the position—which has a certain G. E. Moore ring to it—that there is nonetheless an obvious difference in kind (i.e., of some kind) between statements like "All bachelors are unmarried" and those like "My hat is on the table."⁸ Similarly, in ethics, one might contend that, Quine notwithstanding, some such distinction between "meaning" and "substance" does in fact differentiate importantly distinct kinds of questions.⁹ In any event, if one rejects the analytic-synthetic distinction altogether, this would seem to undermine the objection that OQ is invalid. If the object of OQ is, as just seen, to show that certain statements about "good" or "ought" are not analytic, or true by definition or in virtue of meanings, and so on, then rejecting the analytic-synthetic distinction would presumably imply that the problem with the argument is not its logic, but rather the unintelligibility of its conclusion.

II

A more recent version of Harman's objection has been presented by Putnam which more explicitly connects Moore's argument against naturalism with the theme of concepts (or "notions") and linguistic meaning that is salient in Moore's own characterization."¹⁰ Putnam does not explicitly discuss OQ by name, but focuses instead on Moore's closely related idea of the "naturalistic fallacy": an expression coined by Moore which initially suggests the mistake of inferring normative statements from statements of natural fact, though Moore applies this label also to attempts by naturalists at bridging Hume's gap between "is" and "ought" with a premise defining the latter in terms of the former. OQ, then, is Moore's argument for the claim that the naturalist's "definist fallacy" (as Frankena calls it) really is fallacious or mistaken.¹¹ Putnam agrees with Harman's assessment (which he evidently arrives at independently) that OQ is logically invalid. As Putnam observes, Moore is arguing that if ethical naturalism was true then some statement such as

S: This action is not good even though it is conducive to maximizing utility.

is self-contradictory. Moore presumably demonstrates that it is not self-contradictory by observing that "This action maximizes utility, but is it good?" is an open question. Putnam argues, however, that even if ,S is not contradictory, it does not follow, as Moore must claim, that "although being Good and being conducive to maximizing total utility might be correlated properties, they could not be the same identical property."¹² According to Putnam, OQ rests on assumptions which he and other contemporary philosophers of language reject, the first being Moore's implicit denial of the synthetic identity of properties. This mistake makes OQ subject to the same type of criticism waged by Harman:

One could use Moore's 'proof to show that temperature must be a 'non-natural property,' in fact. For one is not contradicting oneself when one says 'x has temperature Ô but x does not have mean molecular energy E', where E is the value of the kinetic energy that corresponds to temperature T, even if the statement is always false as a matter of empirical fact. So, Moore would have to conclude, Temperature is only correlated with Mean Molecular Kinetic Energy; the two properties cannot literally be identical.¹³

Moore's mistake of assuming that an indentity of properties cannot be asserted by a synthetic statement is presumably generated by another more fundamental mistake, namely that of confusing properties with concepts:

The concept 'good' may not be synonymous with any physicalistic concept (after all, moral-descriptive language and physicalistic language are extremely different 'versions', in Goodman's sense), but it does not follow that being good is not the same property as being P, for some suitable physicalistic (or, better, functionalistic) P. In general, an ostensively learned term for a property (e.g., 'has high temperature') is not synonymous with a theoretical definition of that property; it takes empirical and theoretical research, not linguistic analysis, to find out what temperature is (and, some philosopher might suggest, what goodness is), not reflections on meanings."¹⁴

Putnam further suggests that this criticism of Moore may be seen as an application of Kripke's theory of names.¹⁵ According to Kripke, scientific identity statements like "Temperature is mean molecular kinetic energy" or "Water is H20 " are metaphysically necessary, since the terms "temperature" and "mean molecular energy", or "water" and "H20, " rigidly designate, that is, in all possible worlds, the same physicalistic property, even though we cannot know the truth of such statements a priori. Since these statements are not known to be true a priori, it is presumably an "open question" whether they are true, and consequently Moore's argument would seem to commit him here to the false conclusion that they must be contingently true, that is, that the properties in question cannot really be identical. Similarly, Moore argues that for any natural property N, goodness cannot be identical to N-ness since it is an open question whether N is good, but for Kripke goodness is N-ness just in case that "is good" and "has N" rigidly designate the same property, whether or not this identity can be known a priori, or can be "significantly questioned." On this view of the objection, then, the logical invalidity of OQ is due to the fact that it sets up an epistemic test for metaphysical necessity or essence, and Kripke generally denies that a-priority can be so conflated with necessity.¹⁶ As Putnam concludes:

These ideas of Kripke's have had widespread impact on philosophy of language, metaphysics, and philosophy of mathematics; applied to Moore's argument they are devastating. Moore argued from the fact that [S] can only be false contingently, that P (for some natural property P) could not be an essential property of goodness; this is what the¹⁸ new theory of necessity blocks. All that one can validly infer from the fact that [S] is not self-contradictory is that 'good' is not synonymous with 'conducive to maximizing utility' (not synonymous with P, for any term P in the physicalistic version of the world). From this non-synonymy of words nothing follows about non-identity of properties. Nothing follows about the essence of goodness.¹⁷

In an earlier writing, Putnam acknowledges Moore's insistence that he is illuminating concepts, not words, but Putnam argués that this comes to much the same thing.¹⁸

This criticism of Moore again raises the issue of what OQ is supposed to prove (or disprove). If the argument is an attempt to refute an ethical naturalist's claim about analyticity, or the meanings of the terms "good" and "ought," and so on, then it has not been shown that Kripke's ideas are so devastating. In general, Kripke's rigid-designator theory of necessity does not settle the issue of what terms mean (or the meaning of "meaning," as it has been referred to in the literature) and in fact some of what he says conveniently suggests that an epistemic criterion is appropriate for testing meanings. Kripke disassociates the qualities used to "fix the reference" of a term from the meaning of the term. Various descriptions involving accidental or contingent properties may be used to "define" a term in this reference-fixing sense—as when "gold" is defined as a yellow metal, a "meter" is defined as the length of a stick in Paris, or terms like "water," "temperature," "Hesperus" and "Phosphorus," and so on, are defined with respect to certain overt phenomenological qualities—even though, according to Kripke, such descriptions do not give the meanings of, or become synonymous with, these terms.¹⁹ In the case of natural-kind terms, Kripke notes that it is not self-contradictory to suppose that a thing so described should lack the qualities paradigmatically used to pick it out, or to give a standard dictionary "definition" of it: such as, to suppose that a given tiger is threelegged, or that all "cats" might turn out to be automata or demons instead of animals.²⁰ Analogously, in the present meta-ethical debate, then, one might argue that ethical naturalists have confused the reference-fixing qualities of being desired, pleasurable, conducive to happiness, and so on, with the meaning of "good" or "obligatory," or have attempted to give "definitions" which, as Kripke puts it, could turn out to be false. If so, it is not clear why the test of contradiction posed by OQ, as Putnam construes it, would be irrelevant to exposing their mistake. It is not implausible to suppose that native speakers can test the meanings of their own concepts or words by asking themselves whether certain exceptions are epistemically possible, and in fact this is not obviously so different from Kripke's arguing for his theory by asking readers to consider what name they would apply to an object under certain hypothetical circumstances.

