G.E. Moore’s Naturalistic Fallacy and Open Question Argument Reconsidered by William Piervincenzi Submitted in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Supervised by Professor Robert L. Holmes and Senior Lecturer John Gates Bennett Department of Philosophy The College Arts and Sciences University of Rochester Rochester, New York 2007
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
G.E. Moore’s Naturalistic Fallacy and Open Question Argument Reconsidered
The author was born in Bethpage, New York on May 24, 1969. He
attended the University at Stony Brook from 1988 to1992 and graduated with a
Bachelor of Arts degree in Philosophy. He came to the University of Rochester in
the Fall of 1994 and began graduate studies in Philosophy. He received a Rush
Rhees Fellowship in 1994 and held teaching assistantships for the subsequent four
years. He pursued his research in Ethics under the direction of Professor Robert
Holmes and received a Master of Arts degree from the University of Rochester in
1999.
iv
Acknowledgments
Completion of this dissertation would not have been possible for me
without the kindness and generosity of people too numerous to mention. I wish to
extend special thanks, however, to the following:
Eileen Daly and Eva Cadavid for inspiring me to stick with it and to question
others instead of myself; Greg Janssen for his friendship and encouragement over
the years; Keith McPartland for his enthusiasm and comments, with Greg, on
chapter drafts over coffee at Canal Town; my family, and especially my parents,
Bill and Edna Piervincenzi who stopped asking “when will you be done?” just in
time; Mary Schweizer for good times and pizza; the Diggin family, especially
Cecilia for love and support and especially for Kerry; John G. Bennett who gave
so much of himself through the final months of writing; Robert L. Holmes for his
patience and calming influence; Joe and Jeanne Norwin for being our spiritual
sherpas; all of the Delaware Park Olde Tymers for whom it is always summer;
and Chewbacca, Goldie, and Quincy for quiet comfort when it was needed most.
I extend the most special thanks and love to my wife, Kerry Diggin, to whom this
work is dedicated. (Heaven knows she’s earned it).
v
Abstract
G.E. Moore is justly famous for his arguments against ethical naturalism;
however, after one hundred years of commentary, Moore’s arguments of the
opening chapters of Principia Ethica—his claim that most ethical theories commit
the naturalistic fallacy and his argument that moral goodness is a simple,
indefinable property—are poorly understood. This work seeks to augment our
understanding of these crucial arguments. This work offers an interpretation of
what it means to commit the naturalistic fallacy that is maximally consistent with
Moore's text and examples, distinguishes Moore’s claims about the naturalistic
fallacy from the open question argument, and corrects mistaken interpretations of
the open question argument. In pursuing these goals, I find that the naturalistic
fallacy and the open question argument are complementary, rather than redundant,
arguments. I also find that, contrary to the traditional reading, Moore’s open
question argument actually consists of two arguments, one of which is supported
by at least five sub-arguments, most of which turn on our intuitions about
language and meaning, and none of which is wholly successful. Moore uses the
open question argument to show that goodness is not identical with any complex
natural or metaphysical property. This still leaves the possibility that goodness is
identical with a simple natural or metaphysical property. His claim that
philosophers who attempt to define goodness commit the naturalistic fallacy
vi
undermines support for any such identification, even if it does not conclusively
refute such definitions.
vii
Table of Contents
Chapter 1 The Legacy of Principia Ethica 1
Chapter 2 Goodness, Simple and Indefinable 15
Chapter 3 What is the Naturalistic Fallacy? 35
Chapter 4 The Naturalistic Fallacy: A New Interpretation 88
Chapter 5 Moore’s Proof that ‘Good’ Refers to a Simple,
Non-natural Property 136
Chapter 6 Conclusions 194
Bibliography 211
1
Chapter 1
The Legacy of Principia Ethica
I. Introduction
It is safe to say that after G.E. Moore published Principia Ethica,1 just
over one hundred years ago, philosophical ethics underwent a dramatic change.
All forms of ethical naturalism and metaphysical ethics,2 theories that accounted
for the majority of approaches to ethics, were under attack. Moore offered an
ambitious and vexing collection of arguments that inspired, and continues to
inspire, high praise and scathing criticism.3 And while his arguments did not
sound a death knell for naturalism, they did inspire the development of plenty of
alternative views.4
Moore’s intent in Principia Ethica was to create a new science of ethics,
one that better captured what we had in mind when we called an action or a state
of affairs good. He tried to accomplish this through observations about how we
1 G.E. Moore, Principia Ethica (1903). There are two versions of Principia Ethica in wide use,
which have the same text and different pagination. The newer version is the “revised edition” of
1993. Since this new version contains a useful, previously unpublished preface to a planned but
never written second edition of Principia Ethica, all future citations will be to the newer version.
The appropriate citation is G.E. Moore, Principia Ethica, revised edition, Thomas Baldwin, ed.
(Cambridge 1993) [henceforth ‘Principia Ethica’]. 2 Broadly defined, these are theories that hold that hold that goodness is identical with some
natural or metaphysical property. See, e.g., Principia Ethica, § 27. 3 Alisdair MacIntyre holds Moore’s arguments to play a key role in the decline of Anglo-
American ethical thought, and in the ongoing decay of Western culture. See Brian Hutchinson,
G.E. Moore’s Ethical Theory (Cambridge 2001) p. 3, citing Alisdair MacIntyre, After Virtue, 2nd
ed. (University of Notre Dame Press) 1984, pp 14-19. 4 Most notably, by those who rejected Moore’s positive views, like A.J. Ayer and C.L. Stevenson.
2
use moral language, what we intend, what it is we seem to be attributing to an
action when we call it good, and through arguments that appear to demonstrate
that whatever we have in mind when we use moral language, it can not be
anything like what is proposed by ethical naturalists.
II. Overview of Principia Ethica.
Moore begins Principia Ethica by attempting to identify the scope of
ethical inquiry. He wants to answer the questions: What is the subject matter of
ethics? and, What is it that we purport to study when we engage in moral
reasoning? Clearly, the answer is conduct. But there are other disciplines besides
ethics that study conduct, so in order to differentiate ethics from those, Moore
takes his study a step further. What Ethics studies is good conduct. And since
“all conduct is not good; for some is certainly bad, and some may be indifferent,”5
Moore looks to see what all ethical judgments have in common as their subject
and determines that this common subject is goodness. This prompts Moore to
accept that ethics is the “general enquiry into what is good.”6
To determine what the nature of goodness is, in chapter I, Moore examines
what we mean by the word ‘good.’ We use the word ‘good’ in many ways, most
of which have nothing to do with ethics. For instance, I may say, “This eggplant
is really good” or write, “Good soil and high moisture ensure acceptable rates of
5 Principia Ethica, p. 54.
6 Principia Ethica, p. 54.
3
propagation.” In these contexts, ‘good’ means tasty or fertile, clearly not what we
mean when using ethical language.
There is one sense of this word that Moore concludes is the uniquely
moral sense of the word. This is the sense that Moore calls ‘intrinsic goodness’ or
sometimes ‘intrinsic value.’ For simplicity, I will refer to this predicate as
‘goodness,’ throughout this work. When it appears in single quotes, as in
‘goodness’
I am talking about the word that refers to the property goodness. When it appears
in regular text, without quotes, I use it to refer to the property itself. Moore
asserts throughout Principia Ethica that people make a mistake when they attempt
to define ‘good.’ But clearly, the problem is not in defining the word ‘good.’
Linguistic definition is not what Moore is after. He writes, “A definition does
indeed often mean the expressing of one word’s meaning in other words. But this
is not the sort of definition I am asking for. Such a definition can never be of
ultimate importance in any study except lexicography.”7 A linguistic definition
might tell us how people tend to use a word, or it might define a word by
stipulation. One mistake people make is in claiming that a mere verbal definition
is supposed to be an analysis of the uniquely moral sense of ‘good,’ namely, the
property goodness. While it is a mistake to confuse a verbal definition with a
property analysis, Moore is concerned to illustrate and avoid a different mistake—
7 Principia Ethica, p. 58.
4
that of defining the property goodness in terms of other properties, that is, the
mistake of analyzing goodness at all.
In the rest of the first chapter of Principia Ethica, Moore examines the
characteristics of this property. He concludes that goodness is a non-natural,
simple, indefinable property and an intrinsic kind of value. As we will see in
Chapter 2, Moore takes properties to be either simple or complex. An example of
a complex property is being-a-brother. Moore takes this property to be made up
of the simpler properties being-male and being-a-sibling. Perhaps each of these
can be further broken down. To give a definition of a property is to identify its
component properties and their relation to each other. Goodness, claims Moore,
is a property that admits of no such analysis. It is simple, without component
parts. As such, it is incapable of definition.
To say that goodness is an intrinsic kind of value is to say that it is a
property of objects that they possess in virtue of other properties those objects
possess. In particular, goodness is not part of the intrinsic nature of an object x,
but it is dependent on the intrinsic nature (the natural properties) of x. This makes
goodness an intrinsic kind of value. Using the word ‘intrinsic’ to describe both
the natural properties of an object (all those properties which make up any natural
object) and the object’s value (a non-natural property an object has in virtue of its
arrangement of natural properties) is an unfortunate use of language. Moore
clearly means two different things by this word, and leaves it to context for us to
5
determine which.8 This distinction will be made clearer in Chapter 2, but for
now, it is worth noting that goodness depends on the natural properties of objects,
but is not itself a natural property.
Moore also claims in Chapter I that how valuable an object is, i.e., to what
degree goodness attaches to an object, while being determined purely by the
object’s intrinsic properties, is not necessarily the sum of the value of each of the
object’s parts.9 The value of each individual part of an object or state of affairs is
not determinative of the value of the whole. Moore calls such objects ‘organic
wholes’ or ‘organic unities,’ a term borrowed from Hegelian philosophy.
The final task of the first chapter of Principia Ethica, and the task about
which this author is most interested, is to explain why it is that ethical naturalism
fails. Among Moore’s claims is the view that a fallacy is committed by every
philosopher who offers a definition of goodness. He attempts to describe this
naturalistic fallacy in a number of ways, using examples10
and in the context of
arguing against naturalistic and metaphysical theories.11
Moore does not yet
define naturalism, but it is clear from context that naturalists are defining
8 Writing in 1921 or 1922, in the proposed preface to a never-published second edition of
Principia Ethica, Moore clarified his position on the relationship between goodness and the
intrinsic properties of good objects. Using ‘G’ to denote Goodness (to prevent confusion with
non-moral senses of the word ‘good’), he writes: “G is a property which depends only on the
intrinsic nature of the things which possess it” and “Though G thus depends on the intrinsic
properties of things which possess it, and is, in that sense, and intrinsic kind of value, it is yet not
itself an intrinsic property.” Principia Ethica, p. 22. 9 Principia Ethica, p. 79.
10 Throughout Chapter I.
11 In Chapters II-IV.
6
goodness in terms of properties like those which are the subject matter of
disciplines like biology or psychology.12
As part of his proof that goodness is a non-natural property, and (possibly)
in support of his view that all forms of naturalism commit the naturalistic fallacy,
Moore offers his famous open question argument. The gist of this argument is
that for any complex property you consider as a possible definition of goodness, a
little reflection will show that it differs from goodness, and so, is inadequate as a
definition. This is somewhat of a simplification of the argument, and as we will
see in Chapter 5, there is not unanimity on what the open question argument
really is, nor on what it is supposed to prove.
One way to run the open question argument is as a test that uses our
intuitions about the meaning of the words we use to express moral concepts to
force the reflection required to see that any proposed definition fails. For
instance, if the suggestion was that goodness meant happiness-promoting, we
might substitute ‘happiness-promoting’ for ‘good’ in a sentence and see that that
substitution did not preserve the meaning of the original sentence, as in:
[1] Practicing vegetarianism is good
[2] Practicing vegetarianism is happiness promoting.
We can see that ‘happiness promoting’ has a different meaning than ‘good,’
Moore argues, because even if we accept [2], we may intelligibly ask,
12
He provides a working definition of naturalism in Chapter II.
7
Is practicing vegetarianism good?
The fact that we can know [2] is true, while the truth of [1] is an open question—
i.e., that [2] does not make [1] trivial—shows that [1] and [2] have different
meanings. Moore would say that our suspicion that [1] and [2] do not mean the
same thing is grounded in the inadequacy of happiness-promoting as a definition
for goodness.13
Some version of this test and the naturalistic fallacy are Moore’s
most enduring impacts on moral philosophy and the primary subject of this
dissertation.
Chapters II, III and IV of Principia Ethica are devoted to showing that the
various forms of ethical naturalism and metaphysical ethics commit the
naturalistic fallacy and should be rejected. In Chapter II, Moore tells us that what
characterizes all forms of naturalism is that they purport to define goodness in
terms of some natural object (or property) or other. Moore offers two ways of
identifying natural objects and properties. The first, based on our idea of the
proper scope of scientific inquiry, defines natural properties and objects as those
things that are the subject matter of natural sciences and psychology. Biology,
chemistry, physics, and psychology collectively study such properties as being-
productive-of-life, being-alkaline, being-massive, and being-productive-of-
happiness, therefore these are natural properties. The second way Moore defines
natural objects and properties is in terms of metaphysics. He writes that these
13
This is a vastly simplified version of his argument. Moore’s more elaborate version is the
subject of chapter 5 of this work.
8
properties are objects or properties that currently exist, have existed, or will one
day exist. He offers a test to help us make the determination—Can we imagine
that object or property existing alone in the universe? If so, it is a natural object
or property. This isolation test is purported to be especially useful in identifying
natural properties because Moore takes it to be perfectly conceivable that natural
properties are capable of existence in time (and space) independent of the objects
to which they are usually attached. But this is not the case for non-natural
properties. Naturalistic theories of ethics define goodness in terms of one of
these sorts of objects or properties.
All of these theories, including Hedonism, the subject of Chapter III, are
said to rest on the commission of the naturalistic fallacy. Spencer’s theory of
evolutionary ethics commits the fallacy because it identifies the normative
concept being-ethically-better with being-more-evolved. As Moore points out,
being more evolved may be better than being less so, but it may not be. Surely,
they are not one and the same concept.
Hedonists and all those who base their theories on hedonism—the theory
that goodness is pleasure—are also said to commit the naturalistic fallacy. J.S.
Mill is said to do so in virtue of thinking that the fact that pleasure is universally
desired proves that being-productive-of-pleasure is identical with goodness.
Ethical egoists—those who argue that we ought each to pursue our own best
interests—make two mistakes. Those who are hedonistic egoists commit the
9
naturalistic fallacy by identifying goodness with being-productive-of-pleasure.
And all ethical egoists confuse a universal concept, goodness, with something not
universal—the agent’s benefit, i.e., a property that is necessarily different for
everyone. Utilitarianism, the view that what is right is the act that has the best
results (that which maximizes the proper end(s) of moral conduct), is praised as a
general theory, but criticized on the grounds that utilitarians typically choose as
their end either pleasure, which fails for the reasons Moore has already discussed,
or some other natural property, none of which are identical with goodness.
In Chapter IV, Moore turns his attention to metaphysical ethics. These are
theories that purport to define goodness by identifying it with some metaphysical
property or other—Moore calls these supersensible properties. These are
supposed to be real, existent properties that are beyond our comprehension. A
theory might hold that things are good insofar as they mirror this supersensible
reality. Moore rejects these theories on the ground that they commit the
naturalistic fallacy. This may seem odd, as they are not defining goodness with
respect to any natural property, but Moore notes that though the ultimate
definitions are not natural ones, the commission of the fallacy lies not in the
nature of the definition, but in the definition itself.14
14
This may be read as some proof that the naturalistic fallacy is just a definist fallacy—that to
commit the naturalistic fallacy is to give a definition of an indefinable property, goodness. But
when we attend closely to the text, we see that Moore does not say the fault lies in the mere
definition, but in inferring the definition on inappropriate evidence. See, e.g., Moore’s comments
on Green’s theory, Principia Ethica, p. 189.
10
In Chapter V, Moore moves from discussions of meta-ethics—the
discussion of the nature of goodness, and what sorts of things possess goodness—
to a discussion of practical ethics. The proper question for practical ethics is
‘What ought I (we) do?’ Rather than asking about the value of objects and states
of affairs, it seeks to determine the rightness of actions that bring about such
objects or states of affairs. The answer Moore embraces is that our goal is to try
to maximize intrinsic goodness, and to accomplish this, we ought to do those
things that tend to produce better results in the foreseeable future, measured by
the production of intrinsic goodness.
The final chapter of Principia Ethica is where Moore tells us what things
are intrinsically valuable. To determine the value of objects and states of affairs,
Moore again turns to a kind of isolation test. What value would a thing have if it
existed alone in the universe? Something that has only instrumental value—a
thing that is good only as a means to some other thing—clearly has no value in
such a universe. Moore concludes that there are a great many intrinsically good
things, but two stand out as very great goods, namely love of beautiful things and
love of good persons.
Some effects of Principia Ethica
Principia Ethica has had a profound impact on subsequent analytic ethics.
Convinced by Moore’s assault on naturalism, many came to embrace Moore’s
11
non-naturalism. Some, like W.D. Ross, accept non-naturalism, but reject the
centrality of goodness, opting instead to base their theories on the concepts of
rightness or duty.15
Others have taken a more radical approach, denying the view
that moral language has the kind of universal meaning Moore presupposes. These
views, varieties of noncognitivism, hold that the words we use to make moral
judgments, words like ‘good,’ ‘bad’ or ‘right,’ do not denote moral properties, but
are expressions of some one of our mental states or psychological attitudes. Thus
when I say, “Torture is bad,” I am not attributing the property badness to torture,
rather I am merely expressing an attitude I have (disfavor of some sort, depending
on the variety of noncognitivism) toward torture. Of course, some philosophers
maintain that some form of ethical naturalism is correct, but straightforward
identification of goodness with some natural property or other has become more
rare.16
Some of these naturalists are ‘non-reductive’ naturalists who do not
propose to define goodness in terms of any particular natural property at all.
Plan of this work
This dissertation consists of a close reading and explication of Chapters I-
III of Principia Ethica. After one hundred years of commentary, Moore’s views
on the central arguments of these chapters of Principia Ethica are poorly
understood. For instance, Darwall, Gibbard and Railton, in their "Toward Fin de
15
W.D. Ross, The Right and the Good (1930). 16
Though there are still plenty of straightforward hedonists.
12
Siecle Ethics"17
claim that fifty years ago, it was proved that there was no fallacy
involved in Moore’s naturalistic fallacy18
. At the same time, others criticize
theories like those of evolutionary ethics on the grounds that they commit the
naturalistic fallacy. What a curious thing it must be if it is no fallacy at all, yet the
reason to reject these theories. Moore, himself, claims not to know what he meant
by his claims that all forms of ethical naturalism commit the fallacy.19
It also
seems that critics of the naturalistic fallacy and open question argument generally
criticize Moore’s arguments without being sufficiently careful to distinguish the
open question argument from Moore’s claims about the naturalistic fallacy. One
would hope that after a century of scholarship we would be able to distinguish
theses central tenets of Moorean thought.20
My purpose in writing this dissertation is to offer an interpretation of
Moore’s central claims that demonstrates the subtlety and power of those
arguments. This involves three things:
1) Offering an explication of what it means to commit the naturalistic
fallacy that is consistent with Moore’s text and examples,
2) Distinguishing Moore’s claims about the naturalistic fallacy from the
open question argument, and
17
The Philosophical Review, Vol. 101, No. 1 (January 1992) 18
Probably referring to Frankena’s article “The Naturalistic Fallacy,” Mind 48: 464-77. 19
In his posthumously published preface to the unpublished second edition of Principia Ethica. 20
As we shall see, the failure is not to be laid squarely at the feet of the critics, as Moore is
difficult to interpret, even for himself.
13
3) Correcting mistaken interpretations of the open question argument.
The heavy lifting begins in Chapter 2. There I describe the property goodness and
explain Moore’s claims about it. I focus on what it means for goodness to be
simple and indefinable.
In Chapter 3, I examine interpretations of the naturalistic fallacy offered
by Thomas Baldwin, A.N. Prior, and William Frankena. I also include some
comments made by Moore in the years after Principia Ethica was published. I
find that none of those interpretations, even Moore’s, maintain fidelity to the text
of Principia Ethica. The mistakes, however, are instructive. We will see that
Prior, for instance, conflates the naturalistic fallacy and the open question
argument. The consequences of this are dire, for it allows Prior to minimize both
claims by attacking only one.
In Chapter 4, I offer a close reading of Moore’s statements wherein he
describes the naturalistic fallacy and of his examples of philosophers who commit
the naturalistic fallacy. I find that to commit the naturalistic fallacy is to make
one of three closely related errors. My interpretation of the naturalistic fallacy
suggests that while it is indeed an error, it is not the error that most critics think it
is.
In Chapter 5, I provide a reading of Moore’s open question argument that,
as in the case of the naturalistic fallacy, emphasizes fidelity to the text and rejects
the traditional interpretation. I find that, rather than proving that all natural
14
definitions of goodness are false, the section of Principia Ethica wherein Moore
details the open question argument actually contains many arguments all in
support of the conclusion that goodness is simple and indefinable. I examine
these arguments in detail and offer one of my own that is intended to overcome
some of the problems normally associated with Moore’s arguments.
Chapter 6, the conclusion, summarizes my main points, and emphasizes
the proper role of Moore’s two major arguments of Principia Ethica, Chapter I. I
find that the open question argument is Moore’s proof that goodness cannot be
defined in terms of any complex natural property. But this leaves the possibility
that goodness is a simple natural property. The naturalistic fallacy is then
deployed against any possible simple definition. I argue that the naturalistic
fallacy does not disprove any proposed natural definition, but that commission of
the naturalistic fallacy undermines the rational support any naturalistic theory
could hope to garner.
15
Chapter 2
Goodness, Simple and Indefinable
It is immediately obvious that when we see a thing to be good, its
goodness is not a property which we can take up in our hands, or
separate from it even by the most delicate scientific instruments,
and transfer to something else. It is not, in fact, like most of the
predicates which we ascribe to things, a part of the thing to which
we ascribe it.1
The main theses Moore defends in Principia Ethica are that goodness is a
simple and indefinable property, and that most naturalistic ethical theories commit
the naturalistic fallacy. Chapters 3 and 4 are about the naturalistic fallacy, and in
Chapter 5, I explain Moore’s argument for the indefinability of goodness. But
this does not tell us much about what goodness is. What features does this
property have? What does it mean for a property to be indefinable? This chapter
begins with an examination into what property it is that Moore claims is simple
and indefinable. This is followed by an examination into what it means to be
simple and indefinable.
I. Goodness
Moore’s claims about the naturalistic fallacy aside, perhaps the most
significant thesis he defends in Principia Ethica is the claim that goodness is
1 Principia Ethica, p. 175.
16
simple and indefinable. Of what property is Moore predicating indefinability?
Moore begins his examination of moral theory by asking the question What is
good?2 He does not mean by this question ‘What things are good?’ but rather, he
is inquiring about the property, goodness, itself. By ‘good,’ he means to refer to
that property that he believes is unique to ethics, and upon which all judgments of
right and wrong rest, and not to the many other properties that go by the same
name but which are not relevant to ethics. For example, if I say, “Soup is good
food,” the word ‘good’ does not refer to the property relevant to ethics. I am
likely using the word to refer to some property like tastiness, or being-valued-for-
its-nutritional-content. The question of what things are good is an interesting one,
but before we can do a good job of answering it, we must first figure out what it
means to be good. To what does the word ‘good’ when used in its moral sense
refer? In other words, what property does the word ‘good’ pick out?
By ‘good,’ Moore means “the only simple object of thought which is
peculiar to Ethics.”3 At various times in his career, Moore believed that the other
concepts of ethics—examples include right, obligation, and virtue4—could be
2 Principia Ethica, p. 55.
3 Moore, Principia Ethica, p. 57. This quote raises a few issues worth footnoting. First, Moore
notes that goodness is not the only such property, but its converse, badness, is another. Together
they are the only simple objects of thought peculiar to Ethics. Second, by ‘object of thought’ he
actually means what we mean by ‘property.’ He takes ‘property,’ ‘notion,’ ‘concept,’ ‘object of
thought,’ ‘quality,’ and ‘predicate’ to mean the same thing, at least early in his philosophical
career. In Principia Ethica, Moore no longer uses ‘concept’ to denote objects and properties. As
the vocabulary and conventions of analytic philosophy became standardized, his use of these
words became standardized as well. 4 Note, Moore holds that virtue is not “an unique ethical predicate.” Principia Ethica, p. 231.