It is true, as Putnam points out, that an epistemic contradiction-test will not tell one what natural properties a term refers to, but this again appears irrelevant to the issue of meaning which Moore uses OQ to decide. While Kripke distinguishes meaning and reference-fixing, he does not say (or imply) that the meaning of a term is entirely referential, or a matter of what it rigidly designates. Even if it were, Putnam's objection would still be incorrect. On this view, if "good" and "happiness-maximizing" ("pleasurable," "desired," etc.) do in fact designate the same properties, then the statement that something is good if and only if it is pleasurable, desirable, etc. becomes analytic, or true in virtue of the "meanings" of terms. The objection to OQ, then, should presumably be not that it is logically invalid, but that it is question-begging in its assumption that "X is happiness-maximizing but not good" is non-contradictory, or that the question of whether this statement can be true is really "open." Kripke, in any case, wants to say that scientifically discovered identities like "Hesperus is Phosphorus," "Gold has an atomic number of 79," or "Water is H20" are metaphysically necessary, not that they are analytic, that the terms involved are synonyms or that this is what "gold" or "water" means.²¹

On Putnam's own view, the meaning of a term is not merely the same as its extension, though he is concerned also to refute the view that "meanings are in our heads" or are to be equated with intensional concepts. As Putnam sees it, the meaning of a term like "water" has several components, including extension and stereotypic qualities, and though the latter are part of the "meaning" of any natural-kind term, it is not analytic that all things of that kind exhibit all such qualities (e.g., a tiger may lack stripes, four legs, etc . . . ).²² Since OQ is supposed to establish that "good" is a "simple" notion unlike "complex" natural kinds, Putnam's latter point will not vitiate Moore's conclusion, from the non-analyticity of S, that "maximizes utility" is not even part of the meaning of "good." More importantly, however, as an argument against the analyticity of naturalistic definitions, OQ itself need not involve any further conclusion about this referential or extensional component of meaning. According to Putnam and Kripke, "Water is H20" is, though necessarily true, still "a synthetic identity of properties," and if OQ is supposed to establish only the conceptural point that naturalistic definitions are synthetic, then its validity is compatible with the Putnam-Kripke result in philosophy of language.

III

These considerations prompt a closer look at the prospect of a synthetic identity in ethics, between moral and natural properties, of the same kind established by science in the analogy on which Harman and Putnam base their objection. In defense of the objection, it may be argued that Moore and other anti-naturalists using OQ are, after all, interested in establishing something about moral properties, in addition to any point about the meaning of moral terms. The initial force of the objection derives from the fact that Moore himself does hold a metaphysical view about what goodness is, and he does appear to deny that "good" and "conducive to happiness" are identical properties, even synthetically. It may also be noted that Harman, having presented the above objection to OQ, goes on to suggest that this argument from the analogy between ethics and science works only against what he calls "naturalism in general," and not against "definitional naturalism."²³ Though, as seen below in Part IV, there in some equivocation in Herman's treatment of these, the former is evidently the thesis that moral properties can be reduced to natural ones, in the same way that science produces reductionist identities, such as that reducing color to physical and optical properties, while only the latter "definitional" variants of naturalism assert that "moral judgments [statements?] are definitionally equivalent to natural judgments," that is, yielding a synonymy of meaning.²⁴ The suggestion, then, is that OQ is indeed logically invalid when applied at least to "general" or (non-definitional) reductionist naturalism.

Since OQ is originally Moore's argument, we might begin again by inquiring whether Moore can plausibly be charged with having given a logically invalid argument against this form of naturalism. Several points should be kept in mind here. First, the invalidity objection depends crucially on the assumption that OQ must be an attempt to show that "good" cannot be, as Putnam puts it, a "physicalistic property"; however, if Moore in fact does not clearly delineate properties from concepts, or from word-meaning—as we've seen Putnam himself concedes—then presumably Moore's use of the term "property" cannot be read unambiguously as supporting this interpretation. The passage cited earlier which presents OQ is phrased interchangeably in terms of "predicates," "meaning," and the distinctness of "notions before our minds"—that is, which would seem to make an epistemic test relevant. Again, Putnam himself, earlier in the text containing his objection and elsewhere, has distinguished two senses of property: one pertaining to physicalistic magnitudes, and the like, in science, and an older usage which is merely an equivalent expression for "concept" or "predicate," where the principle of individuation is synonymy.²⁵ Thus, it is odd that Putnam or Harman should suppose that Moore must be addressing the newer, scientific sense of property required by their objection.

The foregoing point is related to a second: Moore's argument against naturalism need not be construed as establishing, by itself, any view about what kind of property goodness is. Here there is an important disanalogy between the application of OQ in ethics, and this style of argument as applied to "temperature" and "mean molecular energy" in science, for in ethics a popular alternative to naturalism is the emotivist or non-cognitivist view that moral language does not ascribe any properties at all, natural or non-natural. Since OQ does not establish non-natural properties in ethics, then, it is misleading for Putnam to criticize OQ on grounds that, applied in science, it would yield the conclusion that temperature is a non-natural property. According to Moore, one cannot conclude that "good" is a simple and indefinable non-natural property until one has eliminated both naturalistic theories of meaning as well as the possibility that good "means nothing at all, and there is no subject of Ethics."²⁶

Admittedly, Moore thought that a second application of OQ could establish that "good," as compared with "pleasant," denotes a "unique property of things"—but again, in the sense of a "distinct meaning" or "unique object before the mind."²⁷ Writing in 1903, Moore clearly underestimates the prospect of non-cognitivist alternatives to naturalism, that is, the possibility of providing some type of subjective, emotivist (prescriptive, etc.) "meaning" for ethical terms; however, the oversight in this second argument for non-natural properties does not vindicate the charge that Moore's first argument against ethical naturalism is logically invalid, and indeed emotivists themselves have found OQ a persuasive argument against naturalism.²⁸

It may be replied that OQ would still be invalid in science, even when limited to a conclusion about what "temperature" is not; however, the objection is otiose if there is a more fundamental disanalogy between ethics and science: namely, if there is, in ethics, no further basis, beyond conceptual or linguistic considerations, for so identifying moral properties with, or "reducing" them synthetically to, natural or physicalistic ones. The guiding spirit of naturalism in ethics is a modern, scientifically-minded quest to make moral statements confirmable by an empirical method, so as to obviate problems of the older tradition of intuitionism. The latter type of theory would make ethics most closely analogous to science (or at least to a prevalent picture of that side of the analogy) by postulating a special realm of metaphysical moral properties and a special extrasensory faculty to perceive them, and this would in any case be the most obvious way to get an objective ontology of properties which moral terms might synthetically denote. Naturalists themselves, however, reject such properties, and Harman too denies this analogy between ethics and science, on grounds that it is not the best explanation of moral experience, a claim which presumably encapsulates standard empiricist arguments that have been around since Hume: 1) the nonverifiability of moral statements by the five senses, the uniqueness of a putative moral sense, disanalogies with ordinary perception, and so on; 2) the degree of moral disagreement in the world, its emotional character, systematic differences between cultures, etc.; and 3) alternative Humean accounts of the origin and function of moral belief, and so on.²⁹ Whether or not one accepts this line of argumentation, the point is that it is accepted by those whom OQ is designed to refute, and if accepted, it would tend to eliminate the prospect of synthetic identities in ethics of the sort to which Harman and Putnam call attention in science. Moreover, this prospect in ethics, which is supposed to make OQ invalid, is eliminated by the same type of argumentation as establishes such identities in science. Another way to put the point is to say that OQ may be supported by supplementary argumentation, such that the combined argument is not logically invalid, even on the Harman-Putnam interpretation. The upshot is that Moore's emphasis, then, on an epistemic or linguistic instrument of analysis should be more useful in ethics than the objection allows.

Admittedly, Moore himself could not avail himself or empiricist arguments against intuitionist or metaphysical theories of morality. On the one hand, Moore's own position is intuitionist. On the other hand, he suggests that OQ is sufficient against other systems of what he calls "metaphysical ethics"—exemplified in the Stoics, Spinoza, Kant or Hegel—which attempt to "reduce" (Moore's term) normative statements to, or infer them from, metaphysical propositions about what is real or rational, etc. Moore rejects these as committing the "naturalistic fallacy," the motor of which, as we have seen, is OQ.³⁰ Certainly OQ by itself, as a semantic test, does not directly refute the existence of metaphysical moral properties. This would suggest that Harman's objection to OQ as invalid against "general naturalism" is more applicable with respect to metaphysical theories of morality like Spinoza's or Hegel's, where the meaning of moral language is not the central issue. Such theories, however, are hardly what OQ is designed to refute, and in fact Moore explicitly applies OQ only against modern, definitional theories of meaning, not against metaphysical property-theories. Even in the latter case, OQ may be helpful in pointing out that an inferential move from metaphysical to moral propositions is not analytic, and therefore requires some additional, extra-semantic basis. Construed in this way, OQ would establish at least a burden of proof which shifts to the moral metaphysician. In terms of Harman's and Putnam's argument, the moral metaphysician must prove a synthetic identity, and if Moore cannot deny such an identity on empiricist grounds, neither can the metaphysician assert the identity on these grounds. Again, then, OQ figures to be at least an important part of an argument against a moral-tometaphysical reductionist program.

Putnam is critical of the foregoing empiricist argumentation, but his criticism is unconvincing. Part of the argument in (3) has been recently developed by J. L. Mackie in his contention that the apparent objective status of moral values is best explained as the result of a psychological process whereby values are "projected" out into reality in order to enhance their effectiveness as a means of social control.³¹ Putnam rejects this view, according to which moral experience is "mislocated subjective feeling," arguing that a theory which postulates an innate sense of justice is "simpler and more sophisticated" than the projection theory, and that a cognitive moral sense is analogous to intuition in logic, or the perception of colors.³² In rebuttal, one might first point out that traditional moral-sense or "perception" theories are, at any rate, not ontologically simpler than non-cognitivism, and that a full-scale projection theory, which must give an account of the specific psychological mechanisms involved in conditioning various types of moral motivation, can get quite "sophisticated." Secondly, since intuition in logic and mathematics enters into scientific explanations, it is indirectly confirmed in a way that the putative moral sense is not, and this is still more obvious in the case of colors: One's eyesight, unlike moral insight, fits into an explanatory scheme corroborated by the other empirical senses, and it is implausible to try to explain colors as projected attitudes conditioned in the same way by social factors. This is not to deny the point—popular since Kuhn, and developed also by Putnam—that values enter into our determination of facts, but even the statement of this thesis recognizes some conceptual distinction, and there appears to be an obvious difference, or asymmetry, in the way that facts and values characteristically arise or interrelate.³³

A final consideration in favor of a projection theory, when fully developed, is its explanation of the psychological phenomenon of moral emotions (associated with [2] above), and the solution it offers to Hume's problem of why moral judgments motivate actions. The property theory suggested by Putnam appears explanatorily redundant, insofar as it must presumably include the same type of theory about how moral emotions evolve, or why social behavior coordinates with moral cognition. But after this psychological, or sociobiological story is told, then, ontologically objective moral properties appear to play no explanatory role. Thus, the Ockham's-razor principle of theoretical simplicity in science favors the projection theory as being not only ontologically simpler, but as fitting into a simpler and more coherent explanation of moral experience. The projection theory should be especially attractive to the extent that one regards even the perception of physical "facts" as functionally determined by their survival value, or human interests in adapting to or dealing conveniently with the world. Here the projection theory, which provides a similar, functionalist account of moral values, would fit into a larger, unified explanatory scheme, part of which would be the value of theoretical simplicity itself.

The preceding motivation argument is not self-contained, but rather connects to another empiricist argument often thought to be independent, and which Putnam independently rejects. Thus, another of Mackie's arguments against intuitionism, this one falling under the rubric of (1) above, is his contention that moral values, conceived as objective qualities in the world, be rejected as both metaphysically and epistemologically "queer" from an empiricist standpoint: that is, requiring an ontology of invisible value-entities unlike any natural objects, an imperatival element of to-be-done-ness as somehow built into certain goals or actions, invisible moral suction pipes between creditors and debtors, or invisible hooks linking individuals with objects to which they have a right, and so on.³⁴ The existence of such prescriptive properties, whose cognition somehow automatically motivates action, would presumably provide a response to the foregoing suggestion that an objectivist moral-property theory must be supplemented by an account of moral psychology, but if the queerness argument works, this avenue of response is unavailable. Putnam rejects the queerness argument:

Mackie defends his claim that goodness is ontologically 'queer' by introducing as a premiss that one cannot know that something is good without having a "pro" attitude towards that something. This amounts to assuming emotivism in order to prove emotivism. The devils in hell are frequently depicted as using "good" with negative emotive force .. . ; contrary to Mackie, 1 do not find such uses to be linguistically improper or to involve any contradiction.³⁵

This response to the queerness argument has problems. If pro-attitudes are separated from moral cognition as Putnam asserts, then this reinstates the objection that a separate account must be given for the evolution of moral attitudes, and there is the problem again that moral properties and the moral sense become explanatorily superfluous. In any event, Putnam's agrument is off target, since the point about queerness is that the moral sense theory implies weird properties, containing an objective, cognizable imperatival element, whether or not a given moral observer has the usual attitude, or has decided to act contrary to the norm. Putnam's linguistic argument against this fails on two grounds. First, language does, in general, reflect this imperatival quality which would make moral properties strange. The uses indicated by Putnam are non-standard, and Putnam himself goes on in the above passage to observe that moral terms do, in general, have commending force.³⁶ Secondly, Putnam's argument seems to overlook the same basic point that he uses against Moore. Even if a speaker may, without contradiction, use moral terms without commending, a moral property may still be synthetically identified as prescriptive. Such a property, exemplified by Plato's form of justice, would indeed be "queer" from an empiricist perspective. The uniqueness of metaphysically objective value enities must itself be at least a factor weighing against their acceptance from that standpoint, but part of their "queerness" is also their lack of any explanatory function in an empiricist scheme. Of course, this entire line of argument presupposes an empiricist scheme, or what Putnam later refers to as "the assumption that the physicalistic version of the world is the Old True Theory."³⁷ Since the plausibility of this theory, however, connects to many issues outside of ethics, on which empiricists like Mackie give further argument, this position in ethics is in an important sense not merely an assumption.³⁸ The full argument is certainly not "assuming emotivism in order to prove emotivism."