17
defined, at least in terms of goodness. For example, in Principia Ethica, he offers
the following analysis of obligation:
‘S is obligated to do x’ means ‘x will produce the greatest amount
of good in the universe’5
But it is not clear whether Moore believed these other ethical properties
could be defined in terms of goodness throughout his career. In Ethics, for
instance, he does not give a definition of rightness in terms of goodness.6 It is
unclear whether this warrants concluding that he abandoned hopes of such a
definition, or whether he merely intended to distance his later work from the
controversy associated with Principia Ethica. In a letter to Russell, written in
1905, Moore writes in response to Russell’s article “On Denoting”7
“Your theory suggested to me a new theory about the meaning of
the ‘meaning’ of words, which might, I think, help us to settle the
question about whether the meaning of ‘right’ is indefinable. But I
can’t work it out right now.”8
Clearly, in 1905, Moore was unsettled about the definability of ‘right.’ Since we
have no evidence that Moore had, by 1905, changed his views about language, it
5 Principia Ethica, p. 198. We recognize that this is an incomplete analysis, because the right side
does not seem to require anything of S. Assuming that ‘ought’ implies ‘can,’ we might solve this
by the addition of ‘and x is among the things S can do.’ Moore provides an analysis of rightness in
the same passage. 6 Moore, Ethics (1911) p 61.
7 Russell, B., “On Denoting,” Mind New Series, Vol. 14, No. 56 (Oct., 1905), pp. 479-493.
8 “Letter to Bertrand Russell of 23 October 1905,” reprinted in part in Philip Pettit, “The Early
Philosophy of G.E. Moore,” The Philosophical Forum, vol. IV, no. 2 (Winter 1972-73), pp. 260-
298, at p. 298.
18
is clear that he was unsure whether rightness was definable (analyzable). We also
have inconclusive evidence from Moore’s reply to his critics in Schilpp. In that
essay, he defends, for the most part, the project of defining ethical properties in
terms of goodness, but in a telling passage, he admits that he is sometimes
inclined to think that no moral terms are names of properties at all, but only carry
emotive meanings.9
Regardless of whether goodness is really the only simple object of thought
peculiar to ethics, or whether some or all of the other concepts are simple as well,
or definable in terms of goodness, it is clear we have ways of thinking about
goodness and these other properties. Despite not being able to define goodness, it
is possible to recognize it for purposes of discussion.10
Similarly, basic properties
of physics—mass, for instance—may not be defined in terms of their constituent
parts, but only by relation to other properties or objects. Nonetheless, such
properties may be identified and discussed. We may think of Moore’s claim
about goodness as the analog to a physicist’s claim about mass. It is a basic
9 He writes: “I must say again that I am inclined to think that ‘right,’ in all ethical uses, and, of
course, ‘wrong,’ ‘ought,’ ‘duty’ also, are, in this more radical sense, not the names of
characteristics at all, that they have merely ‘emotive meaning’ and no ‘cognitive meaning’ at all:
and, if this is true of them, it must also be true of ‘good,’ in the sense I have been most concerned
with. I am inclined to think that this is so, but I am also inclined to think it is not so; and I do not
know which way I am inclined most strongly.” Moore, “A Reply to My Critics,” p. 554 in P.A.
Schilpp, ed., The Philosophy of G.E. Moore, The Library of Living Philosophers collection, vol.
IV, Northwestern University Press (1942). 10
This very feature will be the motivating force behind Moore’s argument that ‘good’ is not
meaningless, i.e., that there is a property that we refer to by the name ‘good.’ This argument is the
second part of Moore’s proof that goodness is simple and indefinable, the subject of Chapter 5.
19
property, we all know it is there, we recognize it, we can discuss it, but we cannot
define it.
Since Moore takes this property to be indefinable, he cannot tell us much
more than this. He can only rely on the fact that we are familiar with the property
in question, that we use it in our judgments about ethics, and that when we reflect
on it, we will recognize it. This is not necessarily problematic. Many disciplines
have core features that we may not be able to define. Nonetheless, we can talk
meaningfully about them. To use an analogy Moore comes back to frequently,
we can talk about yellowness even though I cannot define it. It is enough that we
can identify instances of it when we see it, etc.11
Moore writes “Whenever
[someone] thinks of ‘intrinsic value,’ or ‘intrinsic worth,’ or says that a thing
‘ought to exist,’ he has before his mind the unique object—the unique property of
things—which I mean by ‘good.’”12
Although in this dissertation I have avoided citing to Moore’s later work
except to contrast it with what appears in Principia Ethica, I must make an
exception here. Since Moore sometimes refers to this property as ‘intrinsic
11
Actually, though this is not anything I wanted to talk about in this dissertation, it strikes me that
we may very well be wrong about the features that we think allow us to pick out goodness for
purposes of discussion. 12
Moore, Principia Ethica, p. 68. We see in this quote one of the problems endemic to Moore’s
writing in Principia Ethica. Though it seems that he might be referring to the property being
intrinsically valuable and not the words ‘intrinsic value’ he offsets the phrase with single quotes.
As I have noted elsewhere, this can lead to use/mention problems, some of which Moore noted
himself in his Preface to the Second Edition. On the other hand, he is trying to express the thought
that these are other names for the property goodness. This suggests that they should be offset by
quotes, but then the quote ought to read “Whenever [someone] thinks of the property named
‘intrinsic value’. . . .”
20
value,’ which brings to mind features of goodness that we might not otherwise
think of, Moore’s comments about intrinsic goodness from the Second Preface are
helpful. Moore identifies two important features of intrinsic goodness:
1) Intrinsic goodness is a property which depends only on the intrinsic
nature of the things which possess it, and
2) Though it depends on the intrinsic properties of the things which
possess it, it is not itself an intrinsic property.13
Moore sometimes writes as if (1) implies that if an object has intrinsic
goodness, it has it necessarily, but this implication is not literally true. In
particular, he makes two claims that seem to imply necessity. First, he writes that
judgments that “state that certain kinds of things are themselves good” are, if true,
universally true.14
Secondly, he writes that “a judgment which asserts that a thing
is good in itself . . . if true of one instance of the thing, is necessarily true of all.”15
By ‘necessarily true’ Moore does not mean true in every possible world. Clearly
there are worlds in which a given object, good in this world, would have intrinsic
properties that make it bad, just as my counterpart in Bizarro world might have no
beard and a malevolent streak a mile wide. What Moore means is that in any
world in which an object, good in this world, exists with exactly the same intrinsic
properties, it will be good regardless of the other characteristics of the world.
13
Preface to the Second Edition of Principia Ethica, p. 22. He believes that only predicates of
value have these two features. 14
Principia Ethica, p. 75. 15
Principia Ethica, p. 78.
21
Whatever is good depends only on the non-relational properties of the object such
that for any object, no matter what the rest of the world is like, if an object is good
in one world, its exactly similar counterpart will be good in every world.
It seems that Moore is trying to establish that goodness is a kind of
emergent property, or that it supervenes on the intrinsic properties of the objects
that possess it. His descriptions seem to imply that the supervenience relation is a
strong, local form of supervenience.
We can state this view as follows: For any objects x and y, and any
worlds, w1 and w2, if x in w1 is indiscernible from y in w2 with respect to their
intrinsic properties, then x in w1 is indiscernible from y in w2 with respect to
goodness. On this view,16
the indiscernibility is at the level of individuals, not
whole worlds.
This is a variety of strong supervenience because the indiscernibility with
respect to goodness is across worlds. A weak supervenience view might claim
that if two objects were indiscernible with respect to their natural (or other)
properties in a given possible world, they would be indiscernible with respect to
goodness in that world. On Moore’s view, if an object is good anywhere, an
object exactly like it is good everywhere, whatever else may be true of that world.
The view is a variety of local supervenience because the moral supervenes on the
"local" properties of the objects in question. Relational properties and other non-
16
Unlike, say, a global supervenience view.
22
intrinsic properties do not make up part of the supervenience base.
Moore also believes that most (maybe all) natural and metaphysical
predicates are either contingent or intrinsic. Intrinsic goodness is neither, so it
cannot be natural or metaphysical. We might create from this an argument that
goodness is not definable in terms of natural or metaphysical properties, and
while this would go a long way toward establishing Moore’s claim that goodness
is indefinable, it is not an argument Moore makes.
II. Goodness, simple and indefinable
Simplicity
Moore takes goodness to be simple and indefinable. For Moore, the
indefinability of goodness is a consequence of its being simple. To be a simple
property is to have no other properties as constituents. Complex properties, by
contrast, have such parts. Moorean definition consists in identifying those parts
and their relations to each other and the whole. This conception of properties is
examined briefly, below.
The simplicity of a property and its indefinability are distinct features the
property may have, though one is the consequence of the other.17
And the reason
17
Some doubt is cast on this claim by Moore, himself, in the Second Preface, where he writes that
in Principia Ethica he identified being simple with being indefinable. I do not find this
identification. To be simple is to admit of no definition, but given Moore’s conception of
definition, it appears that indefinability is a consequence of not having parts, rather than being
identical with simplicity.
23
I take this to be so important is that no part of Moore’s proof that goodness is
simple and indefinable actually turns on the indefinability. What Moore does is to
present an argument from the premise that either goodness is simple (hence,
indefinable), or it is complex, or it has no unique meaning. We will see that
Moore does not show that no definition is correct.18
In fact, he only rules out
definitions that identify goodness with some complex property. He then argues
that ‘good’ is not meaningless—that there is some unique property we have in
mind when we use the word ‘good’ in the moral sense. What he does not do, and
this point has been missed by every author I have read, is present in this argument
any sub-argument to the conclusion that no simple natural property, like
pleasantness, is identical with goodness.19
Simplicity has some implications worth drawing, besides the
consequences for definability. To see this, we can consider the property
yellowness.20
Moore uses the analogy between goodness and yellow when he
describes the naturalistic fallacy and to give an example of a natural simple
property.
18
It is typically asserted by those who claim that Moore’s naturalistic fallacy is a definist fallacy
that Moore’s open question argument is designed to show that no definition is correct. But he
emphatically does not show this. Rather, he shows that Goodness is simple. From that simpleness
we might infer that Goodness is indefinable. But we could resist this inference. 19
Strictly speaking, identifying goodness and pleasantness is not to define either, at least in
Moorean terms. 20
Scott Soames, in Philosophical Analysis in the Twentieth Century (Princeton University Press
2003) does a nice job explaining this analogy.
24
Consider the statement ‘Lemons are yellow.’ It is perfectly reasonable to
assert or to believe this statement. Further, the statement does not assert that
lemons are necessarily yellow. We can conceive of an orange or green lemon—it
seems reasonable to think there could be such things, or a world in which such
things existed. And the truth of the statement does not follow from some deeper
necessary truth about lemons, as it might if lemons were defined as “the yellow
citrus fruit,” in which case, we would claim that the expression is analytic. So
Moore would claim that the statement “lemons are yellow’ is synthetic. This is
analogous to statements about good, such as:
Pleasure is good.21
We can hold that such a statement is true without holding that the relation
between pleasure and goodness follows from the definitions of either pleasure or
goodness.22
So, again, Moore would claim that this expression is synthetic.23
If someone were to attempt to define yellowness, by claiming that to be
yellow is to reflect light waves of wavelength Y, we might also take this to be a
synthetic statement. This is because from our knowledge of yellowness, nothing
21
Where the statement is taken as predicating goodness of pleasure. That is, that pleasure is a
thing that is good. 22
Moore takes pleasure to be a simple property as well, so no definition is possible for it. 23
He writes, “[P]ropositions about the good are all of them synthetic and never analytic.”
Principia Ethica, p 58. I think this might be an error on Moore’s part. If a property or object were
defined in terms of goodness, such as “to be a woozle is to be a morally good whatzit,” then the
statement “Woozles are good” would be analytic, and would be about the good, because woozles
would, by definition, be part of the good. Moore might have done better to say instead
“Propositions purporting to define the good (statements purporting to tell us all and only the good
things) are synthetic.” So, “pleasure is good” may or may not be synthetic, depending on the
definition of ‘pleasure,’ but “pleasure is the good” is synthetic (and, Moore would add, false).
25
seems to follow about Y-ish light wavelengths. This is to say that it is not part of
our concept of yellow (and in fact, it is not part of the property yellowness) that
yellow things reflect any particular kind of light. As in the earlier case, we could
imagine a situation in which yellow things reflected wavelengths of a different
size, which suggests a certain contingency about the proposed definition. Thus, it
is synthetic, and not a successful definition.
The reason that goodness and yellowness resist definition in this way is
that they are both simple, and hence, unanalyzable properties. Hence, everything
we might say about them is synthetic.
Analysis
Moore takes this to be so because he takes these properties to be
unanalyzable. Of the many possible kinds of definition,24
Moore is interested in
the kind of definition we might call a property analysis. In Moorean terms, a
successful definition of a property is an analysis that tells us the parts that make
up the object, and perhaps, the relation that those parts stand in to each other and
the thing as a whole. As a consequence of this view of definition, simple
properties—those properties that have no parts—are incapable of being defined.25
Moore writes “definitions which describe the real nature of the object or notion
24
John Fischer, in his paper “On Defining Good,” The Journal of Philosophy, Vol. 51 (1954), pp.
730-736, identifies sixteen different sorts of definition. His list is not considered authoritative. 25
Principia Ethica, p. 59.
26
denoted by a word, and which do not merely tell us what the word is used to
mean, are only possible when the object or notion in question is something
complex.”26
Complex objects or properties, real or mythical, are subject to this
sort of definition. Moore illustrates with the examples of a horse and a chimaera.
Each of these can be analyzed into its component parts, and those parts might be
further analyzable as well, until you get to the level of simples, at which no
further analysis is possible.27
Mary Warnock, in Ethics Since 1900,28
comments on Moore’s contention
that the definition of ‘Horse’ is ‘hooved quadruped of the genus equus.’ This
contention is a source of concern for Warnock because Moore asserts it on several
occasions as an aid to clarify what he means by ‘definition.’ But the result, for
Warnock, is not clarity at all. It is evident that what Moore means by ‘definition’
is analysis, but the definition he offers for ‘horse’ does not look much like an
analysis to Warnock. She writes “No doubt one could list the parts of the horse,
and even state their relation one to another, but to do this would not be to say
‘Hooved quadruped etc.’, with a special meaning; nor would it be taken to be
defining ‘horse’ at all.”29
What Moore expects of a good analysis is that it makes
clear the parts which compose a thing and the relation in which they stand to each
26
Ibid. 27
Principia Ethica, p. 60. 28
Ethics Since 1900, 3rd
ed., (Oxford University Press 1978). 29
Warnock, p. 21.
27
other. Warnock thinks this would not be a definition. An analysis of ‘horse’, she
thinks, simply would not contain pasterns, chestnuts, hooves and tails.
There seem to be two sources of concern underlying Warnock’s criticism.
The first might be that Moore has not given us a single complete and successful
analysis against which we could measure any future analyses we might make.30
With his examples of horses and chimaeras, he gestures in the direction he thinks
a successful analysis lies, but fails to offer one. He says of a chimaera that we can
describe its parts—a snake, a goat, and a lioness—and how it is put together. And
we can further divide each of its parts into parts, because snakes, goats, and
lionesses are, themselves, complex things. “[A]ll are composed of parts, which
may, themselves, in the first instance, be capable of similar division, but which
must in the end be reducible to simplest parts, which no longer can be defined.”31
In some ways, his choice of examples leaves much to be desired. The examples
of horse and chimaera, while illustrating the possibility for defining both concrete
and imaginary entities, are simply too complex to be useful. Additionally, the
lend themselves to a good deal of confusion about what we are actually defining.
And this, it seems, is the second concern underlying Warnock’s complaint.
Are we to consider these definitions of concrete entities or the properties
being a horse and being a chimaera? Moore wants to assert that a certain kind of
30
He gives such an analysis of beauty in Chapter VI, but it looks nothing like an analysis in terms
of parts, so it is not particularly illuminating. 31
Principia Ethica, p. 59.
28
definition—a property analysis—is impossible for goodness, but the examples he
gives of property analyses look like analyses of concrete objects. He divides the
horse and the chimaera into pieces, it appears. This seems like a mistake. The
idea of property analysis as demonstrating the component parts of complex
properties is a defensible project, but I think it looks very little like these
particular examples.
There are some things that can be said in defense of the view that there are
complex properties and that they are made up of other, possibly simple
properties.32
Consider the property being a bachelor. It is reasonable to think that
this property is a conjunction of other properties, viz., being unmarried, being
adult, being male, and perhaps (to eliminate a problem raised by Richard Brandt
that I discuss in Chapter 5) being un-widowed and un-divorced. I embrace this
view of properties because I find the alternative unreasonable. The alternative is
that being a bachelor has no parts and is thus, simple. But then, instantiating that
property is not to instantiate all of its parts. But it seems for all the world like
being a bachelor does mean to be unmarried, and male, and adult, etc.33
I think it
would be a weird sort of world if my being a husband was something beyond my
being married and being male. It would have to be a whole other simple property
I instantiated. I am unsure whether my unease with such a view is that it leads to
32
See, e.g., Jeffrey King, “What is a Philosophical Analysis,” Philosophical Studies 90: 155–179,
1998. 33
Could it be a mere accident that people instantiating all those properties also instantiate
bachelorhood?
29
a weird promiscuity of properties34
or that new property instantiations seem to
bring with them no new information about the object that instantiates them. At
any rate, these considerations suggest an initial plausibility to the kind of view
Moore must advocate.
If the idea of complex properties being composed of component properties
has some plausibility, Moore’s conception of definition as property analysis can
be motivated. The remainder of this chapter considers some requirements for a
successful analysis.
Nature of Analysis
Moore claims that ‘good’ is indefinable. Moore meant by this that the
concept to which the word ‘good’ refers, when used in the sense that is relevant to
ethics, is unanalyzable. In Principia Ethica, Moore offers little explanation of
what he really intended for a successful analysis. He writes that good is
indefinable in that “…it is not composed of any parts, which we can substitute for
it in our minds when we are thinking of it”35
and “…definitions which describe
the real nature of the object or notion denoted by a word…are only possible when
the object or notion involved is complex.”36
This gives us some idea of what he
expects from an analysis—that it only applies to things with parts and that it
34
This is something that does not bother me for other metaphysical entities, like events, or objects,
so why would it be strange here? 35
Principia Ethica, p. 60. 36
Principia Ethica, p. 59.
30
should tell us what those parts are in a way that will enable use to substitute them
for the object when we have it in mind.
In his subsequent writing, for instance in his “Reply to my Critics” and in
his “Preface to the Second Edition of Principia Ethica” he gives a more thorough
treatment of the topic. What follows is an explication of successful analysis as
derived from these sources. Consider a sample analysis of x,
x iff y
Calling x the analysandum and y the analysans, Moore sets the following
criteria for a successful analysis:
1. Both analysans and analysandum must be concepts, ideas or
propositions, but not “mere” verbal expressions.37
This is to ensure that we are clear that we are doing conceptual analysis
and not linguistic analysis.38
The features of language, while interesting to
Moore, are not what he is out to discover. He writes “To define a concept is the
same thing as to give an analysis of it; but to define a word is neither the same
things as to give an analysis of that word, nor the same thing as to give an analysis
of any concept.”39
To be sure, in conceptual analysis, the analysandum and
37
Moore, “Reply to My Critics” in Schilpp, P.A., The Philosophy of G.E. Moore, Northwestern
University Press, Chicago, 1942 , p. 662 38
Moore uses the words ‘concept’ and ‘property’ interchangeably. To do a conceptual analysis is
not to analyze our way of thinking about a thing. It is to analyze the thing itself. 39
Moore, 1942, p 664
31
analysans are expressed verbally, but what is being analyzed is the concept, not its
expression—the shape of the letters, spacing, syntax, etc.
2. Both analysans and analysandum must in some sense be the same
concept.40
The inclusion of ‘in some sense’ is terribly unsatisfying, but it might be
the saving grace of this condition. The problem is that when you are doing
conceptual analysis, the concept you are analyzing cannot be the same concept as
the analysandum. This would make the analysis useless. We usually think that
the analysans and analysandum would need to refer to the same object or property
in the world, but they should not be the same concept. This criticism, however is
a consequence of having been trained in post-Moorean philosophy of language
and metaphysics. Contemporary philosophers use the word ‘concept’ to denote a
way of thinking about a thing. Moore means something else. Philip Pettit writes
that for Moore everything that exists, and everything that is known—including
objects of perception—is composed of concepts.41
Similarly, words are
meaningful because they stand for objects, so the object for which a word stands
is a concept. Given this interpretation of ‘concept’, condition (2) seems perfectly
reasonable. The analysans and analysandum of a successful analysis should refer
to the same object, but in different terms. Consider the analysis of brother:
40
Moore, 1942, p. 666. 41
Pettit, P, “The Early Philosophy of G.E. Moore”, in The Philosophical Forum, vol. IV, No. 2,
Boston University 1972, p. 267 Existing things “can, if anything is to be true of them, be
composed of nothing but concepts.” Moore, “Nature of Judgment,” Mind, (1899) p. 182.
32
x is a brother iff x is a male sibling
‘Brother’ and ‘male sibling’ both refer to the same objects (Moore would
say ‘concepts’) in the world.
3. The analysans must be such that nobody can know or verify that the
analysandum applies to an object without knowing that the analysans does also.42
This condition is troublesome to me. This, I believe, is the source of
Moore’s paradox of analysis. An analysis is not successful unless it meets this
condition, but this condition keeps such an analysis from being informative. It
can offer us no new information. Consider the folk philosophical analysis of heat.
This analysis, thought to be correct and successful, might be worded something
like this:
x is hot iff x has rapidly moving molecules.
Clearly, one might know that some object has rapidly moving molecules
without knowing that that object was hot. It is conceivable that one might have a
sophisticated molecular motion detector and not know that it is also a heat
detector. In fact, anyone unfamiliar with the analysis could know one side of the
bi-conditional applied without knowing that the other applied as well.
If Moore is actually giving an account of this property:
x is a successful analysis of y for S(someone),
then (3) might be a good condition.
42
Moore, 1942, p. 663.
33
4. The expression used for the analysans must be synonymous with that
used for the analysandum.43
This particular condition is one that gave Moore some difficulty. He
thought it was necessary for any successful analysis but he also thought that it was
not satisfiable. This is because he is not sure what criteria to use for synonymy.
If synonymy of expressions means simply that two expressions refer to the same
object, then this condition is clearly correct. But he seems to want something
more of synonymy. In his “Reply to My Critics” he writes:
“It is obvious…that, in a sense, the expression ‘x is a brother’ is
not synonymous with, has not the same meaning as ‘x is a male
sibling,’ since if you were to translate the French word ‘frère’ by
the expression ‘male sibling’ your translation would be incorrect,
whereas if you were to translate it by ‘brother’ it would not”44
This suggests an alternate criterion for synonymy, which, I believe, ought
to be rejected. Issues of correctness in translation (particularly this translation)
are as much issues of style as they are of content. I think the proper response to
Moore’s concern about translation is one of two things: either reject translational
substitutivity as a criterion for synonymy, or reject his assertion that this is a bad
translation. Stylistically it is awkward, but in terms of content it is certainly
acceptable. Another possibility is to simply reject condition (4).
43
Ibid p. 663. 44
Ibid, p. 667.
34
5. The expression used for the analysandum must be different from that
used for the analysans in two ways:
a) The analysans must mention explicitly concepts not used in the
analysandum.
b) The analysans must mention the method by which the concepts are
combined.45
For example, ‘being male’ and ‘being a sibling’ are not mentioned by
‘brother’ and the method by which the concepts male and sibling are combined is
conjunction.
Ultimately, Moore’s claim is that there is no analysis of intrinsic goodness
that satisfies (1-5).46
For Moore, good is unanalyzable because it is simple.
Simple concepts have no parts, unlike brother, which has other concepts as
constituents. Since it has no parts, there are no other concepts, which can be
combined in any way, which could stand as the referents of the analysans.
45
Ibid, p. 666. 46
There is nothing like unanimity on the requirements for a successful analysis. For Moorean
ethics purposes, the part/whole discussion may be enough.
35
Chapter 3
What is the Naturalistic Fallacy?