Despite the traditional contrast between intuitionist and modern naturalistic theories, much of the same empiricist argumentation may be applied against a synthetic identification of moral properties with objective ones, whether physical or metaphysical. In fact both types of reduction may be viewed as forms of "naturalism": While moral-tophysical reductionism is the standard sense of naturalism (hereafter EN1) which OQ is supposed to refute, intuitionism is naturalistic in a broad sense (EN2), insofar as it also conceives of moral values as somehow built into "nature." Indeed, Moore implicitly suggests this similarity in his above charge that "metaphysical ethics" likewise, along with EN1-theories, commits the "naturalistic fallacy."³⁹ Though EN1 appears more congenial to the modern outlook thanEN2, since the former does not postulate any metaphysical pipes or hooks in the universe, EN1 still implies a kind of "objective" ontology of values (goodness, justice, etc.) as mysteriously connected to physicalistic properties of actions (their being desired, pleasant, etc . . . ).⁴⁰ As with EN2, then, there is a queerness argument also against EN1, and if there is not the same epistemic oddity about "intuiting" EN1 properties, there is a corresponding difficulty as to how one is supposed to "see" that moral properties are identical with empirically observable qualities. Putnam's own remarks above suggest that moral properties in fact cannot be so identified with any of the simple physical properties indicated in standard naturalistic definitions, such as the first-person approval theory. Putnam notes that one may intelligibly say that X is good without approving of it, while Moore's OQ makes the same point in reverse: Given that X is approved, one may still intelligibly ask "But is X good?" Similarly, goodness does not appear to be identifiable with any simple natural property of being approved by others, that is by one's society, or with promoting the general welfare (in one's society), and so on. Though the origin and function of morality may be best explained in terms of socially accepted norms which generally promote human well-being—Harman, for instance, goes on to develop such a theory—there appears to be no single descriptive meaning of this sort which statements about goodness or Tightness must always have. If so, this would support the validity of the OQ test by explaining why the open question should be open; namely, that OQ may implicitly operate on the fact that the defect of any particular naturalistic definition is its descriptive inflexibility.⁴¹ The objection by Putnam and Harman would work if there were extra-linguistic reasons to assert synthetic identities in ethics, as there are in science; however, the foregoing argumentation undermines this analogy, and again from a perspective which naturalists themselves endorse. As indicated earlier, naturalists regard their definitions as theories of meaning, and these considerations indicate why naturalistic theories must be viewed in this way, rather than as synthetic reductions.

Putnam's subsequent reaction to ethical naturalism inadvertently suggests a stronger argument for OQ:

. . . We have built a certain neutrality, a certain mindlessness, into our very notion of Nature. Nature is supposed to have no interests, intentions, or point of view .. . It is this same mindlessness of Nature that makes the action-guiding [predicate] 'is right' . . . seem 'queer.' If one physicalistic property Ñ were identical with moral Tightness . . . that would be queer .. . It would be as if Nature itself had values.⁴²

As noted above, there are presumably broad, Quinean grounds for supposing nature to be value-neutral, and while rejecting Mackie's version of the queerness argument, Putnam evidently concurs in rejecting a picture of the world as having moral values truly built-in. This suggests, however, that OQ, as a linguistic test, may be relevant even against a synthetic identity of values with natural facts. If nature really did come equipped with values, we would expect pro-attitudinal meaning to become infused into moral language, to the extent that statements identifying value and nature should indeed begin to feel analytic. Moore's observation that it obviously is not, then, should be relevant linguistic evidence against the moral-property identity theory even in Putnam's sense. On the other hand, if value-neutrality is built into our very notion of "nature," then such fact/value identity statements can never be analytic, which is exactly what Moore's OQ confirms.

IV

As seen earlier, Harman raises two objections to OQ: 1) that it is circular, since the so-called "open question" must simply assume that moral statements are not equivalent to assertions of natural fact; and 2) that OQ is logically invalid, since one cannot infer the falsity of naturalism even if the open question is really open. Clearly, however, Moore cannot be making both alleged mistakes in the same argument. If OQ does assume as a premise the very conclusion it is trying to prove, namely, the falsity of naturalism, then at least OQ must be logically valid. Conversely, if the Harman-Putnam invalidity objection is correct, then Moore's desired conclusion does not follow from his premise, which assures that he has not already logically presupposed the conclusion. Thus, the new invalidity objection, based on an analogy to science, contains some good news for Moore, for if OQ is logically invalid, then at least the usual circularity objection must be incorrect. On the other hand, if the invalidity objection fails, as argued above, OQ is still vulnerable to the charge of circularity. We will see that there is also an alternative way to formulate the invalidity objection so as to accommodate the analysis of Part III. Further examination is needed, then, to determine which, if either, mistake Moore makes.

The conflict between the objections may be removed by saying that OQ is circular when applied to some types of naturalism, and invalid when applied to others. We have already seen that ethical naturalisms can be classified in terms of the kind of "natural facts" to which morality is reduced, and there are additionally different kinds of "reduction"—scientific or explanatory reduction, as contrasted to the semantic type with which Moore is primarily concerned. It may be argued, then, that OQ is circular against semantic naturalistic definitions, and invalid only against non-semantic reductions. Harman suggests this when, having presented the invalidity objection, he remarks, "The open question argument is often put forward as a refutation, not of ethical naturalism in general, but of a more particular version, which we might call definitional naturalism"; however, he goes on to argue that OQ is invalid also against at least some "definii tional" forms of naturalism:

There are, however, various kinds of definitions and the open question argument is not relevant to most of them. For example, a scientist defines water as H20 and, as we have seen, the open question argument applied to this definition does not refute it.⁴³

It is misleading for Harman to pick scientific "definitions" as illustrative of definitional naturalism in ethics. Though Harman never explicitly formulates what "general" naturalism is, the general idea seems to be a type of reduction which depends on explanation rather than "word meanings," and on this characterization scientific reductions are "general" rather than "definitional."⁴⁴ In any case, as has been argued, the definitions of science are established by a type of non-semantic reduction which is disanalogous to anything available in ethics, due to the sort of arguments that Harman himself elsewhere acknowledges. Consequently, "most definitions," including those of science, are simply irrelevant to appraising OQ in ethics; the appropriate conclusion would seem to be not that OQ is invalid against, but that it is merely inapplicable to, non-semantic definitions. As to the application of OQ to semantic definitions, Harman appears to fall back on the circularity objection:

Presumably the open question argument is aimed at someone who claims that a naturalistic definition captures the meaning of a moral term in the sense that moral judgments as we ordinarily use them are synonymous with judgments that describe natural facts. If it really is an open question whether an act that is C is an act that ought to be done—an open question even to someone who knows the meanings of "C" and "ought to be done," how can "C" and "ought to be done" be synonymous? It must be shown, not just assumed, however, that the relevant question is always open, no matter what the natural characteristics C.⁴⁵

The circularity objection, then, is that Moore, and subsequent theorists who use his argument, have merely assumed that the open question is open, which is precisely what ethical naturalism denies.

Contrary to this traditional view, I wish to argue that OQ, properly construed, can provide a helpful type of ethical argument which is neither logically invalid nor circular, even against semantic definitions. Whether proponents of OQ merely assume the falsity of naturalism depends in the first instance on what the test for an "open question" is. In the passage quoted at the outset, Moore says that a question is "open" if it can be significantly asked, that is, if we understand what it means to doubt or to deny an affirmative answer. Clearly, then, for a question to be "significant" it is not a sufficient condition that it be merely intelligible in the sense of being capable of an answer, as is a ("closed") question trivially answered in the affirmative, such as, "Is the Pope Catholic?"; nor is it a necessary condition of "significance" that a question be meaningful in the sense that its answer is informative, i.e., adding to one's present knowledge, as exemplified by the (usually) pointless (but "open") question "Is radioactivity harmful?"⁴⁶ These examples suggest another interpretation, namely that a question is open/significant if it is logically possible or not self-contradictory to answer it in the negative. If this is the criterion of an "open question," then Frankena and others who have waged the standard objection are right: Moore's argument appears circular.⁴⁷ Where "C" is a natural term and "M" is a moral expression such as "ought to be done," the very thesis of naturalism is that some substitutions of "C" and "M" mean the same, that is, that it is not logically possible to answer negatively some question of the form Q, "Are all C's also M's?" On a fourth interpretation, however, the open-question test is based not on logical possibility, but rather on the psychological possibility of being in doubt, or of thinking that one understands such doubt. This comports with the pervasive emphasis in Moore's remarks indicating that the OQ test is essentially subjective—based on "what we are thinking," what is "before our minds," or what we think is before our minds.⁴⁸ Oddly, this construal also squares with what appears to be Putnam's and Harman's own view that the OQ test is basically epistemic in some sense. But when Moore's argument is so construed, and applied against semantic naturalism, the argument must be at least non-circular, since a premise purporting to state only a psychological fact, about what one thinks could be doubted, is not itself merely a statement about what terms mean, what is logically possible, or what ethical naturalism says about this.

On the other hand, the argument is not obviously invalid, either. When applied against semantic definitions, the type of "epistemic" test involved in this version of Moore's argument operates, more specifically, on the basis of pre-philosophic linguistic intuitions. Moore is suggesting that the mere fact, as he puts it, that we ordinarily find at least "questionable" or doubtable a proposition equating a pair of C and M terms in ordinary language, is relevant to determining whether such a theory is correct. There is no reason to suppose that what people think they can imagine, or "have before their minds," cannot be useful in testing theories of meaning—if not theories of metaphysics, as discussed in Part III. There is something common-sensical, after all, about testing theories of word meaning by the linguistic behavior and attitudes of native speakers; indeed, there is otherwise some problem as to the sense in which "theories" of meaning are to be tested at all.

Before subjecting this version of Moore's argument to further critical analysis, it should be noted that the argument thus far is incomplete. Fully elaborated, the main strategy is to use openness as a test for synonymy, and various formulations of this link are available. In the above passage, apart from the point about doubtability, Moore makes three observations in reaction against Russell's definition of "good" as what we "desire to desire": i) that Q1, "Is it good to desire to desire [action] A?", is open (intelligible, informative, etc.); ii) that if Russell's alleged synonymy were correct, Q1 would be equivalent to Q2, "Do we desire to desire that we desire A?", but clearly Q1 and Q2 are not equivalent, Moore says, for when we think of Q1 "we have not before our minds anything so complicated . . ." as Q2 (italics added); and iii) that "moreover" it is obvious by inspection that an affirmative answer to Q1 (viz., "It is good that we desire to desire A") is not equivalent to "It is good that A is good," as presumably it would be if Russell's definition were correct.⁴⁹ In the above notation, then, the main OQ test for synonymy, as suggested by (i), is the following:

7' A C-term is not synonymous with an M-term if the corresponding question of form Q is open (in the psychological sense).

Perhaps in (iii) Moore's point is that Q3, "Is it good that A is good?", is not open; whereas Q1 is—indicating that the terms exchanged are not synonymous. This is not merely a reiteration of the first argument, since Q3 is not a Q-type question in terms of C's and M's. There must be an additional synonymy test:

T2 A C-term is synonymous with an M-term if the result of replacing an occurrence of either by the other in the corresponding Q-question does not change Q's, open/closed status.

On the other hand, the preceding argument does not add much in practice, since it requires that one already know the original question, Q1, to be open, which would seem more obviously to settle the issue under T1. Perhaps in (iii), then, Moore is merely dramatizing the fact that Ql is indeed open—that is, by contrasting it to an obviously closed question whose answer is a tautology. Alternatively, the point in (iii) may be merely to rework (ii), the point being that Q3 is obviously simpler than Ql. The complexity (or simplicity) argument in (ii) suggests several direct tests for synonymy which do not really involve an "open question":

73 A C-term is synonymous with an M-term if the result of replacing one for the other in a statement (question) does not change the complexity of that statement (question), or its practical function.

The latter idea of the practical use of a statement (or question), though not explicitly in Moore, is illustrated in Hare's adaptation of OQ where it is observed that the statement "An A which is C is good" can commend or guide choices, while "An A which is C is C" cannot.⁵⁰ Again, however, it is plausible that Moore intends (ii) not as a separate line of argument, but as reinforcing the first argument: that is, variation in the complexity of terms may be a factor which dramatizes the openness of a question, makes the question open or explains why native speakers so regard it. Similarly, in Hare's argument, the prescriptive or emotive meaning of M-terms, but not C-terms, would be—for contemporary theorists, if not for Moore—an additional factor explaining the psychological "feel" of openness in questions of the form "Are all C's also M's?"