“[F]ar too many philosophers have thought that when they named [the other
properties that good things have], they were actually defining good; that these
properties, in fact, were simply not ‘other’ but absolutely and entirely the same
with goodness. This view I propose to call the ‘naturalistic fallacy.’1
Introduction
Perhaps the most enduring impact of Principia Ethica is the addition of
the expression ‘the naturalistic fallacy’ into the lexicon of analytic ethics. It is the
one contribution of Principia Ethica most recognized and relied upon by other
philosophers, and it is the one holdover of Moore’s arguments that even non-
philosophers embrace.2 Those precious few pages of Principia Ethica in which
Moore describes the naturalistic fallacy have received a great deal of
commentary.3 None of this commentary, however, illuminates why the
naturalistic fallacy is held in such high regard by some and completely dismissed
1 Principia Ethica, p. 62.
2 Though the embrace is hesitant, and perhaps naive. My own informal observations suggest that
most non-philosophers take the naturalistic fallacy to be the attempt to derive an ‘ought’ from an
‘is’—the undertaking of attempting to reach moral conclusions from scientific, or other,
observations. See, e.g., Ursula Goodenaugh, “The Holes in Gould's Semipermeable Membrane
Between Science and Religion,” American Scientist, May 1999, wherein she reviews Stephen J.
Gould’s Rocks of Ages (Ballantine Books, 1999) Goodenaugh writes that “the book attempts to
make a . . . point, namely, that scientists are prone to commit the naturalistic fallacy—to derive
‘oughts’ from ‘ises’—on a grand, overbearing scale.” 3 In addition to the commentary I examine in this chapter, there are a vast number of papers and
book chapters devoted to it. See, e.g., Warnock, Mary, 1978, Ethics Since 1900, 3rd
ed., Oxford:
Oxford University Press, Ch. 1; Lewy, Casimir, “G.E. Moore on the Naturalistic Fallacy,”
Proceedings of the British Academy, Vol 50, (1965) pp 251-262; and Kolnai, Aurel, The Ghost of
the Naturalistic Fallacy, Philosophy, Vol. 55, No. 211. (Jan., 1980), pp. 5-16.
36
by others. This is due, I believe, to a failure by commentators to tease out from
Moore’s somewhat cryptic text a coherent reconstruction of what committing the
naturalistic fallacy entails and what role the fallacy plays in Moore’s overall
project.4
Moore states that nearly all philosophers who advocate naturalistic
ethics—those theories that hold that good or some other moral property is
identical with some natural property or properties—and all, or nearly all,
philosophers who advocate metaphysical ethics—those theories that hold that
ethical truths follow from metaphysical truths5—commit the naturalistic fallacy.
6
Despite one hundred years of philosophical work on the subject, it is not clear
what is being asserted when a theory is said to commit the naturalistic fallacy.
Typically, commentators conclude that Moore’s claim that naturalistic theories
commit the naturalistic fallacy simply amounts to the claim that such theories
assert that good is identical with some natural property, or that Moore’s
comments are so hopelessly confused that his claim is meaningless. For example,
Tom Regan writes, “The ‘naturalistic fallacy’ is the name Moore gives to any
4 For instance, Thomas Baldwin, whose work is discussed in detail, below, writes that it is
Moore’s central claim in Principia Ethica that the naturalistic fallacy is committed by almost all
moral philosophers. I contend that, while this is a claim of Moore’s, it is far from his central
claim. See Baldwin, T., G.E. Moore, Rutledge (London 1990) p. 69. 5 He is not using the term ‘metaphysical’ as we do today. By a ‘metaphysical truth’ he means
some proposition about the metaphysical (i.e., part of the super-sensible) realm. See Principia
Ethica, p. 161. 6 See, e.g., Principia Ethica, pp. 90-93 and 176.
37
attempt to identify Good with something other than itself.”7 Regan, who
provides a thorough treatment of the impact of Moore’s ethical philosophy and
the metaphysics underlying his claims, opts not to analyze just what the
naturalistic fallacy is, beyond this simple assertion. The two most significant
claims Moore makes—that most naturalists commit the naturalistic fallacy and
that goodness is simple and indefinable—are left unexamined. So, even though
Regan points out that nearly two-thirds of Principia Ethica is devoted to the
explanation of the naturalistic fallacy, and finding instances of it in others’ work,8
Regan sidesteps the very issue of what it means to commit the naturalistic fallacy.
I offer that this is because Moore’s text is brutally hard to interpret and subject to
a variety of challenges, including those coming from Moore himself. What error
do naturalistic theories (or those philosophers who assert their truth) commit? Is
it correct to call it a fallacy? In this chapter I examine answers to these questions
as proposed by Thomas Baldwin, Arthur Prior, and William Frankena.
Before examining this secondary literature, it is worth reviewing Moore’s
goals for Principia Ethica and, in light of these, identifying some of the confusion
surrounding the naturalistic fallacy. As we shall see, Moore’s own subsequent
commentary on the Principia Ethica is the root of some of the confusion.
Moore has as his primary purpose in Principia Ethica the scientization of
ethics. In order to establish a science of ethics on a par with the natural
7 Tom Regan, Bloomsbury’s Prophet (Temple University Press 1986), p. 193.
8 T. Regan, p. 197.
38
sciences—which had seen substantial success by the time Moore wrote Principia
Ethica—Moore believes it is necessary to establish precisely what the proper
subject of ethics is. As noted in the first chapter, he takes the proper subject of
ethics to be goodness.
Goodness, unlike some of the properties that are the proper subject of the
natural sciences, is a property about which there is nothing like unanimity in the
views about its nature. The paradigms of sound reasoning—mathematics and the
natural sciences—might face this problem, especially at the peripheries of those
sciences,9 but not about their central concepts; there is something like unanimity
about what angles are or what it is that chemistry studies. And this is no surprise.
It takes a certain solidification about central concepts before advancement is
possible. In keeping with goal of solidifying the central concepts of ethics, a big
part of Moore’s work is to identify what goodness is.
One might infer from this that to identify what goodness is, it may help to
determine what it is not. Disabusing fellow philosophers of their false views
about the subject matter of ethics would go a long way toward establishing a core
science of ethics. One way to view Moore’s writing on the naturalistic fallacy,
then, is that it is an attempt to determine what it is that goodness is not.
9 For example, there is debate about the nature of fields in physics. See, e.g., Auyang, S. Y., How
is Quantum Field Theory Possible? , (New York: Oxford University Press 1995) cited in Michael
Esfeld, “Holism in Cartesianism and in Today’s Philosophy of Physics,” Journal for General
Philosophy of Science / Zeitschrift für allgemeine Wissenschaftstheorie 30 (1999), pp. 17–36
See Frankena p 468, fn 1, citing to Whittaker, The Theory of Abstract Entities, pp 19ff. 84
Frankena, p 468. 85
The starting point for Mill’s disastrous analogy. I think a far better example of an argument that
purports to violate Hume’s law would have nothing in it that was derived from a definition of
good, as is the suppressed premise Frankena will eventually supply. An argument that proceeded
from the pain of childbirth to the conclusion that we ought to stop reproducing would probably be
truer to the spirit of What Hume was railing against.
78
of the argument is “pleasure is good since it is sought by all men.”86
Note first
that this argument’s conclusion, that pleasure is good, is ambiguous in that it may
mean that pleasure is identical with the property goodness, or it may mean that
pleasure is a good thing—that it has the property goodness, but is not identical
with it.
Frankena reconstructs the argument as follows:
(a) Pleasure is sought by all men.
(b) Therefore, pleasure is good.
This argument, of course, is fallacious. But the fallacy at work here is not
something so complicated as a violation of Hume’s law. Rather, it is merely that
the argument is logically invalid. Nobody claims that the Epicurean argument is a
violation of the naturalistic fallacy by virtue of its invalidity. Any philosopher
interested in advancing the arguments would fix the invalidity issue first, and then
assault the argument.87
The interesting question, and the one Frankena takes up next, is whether
there is a fallacy involved when we supply the premise suppressed in the
Epicurean argument. The argument, as Frankena reconstructs it is as follows:
86
Frankena, p. 468. 87
In fact, since I take Moore to be arguing not that the mere definition of good in natural terms is
the naturalistic fallacy, but rather that the fallacy occurs in the defense of that definition, it appears
that this is Moore’s approach as well.
79
(a) Pleasure is sought by all men
(b) What is sought by all men is good (by definition)
(c) Therefore, pleasure is good.88
Premise (b) raises warning flags, just as the conclusion of the original invalid
argument did, and for the same reasons. The conclusion, we saw, was ambiguous
between being a definition and attributing a property to pleasure, that of being
good. (b) is ambiguous between offering a definition of good as being that which
is sought by all men, and offering a property attribution to those things sought by
all men. Frankena takes the proponents of the Epicurean argument to be offering
a definition of good or to be intending to assert a proposition that is true by
definition in (b). But note, given this reconstruction, if (b) is a definition, or
asserts an identity, then (c) must also do so to preserve validity. In such a case,
the argument has no moral conclusion at all. Then there would be no question
about whether the argument violated any of the three intuitionists’ theses. So, to
preserve validity and to preserve the Epicureans’ likely intentions, (b) must be
read as a property attribution that follows from a definition or identity statement.89
Frankena writes that the suppressed premise, (b) is the reason that the
argument “involves the naturalistic fallacy.”90
In identifying goodness with the
property being-sought-by-all-men, the Epicurean has committed the fallacy.
88
Frankena, p. 469. 89
It is easy to recast the whole argument in a way that avoids this difficulty, but doing so is not
necessary for our purposes. 90
Frankena p 469.
80
Frankena wants to show that in fact, to identify goodness with any other property
would be to commit the fallacy. If this is correct, then committing the naturalistic
fallacy is not the same as deducing ethical propositions from non-ethical ones.
That is, it is not a violation of the first of the three Intuitionists’ theses about the
bifurcation of moral from non-moral. Rather, Frankena takes Moore to be
claiming that the identification, itself, is the fallacy. That is, the fallacy is
committed when a definition of good in terms of any other properties is given.
Frankena’s claim is that this argument is a case where there is no longer a
violation of Hume’s law, but premise (b) is supported by an instance of the
naturalistic fallacy. If the fallacy occurs where there is no Hume’s law violation,
then committing the naturalistic fallacy is not to violate Hume’s law.
Since Frankena takes Moore’s claim to be that the identification of
goodness is the fallacy, he also claims Moore’s position is to beg the question on
the definability of goodness. Frankena writes that Moore takes (b) of the original
argument to be analytic, but not ethical, because he takes goodness to be
indefinable. Someone who thought goodness could be analyzed and thought that
(b) was a statement of the correct analysis of goodness would take (b) to be
analytic but not ethical because rather than predicating a property of that which is
sough by all men, it tells us something else. It tells us the logical relationship
between that which is sought by all men and goodness. This is a fairly
commonsensical view, but Frankena seems dissatisfied. Frankena derives his
81
view of Moore from a passage in which Moore notes that two naturalists with
competing ideas of the nature of goodness, in trying to refute each other will rely
on purely psychological statements. For example, someone who held that
goodness was desirableness and who wanted to refute the view that goodness was
pleasure would argue that pleasure is not the object of desire. This is not an
ethical/moral proposition, but a psychological one. To generalize from this
example that Moore takes all statements of the form “x is good” to be non-ethical
is a serious mistake in interpretation. But the proposition that pleasure is morally
good most certainly is a moral proposition, whether true or false. So, if (b) were
taken as property attributing, then it is ethical. If Moore takes (b) to be analytic
and not ethical, it would be because he took it to be a definition of goodness.
Regarding Frankena’s claim that Moore’s view begs the question against
naturalism, it hardly seems to be begging the question to say that a definition is
wrong because the property that is being defined is incapable of definition. It
would be begging the question to say that definition X is wrong on grounds that
definition Y was correct.91
But to say X is wrong because it is conceptually
impossible to give such a definition is not begging the question. If Moore never
offered an argument that goodness was indefinable, I might be inclined to agree
with Frankena. But he does. It is the open question argument, and perhaps, an
induction from multiple iterations of the argument. I examine, in Chapter 5,
91
The position taken by Prior’s inconsistent naturalist.
82
Moore’s argument that goodness is indefinable, and find that it is significantly
stronger than critics claim.
To support the claim that Moore takes the commission of the naturalistic
fallacy to be to identify goodness with some property (to offer a definition of
goodness), Frankena offers a number of passages from Principia Ethica,
reproduced here, in full:
[F]ar too many philosophers have thought that when they named
those other properties [belonging to all things which are good] they
were actually defining good; that these properties, in fact, were
simply not ‘other’, but absolutely and entirely the same with
goodness. This view I propose to call the ‘naturalistic
fallacy’. . . .92
and,
“I have thus appropriated the name Naturalism to a particular method of
approaching Ethics….This method consists in substituting for ‘good’ some one
property of a natural object or of a collection of natural objects….”93
and,
“…the naturalistic fallacy [is] the fallacy which consists in identifying the simple
notion we mean by ‘good’ with some other notion.94
92
Principia Ethica, p. 62, (quoted in Frankena, p 470). 93
Principia Ethica p. 91-92, (quoted by Frankena at 470). 94
Principia Ethica p. 109, cf. pp. xiii, 125, (quoted by Frankena at 470).