If nothing further were said, the foregoing linguistic interpretation of OQ would be unlikely to impress many of Moore's contemporary critics, though the appropriate objection is neither circularity nor the new type of invalidity objection waged by Harman and Putnam. The latter objection is in fact compatible with at least the essential insight of the circularity objection. Harman's and Putnam's point was that even if a naturalistic definition is not analytic, Moore cannot validly infer that the definition (i.e., as stating a synthetic identity) is false, but one might argue at the same time that the OQ test fails, in the first step, to show the non-analyticity of such definitions. The central difficulty is that native speakers may simply be wrong about what terms mean. The precise character of this fallibility problem may be put in perspective by considering Moore's replacement arguments, which begin to look much like the infamous intentional fallacy:

1. Descartes knows that his body is spatially extended and divisible.

2. Descartes does not know that his mind is spatially extended and divisible.

3. Leibniz's Law (ranging over intentional properties).

4. Therefore, mind and body cannot be identical.

Compare:

1'. Moore knows (does not doubt) that X is desired.

2'. Moore does not know (can doubt) that X is good.

3'. LL

4'. Therefore, being desired is not identical with being good.

When we apply the stock objection that has been given to the Cartesian mind-body argument, this formulation of Moore's nature-morality argument is either logically invalid, i.e., if (2') is read de dicto, or circular, where (2') is taken de re and where one assumes that it is true. (Another standard objection, that intentional properties do not exist, or are not subsumed under Leibniz's Law, would generate a different kind of invalidity objection.) The above formulation of OQ, however, is not exactly Moore's argument, since Moore does not appear to rely in this way on intentional properties, and as argued in Part III above, OQ does not have to be an epistemic test of the referential meaning of terms. Moreover, the comparison of Descartes and Moore underscores the fact that, as just seen, the OQ test is subjective or psychological—that is, Moore's "significantly asks" is not a de re intentional verb. When construed de dicto, Moore's argument is again plausible, since presumably even the Cartesian argument, read de dicto, may validly and non-circularly establish that the terms "mind" and "body" differ in connotative, if not denotative, meaning. Still, critics may reasonably contend that Moore's argument for non-analyticity commits a similar, less circuitous form of argumentum ignoratium.⁵¹ This is suggested especially by Moore's remarks on complexity—to wit: Native speakers do not know that X and Y are synonymous, possibly because Y is more complicated; therefore, they are not. The appropriate objection to OQ, then, is that it is logically invalid after all, but on grounds entirely different from what Harman and Putnam indicate in their analogy to science.

Though more plausible than the standard circularity objection, the foregoing objection to OQ is unsuccessful. Admittedly, native speakers are not experts at judging philosophic theories of meaning; on the other hand, neither does the above argumentation establish that pre-philosophic linguistic intuitions are irrelevant to testing such theories. In general, this charge of invalidity derives much of its strength from interpreting OQ as i) the claim that linguistic intuitions are a conclusive refutation of ethical naturalism, rather than merely relevant evidence against it, and as ii) an attempt to refute all naturalistic definitions in toto, without the need to apply the test seriatum to particular definitions so as to accumulate the evidence. Moore can deny both, and even if he did hold (i) or (ii), those who use OQ today need not. It is true that Moore himself concludes that no naturalistic definition works, but he has no doubt considered all of the prominent proposals by naturalists, against which OQ has seemed effective, and he may simply be drawing a good inductive inference that it will continue to be effective against any other definitions of the same sort. Having presented OQ, as applied to Russell's definition, Moore does explicitly say that one must continue to apply the test, for example, against a hedonistic definition:

But whoever will attentively consider with himself what is actually before his mind when he asks the question 'Is pleasure (or whatever it may be) after all good?' can easily satisfy himself that he is not merely wondering whether pleasure is pleasant. And if he will try this experiment with each suggested definition in succession, he may become expert enough to recognize that in every case he has before his mind a unique object, with regard to the connection of which with any other object, a distinct question may be asked, (italics added)⁵²

While it is clear that Moore does not think one should infer the falsity of all naturalistic theories from one linguistic experiment, he also does not contend that native speakers are automatically "experts" here without "attentively considering" their linguistic intuitions, or (as he puts it on the previous page) exercising some "reflection."

These considerations raise larger issues, concerning the relationship of ethics to linguistics, which cannot be explored in the present essay; however, the initial plausibility of the above argument is in fact supported by contemporary reflection related to science. The argument has been that Moore's test at least focuses on relevant data, having evidentiary weight for (dis)confirming naturalistic theories of meaning. Here one must additionally observe that the format of Moore's test is at any rate not merely to treat native speakers as linguistic experts who pass judgment directly on a given naturalistic theory of meaning—they are not simply asked whether they think a given definition is correct. Instead, the point of posing a question to speakers, and asking them whether they seem to understand its denial, is presumably to "catch them in the act," so to speak, of actual linguistic behavior, not of meta-linguistic theorizing. Surely, this first-order linguistic behavior should have some general relevance to testing a theory of synonymy or analyticity. Thus, from a Quinean point of view, the analyticity of statements is determined by a behavioristic test that gauges the degree of one's reluctance to deny or revise them, as determined ultimately by the overall fit of one's conceptual scheme with experience.⁵³ Similarly, Moore's test measures the intuitive reluctance of speakers to deny, for instance, that a pleasant thing is good. Whether or not one finds the argument compelling in every instance, reliance upon this sort of linguistic data, as a general test of semantic definitions, appears to be neither circular nor invalid.

NOTES

¹ G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1980), pp. 15 ff. First published in 1903.

² Gilbert Harman, The Nature of Morality (Oxford: Oxford University Press, 1977), p. 19. Cf. W. K. Frankena, Ethics (Englewood Cliffs: Prentice-Hall, 1973), pp. 99 ff.

³ Harman, loc. cit.

⁴ Naturalistic definitions may also be proposed as stipulative or reformative, and admittedly ethical naturalists themselves may not always distinguish these; however, historical naturalists characteristically claim that their definitions capture at least the substantial content of what ethical terms mean in everyday usage. Cf. Richard B. Brandt, Ethical Theory (Prentice-Hall, 1959), pp. 156 ff.

⁵ Moore, op. cit., p. 7.

⁶Ibid., pp. 17-19, 66 ff.

⁷ Harman, op. cit., p. viii.

⁸ See Hilary Putnam, Mind, Language and Reality in Philosophical Papers, Vol. 2 (Cambridge: Cambridge University Press, 1975), pp. 14, 36. On the reaction to Quine's arguments, cf. also Roderick Chisholm, Theory of Knowledge, 2nd edition (Englewood Cliffs: Prentice-Hall, 1977, 2nd edition), p. 61.

⁹ If the meta/normative distinction has been "undermined" at all, surely ethical theorists like Hare have had more to do with this than Quine. Actually, however, even if one agrees (with Hare) that substantive normative conclusions are to be founded on linguistic intuitions, this would presumably show only that there are interesting connections between "meta-ethics" and "normative ethics," not that a useful schema for separating (or so relating) ethical issues has been undermined. It is perhaps still less plausible, as Harman also suggests in the cited passage, that Quine's attack on the analytic-synthetic distinction is what precipitated the shift away from linguistic meta-ethics toward "substantive" normative concerns in the 60's and 70's. Oddly enough, having expressed his dissatisfaction with the older linguistic tradition of metaethics (c. 1930-1960—cf. op. cit., pp. vii, ix), Harman proceeds to develop a theory of moral relativism which relies upon linguistic arguments: cf. Chs. 9 and 10.