83
Finally, he points to another section95
where Moore notes that the fallacy is also
involved when someone identifies some natural property with some other natural
property. While he would not call this error a ‘naturalistic fallacy,’ he takes it to
be the same mistake. Frankena points out, rightly, that Moore would say the same
about identifying metaphysical objects or propositions as well.96
We can see, then, that the fallacy is not limited to cases of defining
goodness, and that it can occur when defining many properties, moral and non-
moral. Frankena says that it is really a generic definist fallacy (or, at least, that
the fallacy underlying the naturalistic fallacy is a generic definist fallacy). The
quoted Principia Ethica passages suggest to Frankena that the fallacy is “the
process of confusing or identifying two properties, of defining one property by
another, or of substituting one property for another . . . . [T]he fallacy is always
simply that two properties are being treated as one.”97
If this were the proper way
to view the fallacy, we can see that it surely is no violation of Hume’s law. In
fact, ultimately, I accept that it has little bearing on Hume’s law, but I do not
accept Frankena’s conclusions. Formulated this way, the naturalistic fallacy
reflects the literal meaning of Bishop Butler’s inscription on Principia Ethica:
“Everything is what it is, and not another thing.”98
95
Principia Ethica p. 65. 96
If, for example, someone claimed that being a set was identical with being ticklish, or being
incorporeal. 97
Frankena, p 471. 98
Frankena, p 472, quoting the motto from Principia Ethica.
84
So Frankena finds that one version of the naturalistic fallacy to be simply
the identification of goodness with any other property. While it is clear that
identifying two distinct properties is an error, is there a reason to think that
anyone who defines goodness does this? Frankena proceeds as though the
existence of the fallacy itself is supposed, by Moore, to prove that such a
definition is impossible.99
That is, that Moore assumes that goodness is
indefinable, and then uses that to claim that any proposed definition is false.100
I
reject the claim that Moore does this. His ground, in each case, for accusing a
philosopher of committing the naturalistic fallacy is not simply that they have
defined goodness.
Frankena admits that this version of the naturalistic fallacy may be a
misinterpretation of Moore and that the passages cited support a weaker version
of the definist fallacy. On this weaker version, it is an error to attempt to define
an indefinable property. This just raises the question, what properties are
indefinable? If this is the fallacy as Moore intends it, he might be held to be
begging the question in favor of his view by assuming goodness is indefinable
and then accusing competing views of committing the naturalistic fallacy.
99
This is, I suspect, because he, like Prior, does not view the open question argument to prove
anything other than that there is a fallacy. 100
Frankena writes: “To be effective at all, it must be understood to mean, ‘Every term means
what it means, and not what is meant by any other term.’ Mr. Moore seems implicitly to
understand his motto in this way in section 13, for he proceeds as if ‘good’ has no meaning, if it
has no unique meaning. If the motto be taken in this way, it will follow that ‘good’ is an
indefinable term, since no synonyms can be found. And then the method is as useless as an
English butcher in a world without sheep.” Frankena, p. 272-73.
85
We can see the appeal of such a position. After all, as I have noted,
Moore includes an independent proof of the indefinability of goodness in § 13.
So the supposition is that having proved goodness is indefinable, Moore can
accuse anyone who has attempted such a definition of committing the naturalistic
fallacy, and thereby, dispense with their views. This would not be an accurate
description of Moore’s method (though other intuitionists might be less careful).
Moore does not use the naturalistic fallacy as a weapon. He does not merely
assert that Mill’s view attempts to define an indefinable property and so is false.
Rather, he identifies a particular sort of error as the naturalistic fallacy, thereby
undermining the argument by which the offending definition was derived. This
argument against Mill will be treated more fully in Chapter 4.
Another, more important reason to reject the view that to commit the
naturalistic fallacy is to attempt to define an indefinable property is that Moore
points out examples of the naturalistic fallacy being committed in definitions of
definable properties. For example, Moore says that the naturalistic fallacy is
commonly committed with respect to beauty, just as it is committed with respect
to goodness.101
If you think the naturalistic fallacy is a mere definist fallacy, then
you would expect that Moore would have to claim that beauty is unanalyzable and
indefinable too. But he does not. In fact, three sentences after he makes this
pronouncement, he offers a definition of beauty that does not commit the
101
Principia Ethica, p. 259.
86
naturalistic fallacy. To commit the naturalistic fallacy with respect to beauty is to
suppose that since beautiful objects all cause some certain feeling in us, that to be
beautiful is to cause such a feeling, or beauty is identical with a tendency to cause
such feelings. I contend that the fallacy here is not in defining beauty in terms of
our subjective feelings, but in the inference from the conjunction of beauty and
that feeling to the identity of those properties.
The upshot for Frankena is that of the possible uses of the naturalistic
fallacy he has identified, identification with violations of Hume’s law is a mere
logical fallacy, and not broad enough to capture the full range of arguments and
theories to which the term is applied, and of the two versions of the definist
fallacy, one seems overly strong, and the other may be refuted by Moore’s actual
practice. We shall see whether this is the case in the next chapter.
Conclusion
Baldwin, Frankena and Prior each offer analyses of Moore’s use of the
term ‘naturalistic fallacy.’ Of these, Frankena’s analysis has been the most
influential, with those who think there is some problem with the naturalistic
fallacy generally adopting the position that it is a triviality. Typically, those who
do not see a problem, simply take it to be the same as Hume’s law. In the next
chapter, I explain why it is not a mere triviality. By noting the proper role of the
87
naturalistic fallacy in Moore’s thought, and placing it within the context of his
overall assault on naturalism, I show that, rather than being question begging, it is
an explanation of how naturalists arrive at their conclusions. Moore’s proof that
those conclusions are mistaken, the open question argument, is a different
argument altogether.
88
Chapter 4
The Naturalistic Fallacy: A New Interpretation
In the previous chapter, I showed that there are a number of different
interpretations of what Moore means by his claim that naturalistic theories of
ethics commit the naturalistic fallacy. In this chapter I offer a new interpretation
of the naturalistic fallacy that more closely matches the text of Principia Ethica.
Frankena’s work demonstrates that of the many possibilities considered
for what it means to commit the naturalistic fallacy, two deserve serious
consideration. The first is that to commit the naturalistic fallacy is to define an
indefinable property. As I showed in the last chapter, this interpretation is not
consistent with Moore’s text because he points out that the naturalistic fallacy is
frequently committed with respect to beauty, a definable property.
The second interpretation we must consider is that to commit the
naturalistic fallacy is simply to inaccurately define a property. For purposes of
this discussion, let’s call this ‘the general definist fallacy.’ This interpretation has
much to recommend it. After all, every example that Moore offers of committing
the fallacy is a case of someone getting a definition wrong: There is the example
of identifying yellowness with a certain type of light vibration; that of mistakenly
thinking that since I am pleased, I am the definition of being pleased; the case of
89
believing since oranges are yellow, being an orange is the same as being yellow,
etc. Every one of these definitions is supposed to illustrate the naturalistic fallacy
and every one is an example of an incorrect definition. This seems like pretty
good evidence that to commit the naturalistic fallacy is to incorrectly define a
property, that is, to think of two properties that they are one and the same.1
Despite this, I prefer a different view.
The view I favor is something akin to Prior’s view, as articulated in the
last chapter. Moore’s examples in the sections in which he describes the
naturalistic fallacy, his analyses of individual theories that are said to commit the
naturalistic fallacy, seem to stress something besides the misidentification. There
seems to be an explanation for why the mistake was made. This suggests to me
that to commit the naturalistic fallacy is to infer from inadequate evidence that
what we think of as two properties really is one. The details of this view are
worked out below.
Given that the general definist fallacy is well supported by Moore’s
examples, the evidence required to prove this is a bad interpretation would need to
be strong. The best evidence would be a Moorean example where someone got
the definition right, but was still accused of committing the naturalistic fallacy.
This would prove that the general definist interpretation was false, though it
would not necessarily prove any alternative. Unfortunately, there are no such
1 In essence, Baldwin’s second interpretation, generalized to apply to properties besides goodness.
90
examples in Principia Ethica. We might also look to examples Moore gives of
successful definitions. If we saw that Moore was very careful not to infer from
the co-instantiation of properties that they were identical, as seems to be the
mistake highlighted in several of the cases we’ve looked at, we might infer from
this that his care was directed at avoiding the naturalistic fallacy. Again, we do
not find this sort of care. In fact, it is not clear what process Moore uses to define
such properties as beauty or virtue.
When he defines beauty, he asks what properties can be removed from an
experience while maintaining its beautiful character.2 He also imagines the value
of the experience if certain features are added. He seems to be determining what
features are necessary to our experience of beauty. But there seems to be no
inference from these observations to the definition Moore offers. He writes that
“the beautiful should be defined as that of which the admiring contemplation is
good in itself.”3 Note that this is a definition of the beautiful, and does not
necessarily define the property being beautiful. Moore then takes the next step
and defines being beautiful as being an essential part of an intrinsically valuable
experience.4
2 Principia Ethica, p. 243.
3 Principia Ethica, p. 249.
4 Presumably, the contemplation.
91
One way Moore could have arrived at this definition is by intuition.
Sidgwick’s use of this method is commended as a route to determining the good,5
so perhaps it could be used to determine the beautiful as well. But if so, we still
have to overcome the hurdle of getting from the beautiful to beauty. When it
comes to good, Moore says that people theorize about the good and then decide
whatever they chose as the good is the very meaning of good. And this maneuver
is forbidden, presumably because it involves the naturalistic fallacy. But it looks
like that what he has done here. If so, this would weigh against the suggestion that
the naturalistic fallacy is a kind of inference from inadequate evidence to a
definition.
The discussion so far makes the prospect of defending the view I
suggested in Chapter 3 look pretty dim. But it is not as bad as it seems. First, the
problem raised by the definition of beauty is highly speculative. It is not at all
clear that Moore actually made the inference I claim is forbidden. Second,
despite Moore’s protestations that he does not describe a bad inference, and so has
not actually found anything that should be called a fallacy, 6 it genuinely seems
that he has identified such an inference.7 Moreover, it is hard to believe that
Moore does not understand what a fallacy is. I find it easier to believe that he
stated his examples in ways that sometimes failed to emphasize the bad inference,
5 Though his particular conclusion is not embraced. Principia Ethica, p. 145.
6 Second Preface, Principia Ethica, p. 20-21.
7 See example below, p. 102.
92
instead focusing on the falsity of the view.8 The fact that when he actually
accuses particular philosophers of having committed the naturalistic fallacy, he
never claims that their definitions are wrong but merely that they committed the
fallacy also suggests that the error is not the definist fallacy.9 Finally, the
interpretation I favor is consistent with the general definist fallacy for every
example we have, while providing a degree of explanatory power, in that it
explains why the faulty definitions were proposed. This is certainly not a
requirement of a good interpretation of the naturalistic fallacy, but it does not
hurt. What remains, then, is to look at the specific examples Moore gives us in
more detail, and see if any of them seem to weigh more heavily in favor of either
view.
Prior and Frankena, despite offering distinctly different interpretations of
what it means to commit the naturalistic fallacy, each treat it as though it were an
argument against naturalistic theories. In so doing, they share a common error.
The first section of this chapter explains the error and shows that Moore was
explicit in his assertion that theories are not false by virtue of resting on an
instance of the naturalistic fallacy. Evidence for this claim is provided in two
8 When it was a complex definition. A corollary of my view of the open question argument is that
Moore cannot prove that any simple property identity is false by use of the argument. He may
show that a given simple definition is wrong by means of a specific argument, perhaps by showing
that goodness and the property are not even coextensive, so surely not identical, for instance. 9 The closest he comes to claiming that the definition offered by a philosopher is wrong is when he
writes “The direct object of Ethics is knowledge and not practice; and anyone who uses the
naturalistic fallacy has certainly not fulfilled this first object. . . .” Principia Ethica, p. 71. Though
this could mean either a mistake in the definition, or a failure in its justification.
93
forms—direct and indirect textual evidence. By ‘indirect evidence’ I mean I
examine one or two of Moore’s arguments against particular forms of naturalism,
and demonstrate that at no point in these arguments does Moore assert that they
are false for resting on the naturalistic fallacy. Examination of such arguments
makes it clear that Moore does not intend us to believe that naturalistic theories
are false for resting on the fallacy. Rather, it appears that, in each case, Moore is
merely pointing out that there is such an error being made that explains where
philosophers who advocate such theories have gone wrong—why it is that they
embrace and endorse false theories. But the theories are alleged to be false by
virtue of some other error.10
The proof that such identities are false ones is not
proved by the naturalistic fallacy, or by accusation that advocates of those
theories commit the fallacy. The actual proof is provided by the open question
argument. This proof is examined in detail in the next chapter.
After demonstrating that Moore does not take the fact that a theory rests
on the naturalistic fallacy to be a fatal flaw for the theory, I present my own
account of the fallacy. I draw on Moore’s description of the fallacy, and on his
arguments against the theories of Bentham, Mill, and Spencer. I argue that a
single interpretation is consistent with each of Moore’s usages. In the course of
this discussion, I consider several criticisms, including those inspired by Moorean
10
The theories, if they are false, are so because they are wrong about what things ought to be done
or because they advocate an incorrect theory of the good. And the evidence for their theories are
inadequate because by having committed the naturalistic fallacy, the advocates have undermined
the possible evidence for their normative claims.
94
text that, on an initial read, seems to contradict my interpretation of the
naturalistic fallacy.
§ 1) The proper role of the Naturalistic Fallacy
We saw, in the last chapter, that Moore scholars have noted that there is
debate about the role of the naturalistic fallacy in Moore’s ethical theory.
Interestingly, while this debate is noted, none of the philosophers whose work we
examined attempted to resolve the debate. Moreover, at least two of them,
Frankena—who pointed out that there are some intuitionists who use the
naturalistic fallacy like a weapon—and Prior—who pointed out that the
naturalistic fallacy, as he conceives it, is an effective weapon (at least against
those he called “inconsistent naturalists”)—take sides on the matter. That is, both
assume that the naturalistic fallacy is intended to refute naturalistic theories of
ethics. This is the very basis of Frankena’s claim that it is question begging. It is
also the reason I reject Prior’s `characterization, despite agreeing with him, by and
large, about what it means to commit the naturalistic fallacy.
Since the basis of my criticisms of Frankena and Prior rest, in part, on
their having gotten the role of the naturalistic fallacy wrong, it is incumbent on
me to provide some proof that Moore, at least, did not intend the naturalistic
95
fallacy to be a refutation of anything. My claim rests on two different types of
support—text and inference from Moore’s own practice.
Moore, himself, clarifies the proper role of the naturalistic fallacy while
discussing Bentham’s commission of the naturalistic fallacy:
Now, I do not wish the importance I have assigned to this fallacy
to be misunderstood. The discovery of it does not at all refute
Bentham’s contention that greatest happiness is the proper end of
human action, if that be understood as an ethical proposition, as he
undoubtedly intended it. That principle may be true all the same;
we shall consider whether it is so in succeeding chapters. Bentham
might have maintained it, as Prof. Sidgwick does, even if the
fallacy had been pointed out to him. What I am maintaining is that
the reasons which he actually gives for his ethical proposition are
fallacious ones so far as they consist in a definition of right.11
Here, Moore is pointing out that fact that Bentham has committed the naturalistic
fallacy is not the feature that makes his theory false. Whether it is true or false
rests on other grounds. Moore considers these grounds in Chapter III, under the
topic of Hedonism, generally. I examine Moore’s argument, and in particular, the
role of the naturalistic fallacy in that argument below. But regardless of the
theory’s truth, or lack thereof, the reasons that Bentham gives for accepting that
11
Moore, Principia Ethica, Ch. I, § 14, ¶4.p 71 (emphasis in original) (all citations to Principia
Ethica are to the revised edition, Cambridge 1993).
96
theory are the proper target of the naturalistic fallacy. This is why Moore writes,
when speaking of Bentham’s reasons that Bentham “did not perceive them to be
fallacious” and that “if he had done so, he would have been led to seek for other
reasons in support of his Utilitarianism . . . .”12
Plainly, the text indicates that
Moore considers the role of the naturalistic fallacy to be to undermine the reasons
for holding and promoting a theory, and not to prove its falsity. This text is the
most direct support for my contention that the naturalistic fallacy is not in any
way Moore’s argument that any particular ethical view is false. Rather it is
merely his diagnosis of where a philosopher has gone wrong in forging or
defending his theory.
Unfortunately, though this passage is helpful in clarifying the proper role
of the naturalistic fallacy, it does not support my claim that to commit the
naturalistic fallacy is anything other than to commit the general definist fallacy. It
does, however, prove that to interpret the naturalistic fallacy in such a way as to
make commission of the fallacy the false-making characteristic of the normative
theory, rather than a mistake that undermines our confidence in it would be a
mistake. It also give us hope that since Moore was so concerned with the reasons
given for ethical views, he might be similarly concerned with reasons for
accepting definitions of moral properties.
12
Principia Ethica, p. 71.
97
Though this passage has been generally overlooked, I have seen it cited
two times. The first paper I have found to make appeal to the quote above is
Everett Hall’s “The ‘Proof’ of Utility in Bentham and Mill.”13
Hall treats Moore
as though he had claimed that Bentham had committed the naturalistic fallacy in
its definist configuration. That is, he accepts Frankena’s interpretation of the
fallacy, and accuses Moore of begging the question against Bentham. Arguably,
it is possible that Moore was accusing Bentham of committing a definist fallacy,14
but I consider is a better explanation, below.15
The only other reference I have found to this passage commits a pretty
grievous error. Julius Kovesi cites to this passage as proof that Moore thinks the
error in Bentham’s reasoning lies in the fact that he presented an argument
wherein he defined the good. “The fallacy consists in the reasoning adduced to
show what the good is; whereas it should be laid down what the good is ‘as an
axiom’.”16
This is most definitely not what Moore asserts. In fact, Moore writes
“it might be perfectly consistent for him to define right as conducive to the
general happiness provided only (and note this proviso) he had already proved, or
laid down as an axiom, that general happiness was the good.”17
Moore thinks it is
perfectly acceptable to try to prove what the good is, and doing so, in addition to
13
Ethics vol. 60, No. 1 (Oct 1949) pp. 1-18, at 15. 14
Though I attempt to refute this charge, below. 15
In my detailed explication of Moore’s analysis of Bentham, p. 113 16
Kovesi, J., “Principia Ethica" Re-Examined: The Ethics of a Proto-Logical Atomism,”
Philosophy, Vol. 59, No. 228. (Apr., 1984), pp. 157-170 at 170. 17
Principia Ethica, p. 70. (Emphasis in original)
98
laying it down as an axiom, is the key to avoiding the naturalistic fallacy for
Bentham.18
As I argue below, the fallacy lies in making the inference from what
things are good to a definition of goodness. Kolvesi confuses the good, which
Moore takes to be those things that instantiate goodness, with goodness itself.
The upshot is that a key passage in Principia Ethica—one that effectively
undermines some of the most potent criticisms asserted against Moore’s ethical
philosophy and perhaps leaves a little hope for a non-definist view of the
naturalistic fallacy—has been rarely cited and, perhaps, badly misinterpreted.
Hall accuses Moore of committing the definist fallacy, and Kolvesi takes Moore
to reject a proposition that he explicitly endorses. While Moore believes Bentham
has committed the naturalistic fallacy, he does not think his normative theory is
false on that ground, and he does not take the fallacy to be in attempting to prove
that a particular property is the good.
Moore gives similarly deferential treatment to the theories of Mill and
Spencer. Though he brutally assaults their arguments as artless commissions of
the naturalistic fallacy, and argues that each theory is without merit, he does not
argue that they are false for being based on the fallacy. Moore’s refutations of
Mill’s hedonism and Spencer’s evolutionary ethics are based on other grounds.
18
This is not to say that he thinks such proof is possible. We see in his discussion of intuitionist
hedonism that he takes direct proof of propositions about the good to be impossible. Such claims
are only open to ‘indirect proof.’ Principia Ethica, p. 128.
99
As is the case with Bentham’s views, Moore takes Mill to task for
committing the naturalistic fallacy. After several pages of accusing Mill of
naïveté and artlessness, and giving him some good-natured ribbing (“Well, the
fallacy in this step is so obvious, that it is quite wonderful how Mill failed to see
it”19
) for having committed the naturalistic fallacy, Moore turns his attention to
disproving the theory. Again, we see that a theory’s resting upon the naturalistic
fallacy does not demonstrate that the theory is false. It merely undermines our
confidence in the arguments mustered for its support. Moore writes “So far I
have been only occupied with refuting Mill’s naturalistic arguments for
Hedonism; but the doctrine that pleasure alone is desirable may still be true,
although Mill’s fallacies cannot prove it so. This is the question which we have
now to face.”20
The remaining thirty-five pages of Moore’s chapter III are
devoted to trying to show that Mill’s principle is false. If Moore took commission
of the naturalistic fallacy to be a fatal flaw in a theory there would be no need for
such arguments.21
Most importantly, for my purposes, Moore does not claim that
Mill has gotten the definition of goodness wrong.
Herbert Spencer’s evolutionary ethical theory gets similar treatment.
Moore identifies an instance of the naturalistic fallacy in Spencer’s identification
19
Principia Ethica, p. 118. 20
Principia Ethica, p 126. 21
Nor for the apologies he gives for the sorry state of ethical argumentation. He recognizes that
the sort of refutation that we see in logical proofs is not possible in the context of ethics, at least
with respect to arguments about what things are good.
100
of increasing ethical sanction with being more evolved. In other words, he has
defined being more evolved as an end in itself. Moore is unsure whether Spencer
has really committed the fallacy because “Mr. Spencer’s language is extremely
loose,”22
but whether he has or has not committed the fallacy, his theory must be
further evaluated for its truth. Again, commission of the fallacy is not sufficient
to demonstrate that a theory is false.
It is clear that Moore, at least, does not use the naturalistic fallacy as a
weapon against naturalists in the way that Frankena and Prior seem to assume. In
no case does he dismiss a theory as false on the ground that its proponent has
committed the fallacy in formulating or defending the theory. Nor does he claim
that the property identities these theorists assert are false on the ground that
goodness is indefinable. Surely, to assert that would be to beg the question.
Rather, Moore rejects the reasoning behind the theories as being fallacious, and
then turns his attention to arguing against the other merits of the theories. Finally,
Moore does not claim that any theorist has committed the fallacy merely on the
grounds that he has offered a definition of goodness, an indefinable property.
This last point is demonstrated in the next section, wherein my positive account of
the naturalistic fallacy is presented.
Again, nothing in the preceding material proves that Moore does not take
the naturalistic fallacy to be a definist fallacy, but neither does it prove that he
22
Principia Ethica, p. 100-101.
101
does. In order to be a definist fallacy, in order to avoid begging the question,
Moore would need proof that goodness is indefinable. He would also need proof
that it is not identical with any natural simple property, because definition is a
matter if identifying the constituent parts of a property—if there are no parts,
there can be no definition, but there could still be true identity statements that
identify goodness with a simple property. If Moore does not have a proof that
goodness is not identical with a simple natural property—and I contend he does
not—then in order to defend his view that goodness is non-natural, he would need
a view of the naturalistic fallacy such that commission of the fallacy could
undermine confidence in proffered simple identity statements. A definist fallacy
would not suffice. The rest of this chapter develops a view that would.
§ 2) The naturalistic fallacy: a close reading
It is not entirely surprising that most interpretations of the naturalistic
fallacy are unsatisfying. Moore uses a number of examples to illustrate it, and
they do not always appear, at first glance, to demonstrate the same mistake.
Furthermore, sometimes his descriptions are downright cryptic, as, for example,
when he writes of a particular class of naturalistic theories:
In other words they are all theories of the end or ideal, the adoption
of which has been chiefly caused by the commission of what I
102
have called the naturalistic fallacy: they all confuse the first and
second of the three possible questions which Ethics can ask.23
This statement, for all the world, appears to mean that to commit the naturalistic
fallacy is to confuse two questions. Compounding the confusion is that Moore is
not very clear about which two questions he is referring to. The most likely
candidates appear in the opening sections of Chapter I, thirty or more pages
before this passage, but these same questions appear again, just before this
passage, but in a different order.24
Interpreting this is tough—Moore may be
defining the naturalistic fallacy, he may be offering an alternative way to think
about the problem without suggesting it means anything other than what he had
previously established, or he may be doing neither of these, and instead merely
suggesting an alternative error was committed by the naturalist.