¹⁰ Hilary Putnam, Reason, Truth and History (Cambridge: Cambridge University Press, 1981), pp. 205 ff.

¹¹ The classic piece on this is W. K. Frankena's "Naturalistic Fallacy," reprinted in the collection of his essays ed. by Kenneth E. Goodpaster, Perspectives in Morality (Notre Dame: University of Notre Dame Press, 1976). Frankena makes four major points: (1) that any such inference from is to ought can be made logically valid simply by adding a naturalistic definition as a premise (p. 4); (2) that if this premise is regarded as an "ought" proposition, then Moore is not merely endorsing Hume's is/ought bifurcation (p. 5); (3) that, in any case, if Moore is objecting to a type of definition, then the term "fallacy" is a misnomer (loc. cit); and (4) that Moore has no real argument against such definitions (pp. 8f). Though (1) (3) are technically correct and conceptually illuminating, it is reasonable to view Moore's "naturalistic fallacy" as being essentially in the spirit of Hume's reaction against the "imperceptible" inferential move from is to ought in "vulgar morality." Moore reinforces Hume's bifurcation by also disallowing any "perceptible" or overt definitional connections between "is" and "ought" in less vulgar systems of ethics, and as Frankena later saw (cf. Ibid., p. 210), OQ is the argument at work behind the scenes here.

¹² Putnam, Reason, Truth and History, op. cit. p. 206.

¹³Ibid., pp. 206-7.

¹⁴Ibid., p. 207.

¹⁵Ibid., pp. 207-8.

¹⁶ Putnam explicitly discusses neither Moore's concept of an "open question" nor Kripke's concept of "rigid designation." See Saul Kripke, Naming and Necessity (Cambridge, MA: Harvard University Press, 1980), pp. 34-39. (Putnam does not cite specific pages.) Cf. also id., "Identity and Necessity," Identity and Individuation, ed. by Milton K. Munitz (New York: New York University Press, 1971), pp. 149 ff.

¹⁷Ibid., p. 208.

¹⁸ Putnam, Mind, Language and Reality, pp. 3, 8 ff.

¹⁹ Cf. Naming and Necessity, pp. 26, 54 ff., 116 ff., 135 ff.

²⁰Ibid., pp. 119 ff, 122, 125-26.

²¹ Cf. ibid., pp. 103 ff., 125, 128, 141 ff. Kripke does not try to explicate the concept of analyticity. He stipulates an "analytic" truth to be "one which depends on meanings in the strict sense and therefore is necessary as well as a priori": pp. 56, n. 21; 122f, n. 63; and cf. pp. 39, 56n. Since the above truths are not known a priori, then, this presumably indicates that the terms do not "mean" the same even though they refer to the same objects. In the p. 122 note Kripke also recognizes a sense of analyticity whereby a statement can be known a priori due to its being true by "definition" in the reference-fixing sense. This would make some analytic truths contingent, but they still do not involve synonyms for Kripke: cf. pp. 14 ff. 63.

²² Addressing the example of the "twin earth" in which what is called "water" turns out to be not H20 but rather some other chemical composition XYZ, Putnam sees two options for a theory of meaning to go with Kripke's theory that names of natural kinds are rigid designators: 1) to say that the word "water" in the two worlds (our earth and the "twin") has the same meaning but different extensions, i.e., to identify meanings with concepts, or ii) to say, as Putnam prefers, that "meaning determines extension," so that different extensions automatically indicate different meanings: cf. Mind, Language and Reality, pp. 223 ff., 234, 245 ff; "water" is defined at p. 269; for the point about meaning and ambiguity, cf. pp. 140 ff., 250, 256.

²³ Harman, op. cit., p. 19.

²⁴Ibid., pp. 19-20; ethical "reduction" is discussed on the previous page, and is introduced at pp. 13-14 in connection with the idea of explanation.

²⁵ Cf. Reason, Truth and History, p. 84, and "On Properties," reprinted in Mathematics, Matter and Method in Philosophical Papers, Vol. 1 (Cambridge: Cambridge University Press, 1979), pp. 305 ff. Putnam illustrates what "predicates" are with the timely example of "Is existence a predicate?" (i.e., one of Moore's famous titles), though observing that this is not merely a syntactical issue.

²⁶ Moore, op. cit., p. 15. Harman echoes the same point where he notes that moral "nihilism" is an alternative to naturalism: Ibid., p. 17. Oddly, he refers to nihilism here as a form of naturalism, though "extreme nihilism," as Harman conceives it (i.e., moral skepticism) appears neither to "reduce" morality to anything, nor to be a "definitional" theory of morality.

²⁷Ibid., p. 17. The arguments against naturalistic meaning, and against the meaninglessness of 'good', are given in paragraphs (1) and (2), respectively.

²⁸ For example, with "emotivist" being used here in a broad sense: R. M. Hare, The Language of Morals (Oxford: Clarendon Press, 1963), pp. 90 ff. An emotivist theory of what moral terms do mean would provide a type of argument, other than OQ, for criticizing all naturalistic definitions. Cf. C. L. Stevenson, Ethics and Language (New Haven: Yale University Press, 1944).

²⁹ Harman discusses the problem of observation in The Nature of Morality, pp. 6-9. He does not explicitly address "intuitionism," nor does he articulate this auxiliary argumentation, which however is needed to make good the claim that intuitionism is not the "best explanation." Accordingly, it should be understood that empiricists need not regard any single argument in the above categories as conclusive. Instead, the arguments may be regarded as cumulatively constituting a case against intuitionist or rationalistic ethics. This case also connects to arguments beyond ethics—cf. n. 38 below.

³⁰ Moore, op. cit., pp. 114 ff. In Moore's terminology, the "metaphysical" is "supersensible," or what does not exist in time (pp. 110 ff.). By contrast, Moore's intuited "non-natural" quality of goodness attaches to natural objects which are perceptible by the senses. See Hasna Begum, "Moore on Goodness and the Naturalistic Fallacy" Australasian Journal of Philosophy, vol. 57 (1979), pp. 251 ff.

³¹ J. L. Mackie, Ethics: Inventing Right and Wrong (Harmondsworth: Penguin Books, 1977), pp. 42 ff. Cf. C. L. Stevenson's characterization of non-natural qualities as "an invisible shadow cast by confusion and emotive meaning": Ethics and Language, op. cit.), p. 109.

³²Reason, Truth and History, op. cit., pp. 142, 144-47. Mackie is not explicitly addressed here.

³³ As one commentator recently suggests, our beliefs about the injustice of slavery are not forced on us by the physical properties of objects, i.e., in the same way as our belief that fire engines are red is forced on us: Dworkin in Ronald Dworkin and Contemporary Jurisprudence, ed. by Marshall Cohen (Totowa: Rowman & Allanheld, 1984), p. 277. Putnam is concerned here to deny that belief in justice is like belief in ghosts, a view which he associates with the projection theory; however, Hume too distinguished belief in justice from mere superstition without postulating moral properties: An Enquiry Concerning the Principles of Morals (Indianapolis: Hackett Publishing Co., 1983), Sec. III, p. 31. On the comparison of ethics and mathematics, cf. Harman, The Nature of Morality, pp. 9 ff.

³⁴ Mackie, op. cit., pp. 40, 74.

³⁵Reason, Truth and History, op. cit., p. 209.

³⁶Ibid., pp. 209-10. Putnam observes (p. 210) that "good" is more often used without personal approval than are "should," "ought," "right," etc.; but even the latter have intelligible non-standard, non-approving usages: Cf. Harman, The Nature of Morality, pp. 47 ff., 122-24. Nor does Mackie say that all uses of "good" imply personal approval: Cf. p. 55.

³⁷Reason, Truth and History, op. cit., p. 210.

³⁸ Thus, an intuitionist or moral objectivist, who does not merely assume that the physicalist version of the world is false, might go on to argue that empiricism is implausible in other areas of philosophy, and therefore unreliable in ethics—an argument which empiricists may go on to rebut: Cf. Mackie, Ethics, op. cit., p. 39. A more modest argument would be that empiricism—presumably no less than rationalism—rests on an "assumption" in the sense that the choice, or "success," of any ultimate principle of knowledge depends at least partly on one's initial, common-sense judgement as to what metaphysical objects exist, or what obviously is known (i.e., another G. E. Moore shift, writ large). Thus, the games of metaphysics and epistemology must be played simultaneously, with ethics as one, of many, substantive application areas. Chisholm has argued that the epistemological problems of morality, other minds, and God are interrelated in this way: Theory of Knowledge, op. cit., pp. 3, 119 ff. Contrary to Chisholm's own "choice" here, one might conclude that there is no ethical knowledge of the kind commonly supposed, if a non-cognitivist explanation can be given for that supposition, and on the basis of empiricist principles (Quinean rather than positivist) which have acceptable implications in other areas. This much broader claim, at any rate, underlies the foregoing arguments in ethics.

³⁹ See n. 30 above. Cf. R. L. Franklin, "Recent Work in Ethical Naturalism" in American Philosophical Ouarterly Studies in Ethics, (1973), p. 55, distinguishing naturalism in the first sense (EN]), according to which moral statements express empirically or scientifically ascertainable facts, apart from the will of God or intuition, from the second, broader sense (EN2), according to which moral statements express facts, even if non-natural or metaphysical ones. EN2, then, includes what might be called "ethical super-naturalism," where moral properties are intuited and revealed by God, and perhaps it would have been more accurate had Moore characterized the fallacy of (non-theistic) metaphysical ethics as committing the "super-naturalistic fallacy."

⁴⁰ Cf. Mackie, Ethics, op. cit., pp. 38, 41. Of course, desires, pleasant experiences, etc., are "subjective" in a sense, and Mackie so labels a naturalistic approval definition as denying the metaphysical brand of moral "objectivity" in Plato's forms: Cf. pp. 17, 23 ff. Still, the psychological or social phenomena involved in EN1 definitions constitute a type of "objective" fact which moral statements may describe, and to which values attach.

⁴¹ Cf. ibid., pp. 50, 60-61.

⁴²Reason, Truth and History, op. cit., p. 211.

⁴³ Harman, The Nature of Morality, op. cit., p. 19.

⁴⁴ Cf. p. 18, and nn. 23f above. Earlier Harman characterizes naturalism as the "general view" that "all facts are facts of nature" (p. 17), and subsequently observes that the "general naturalist" abandons religion (p. 21). Thus, a "general naturalist" need not subscribe to ethical naturalism, which in turn comes in "general" and "definitional" variants. Harman thus appears to exclude from the ambit of "naturalism" the EN2 theories discussed in Part III.

⁴⁵Ibid., pp. 19 ff.

⁴⁶ Cf. Principia Ethica, op. cit., pp. 15f. Moore associates the notion of an "open" question both with its "intelligibility" and the "information" it calls for; however, as just seen, these latter notions are overly broad. Of course, a question may "call for information" which one already has, and this broader sense of an answer's being "informative" does seem to be central to the notion of "openness." Cf. Brandt's discussion, which mentions this only as indicating the difficulty of equating open (or intelligible) questions with those that are not "pointless": Ethical Theory, op. cit., pp. 164-65. But Moore presumably has in mind non-linguistic "information," and the more basic test of when a question calls for any such information appears to be doubtability, the idea with which Moore concludes the passage, and on which the circularity objection focuses.

⁴⁷ Cf. note 2 above, and Roger Hancock, "The Refutation of Naturalism in Moore and Hare," The Journal of Philosophy, vol. 58 (1960), pp. 328 ff.

⁴⁸Principia Ethica, op. cit., p. 16. Cf. Frank Snare, "The Open Question as a Linguistic Test," Ratio, vol. 17 (1975), p. 126. This essay contains the basic move for defending Moore, though without regarding this as Moore's own view of OQ (p. 129), and more importantly, without supplying the needed argumentation developed below. When this argument is supplied, it is not at all obvious that Moore himself—who was clearly cognizant of the fact that he was trying to refute a certain type of linguistic theory—did not view the OQ argument as involving a linguistic test.

⁴⁹Ibid., pp. 15 ff.

⁵⁰ R. M. Hare, The Language of Morals, op. cit., pp. 82 ff., 89 ff., 145, 195. Compare also Fred Feldman, Introductory Ethics (Prentice-Hall, 1978), pp. 199-203. Feldman formulates two versions of OQ, the first being roughly a restricted "complexity" version of 73 (p. 200), and a second corresponding to (p. 203). This discussion is typical of the literature in that the operative notion of 'openness' is left unanalyzed, and consequently something like the Tl-variant is charged with circularity.

⁵¹ The intentional fallacy may be viewed as merely a fancy type of argument from ignorance.

⁵²Principia Ethica, op. cit., p. 16.

⁵³ W. V. O. Quine, "Two Dogmas of Empiricism" in From a Logical Point of View (Harvard, 1953), pp. 24, 27, 31, 37 ff., 41. In the Quinean chain of analysis, synonymy for single terms is defined as interchangeability in truth-functionally synonymous statements, which in turn are characterized by analyticity—the latter resting finally on a behavioristic criterion. Cf. M. White, "The Analytic and the Synthetic: An Untenable Dualism" in Semantics and Philosophy of Language, ed. by Leonard Linsky (Urbana-Champaign: University of Illinois Press, 1952): the notion of "self-contradiction" characterized by "a certain feeling of horror or queerness" which native speakers have, or their "degree of discomfort" (p. 325).