Nonetheless, an interpretation consistent with each of Moore’s uses of the
words “naturalistic fallacy” can be found. In this section, I offer such an
interpretation.
a. Goodness and yellow
Moore’s most clear explanation of the naturalistic fallacy appears in the
passage in which the naturalistic fallacy is first introduced. Moore writes:
23
Principia Ethica, pp. 89-90. 24
Compare Principia Ethica, Ch. I, pp 54-58, and Ch. II, p 89.
103
It may be true that all things which are good are also something
else, just as it is true that all things which are yellow produce a
certain kind of vibration in the light. And it is a fact, that Ethics
aims at discovering what are those other properties belonging to all
things which are good. But far too many philosophers have
thought that when they named those other properties [belonging to
all things which are good] they were actually defining good; that
these properties, in fact, were simply not ‘other’, but absolutely
and entirely the same with goodness. This view I propose to call
the ‘naturalistic fallacy’….25
We saw in Chapter 3 that Baldwin uses this passage as evidence that to
commit the naturalistic fallacy is to deny the indefinability of goodness. And we
also saw that Frankena relies on this passage (among others), claiming that it
means that nothing is identical with goodness26
and that to commit the naturalistic
fallacy is to assert of something other than goodness that it is identical with
goodness.27
Frankena’s emphasis is on the act of attempting the definition. This,
he thinks, is what Moore has labeled the “naturalistic fallacy.” But this is not the
only reasonable interpretation, nor is it the one that most accurately represents
Moore’s text. The view Moore proposes to call the naturalistic fallacy is not the
25
Principia Ethica, § 10, p. 62. 26
He takes this passage to be a mere demonstration of Butler’s maxim that appears as the motto on
the cover page of Principia Ethica, “Everything is what it is, and not another thing.” 27
Frankena, p. 470-71.
104
belief that defining goodness is possible, or even that any particular identity
statement is right. Rather, Moore writes that it is the belief that when one names
the other properties, those properties that are coextensive with goodness, one is
defining goodness.
Consider Moore’s example step by step. The naturalist observes that
every instance of goodness is also an instance of pleasure, just as he might notice
that every instance of yellow is accompanied by y-vibrations. From this (limited)
coextension of goodness and pleasure, the naturalist deduces that goodness is
identical with pleasure. I take it that this deduction is Moore’s target. To commit
the naturalistic fallacy is to make the inference from coextension to identity.
Moore calls it a fallacy because he takes the deduction itself to be invalid, not
simply because he believes the conclusion is false.
We know that such deductions are invalid because coextension, even
necessary coextension, is not identity. Consider, for example, the properties
being renate and being cordate. As it turns out, these are coextensive properties.28
Everything that has a kidney has a heart, and vice-versa. But just a moment’s
reflection makes it clear that they are not identical properties. The same is true
for yellowness and y-vibrations; just a second of careful thought dispels the
possibility that they are identical. They are merely properties that are
coextensive, or perhaps they share an even weaker relation. After all, the
28
At least in philosopher biology.
105
yellowness of an object only causes y-vibrations under certain conditions. That
is, the y-vibrationality is only a disposition that yellow things have.
Now Frankena is quite right that to make the move from this claim about
the naturalist to the further claim that her definition is wrong is impermissible.
Moore also acknowledges this in the passage discussing Bentham, above. When
Frankena claims that Moore begs the question, he seems to have in mind a
different Moorean argument. Moore would have begged the question against the
naturalist if he argued as follows:
1) Goodness is coextensive with pleasantness
2) Goodness is not identical with pleasantness
3) If (1) and (2) then goodness is not identical with pleasantness.
4) Therefore, goodness is not identical with pleasantness.
Since (2) is the very conclusion that Frankena takes Moore to be reaching
when he claims that someone who defines goodness as pleasantness is committing
the naturalistic fallacy, to make such an argument would be to beg the question
against the naturalist. To see this, we need only to look at the reason Moore
might assert (2). Frankena thinks the reason Moore believes that goodness is not
identical with pleasantness is that to believe such a thing is to commit the
naturalistic fallacy. The proof that someone has committed the naturalistic fallacy
requires one to deny the naturalist’s conclusion as one of your premises. This is
106
plainly question begging. This is why it is so important to emphasize that this is
not what Moore actually does.
Actually, this example is worse still for the naturalist, because he has not
started from the premise that goodness and pleasantness are coextensive. Instead,
the naturalist has started from the premise that all instances of goodness are
instances of pleasure. From this, it would be, perhaps, worse to infer the identity
of goodness and pleasantness, because all it would take to disprove the identity
would be a single disconfirming instance of pleasure that was not good. At least
if the properties were coextensive, there could be no such disconfirming
instance.29
The mistake our naturalist has made is to believe that goodness and
pleasantness are identical on the ground that he has always found them together.
This suggests that, in this case, to commit the naturalistic fallacy is to infer from
the fact that every good thing also is pleasant that goodness is identical with
pleasure. In general terms, it is the inference from the constant conjunction of one
property with another to their identity. This constant conjunction is a weaker
relationship than coextension, as Moore seems to describe it as a one way relation
here.30
I consider the ramifications of this below, in my discussion of Moore’s
29
Clearly, if the properties were merely contingently coextensive, there could be a possible
disconfirming instance of pleasure. But since the inference is bad even if the properties are
necessarily coextensive, it seems trivial to compound the trouble. 30
Indeed, in the cases where naturalists have claimed that some natural property is the good
(where there is real coextension), Moore proceeds as though they derived that from the identity
107
treatment of particular theorists. For the moment, we should progress on the
theory that to commit the naturalistic fallacy is to infer from either the
coextension of two properties or the mere supposition that all instances of one
property are also instances of another, that the properties are identical.
Symbolically:
NF1) ∀x (Px ↔ Qx) → (P=Q)
or
NF2) ∀x (Px → Qx) → (P=Q)
b. “I am pleased”
The next place we see Moore invoking the naturalistic fallacy is several
pages later, in § 12. Here, there are two related passages, each of which is
relatively easy to misinterpret alone, but when read together, support my
interpretations over any others.
Suppose a man says ‘I am pleased’ . . . . We may be able to say
how [pleasure] is related to other things: that, for example, it is in
the mind, that it causes desire, that we are conscious of it, etc., etc.
We can, I say, describe its relations to other things, but define it we
between goodness and the property in question, after having first committed the naturalistic fallacy
by inferring the identity from the mere constant conjunction.
108
can not. And if anybody tried to define pleasure for us as being any
other natural object; if anybody were to say, for instance, that
pleasure means the sensation of red, and were to proceed to deduce
from that that pleasure is a colour, we should be entitled to laugh at
him and to distrust his future statements about pleasure. Well, that
would be the same fallacy which I have called the naturalistic
fallacy.31
At a first read, it is easy to take Moore to be saying that to commit
the naturalistic fallacy is, in this case, to define pleasure. And since he has
identified pleasure as being indefinable, we might generalize to the
conclusion that to commit the naturalistic fallacy is to give a definition of
an indefinable property.32
But, as I noted in my discussion of Baldwin,
something a little more complex and interesting is going on here. It is not
simply that someone is trying to define pleasure. It is that they are
attempting to do so by virtue of its relation to other natural properties or
objects. That is, she is identifying pleasure with other attendant
experiences, like sensations of red, or with its causal relata, like the desire
it causes in us. The fact that this is the error Moore has in mind becomes
clearer as we read farther in the passage.
31
Principia Ethica, § 12, pp 64-65. 32
This seems to be the assumption that Baldwin takes from this passage, and it is definitely the
interpretation that Frankena accepts.
109
Moore goes on, in this passage, to note that while pleasure is
indefinable, it is not the case that we cannot meaningfully assert that we
are pleased. It is enough to understand that being pleased means to have
the sensation of pleasure.33
It seems to me that he is addressing concerns
that if we cannot give a definition of a property, we cannot meaningfully
assert that we have that property. But Moore’s point is that of course we
can assert that we instantiate a property that is simple and indefinable. But
the attempt to define such a property by its connection to other properties
we might experience at the same time is an error. Moore wraps up this
passage by introducing another mistake, one that at first blush, seems
distinct. But once again, a little careful analysis shows that it is closely
related to the mistake highlighted above.
If I were to imagine that when I said ‘I am pleased,’ I meant that I
was exactly the same thing as ‘pleased,’ I should not indeed call
that a naturalistic fallacy, although it would be the same fallacy as
I have called naturalistic with reference to Ethics. The reason of
this is obvious enough. When a man confuses two natural objects
with one another, defining one by the other, if for instance he
33
Moore phrases this in linguistic terms, quoting off ‘pleased’ and ‘having the sensation of
pleasure.’ But this is a mistake. He is not making a mere linguistic point. His point is about the
property itself. I take it that this is why he admits to making use/mention errors in his discussion
of the naturalistic fallacy. But those use/mention errors need not be fatal to the view. We may,
with a little sensitivity to the text, discern where he is discussing words versus the properties to
which those words refer.
110
confuses himself, who is one natural object, with ‘pleased’ or with
‘pleasure’ which are others, then there is no reason to call the
fallacy naturalistic. But if he confuses ‘good,’ which is not in the
same sense a natural object, with any natural object whatever, then
there is a reason for calling that a naturalistic fallacy; its being
made with regard to ‘good’ marks it as something quite specific,
and this specific mistake deserves a name because it is so
common.34
There are two important things happening in this passage. The first is that
Moore provides another example of the fallacy. As I noted in Chapter 3, it is of a
case that we might never expect to see actualized. After all, it is hard to imagine
anyone mistaking the ‘is’ of predication with the ‘is’ of identity when the
identifier, himself, is the subject of the identity he asserts. It is hard to confuse
myself with something else.35
The man in Moore’s example has concluded from
the fact that he is pleased that he is identical with the property being pleased.36
It
appears in this example that the hapless identifier is confusing the ‘is’ of
predication with the ‘is’ of identity. It seems that he has thought about the true
34
Principia Ethica, p 65. 35
Though it is not impossible. I once woke up with my arm under my body, and asleep, so I could
not feel my weight upon it, and thought that it was not my arm. This, in a sense, is to confuse
myself with something else . . . 36
Again, we have a use/mention problem here. But it is clear, from the sentences immediately
following the initial description of the confusion, that by ‘pleased’ offset in single quotes, Moore
intends us to understand, not the word ‘pleased,’ but the property to which that word refers.
111
sentence, “I am pleased” which means, to him, that he has the property being
pleased (or alternately, that he is experiencing pleasure) and derived from it the
identity statement “I am pleased” which means that he is identical with the
property being pleased. This passage seems particularly amenable to this sort of
reading.37
But such a reading would be a mistake. Moore noted in his preface to
the second edition of Principia Ethica that this passage is amenable to such a
reading, though it is not what he intended. Rather, he takes confusing the ‘is’ of
predication with the ‘is’ of identity to be the source of the commission of the
fallacy in this particular instance (in my terms, having made that mistake is the
reason for the bad inference).38
While confusing predication and identity might
be at work here, also at work is the inference from co-instantiation to identity. In
this case, our confused identifier instantiates both being pleased and being
himself. And on the basis of that he takes being himself and being pleased to be
identical. Read in the context of the passage immediately preceding this one,
where the error is plainly inferring the identity of properties that occur in the same
object, it is clear that this is interpretation is to be preferred over the
identity/predication problem. This is an even more dramatic error than inferring
the identity of two properties from their uni-directional relation. Here, the
37
Moore’s use/mention trouble on the following page in discussing oranges, yellowness and
sweetness also might lead to one accepting this interpretation. Concluding from the fact that
oranges are yellow that orange and yellow are the same property looks like it might be a problem
caused by confusing the ‘is’ of predication with the ‘is’ of identity. But, again, in context it
appears to be a case of confusing coextension with identity (a closely related problem, it seems). 38
Principia Ethica, preface to the second edition, p. 20.
112
confused identifier seems to be making the inference from a single instance of the
conjunction of two properties. Symbolically, the fallacy looks like this:
NF3) ∃x (Px & Qx) → (P=Q)
The second thing Moore does in this passage is to demonstrate to us that
the naturalistic fallacy is a general purpose fallacy. That is, it is a mistake that can
be made in many contexts, not merely in the context of ethics. This is important
to keep in mind, because it can help keep up from accepting interpretations of the
fallacy that will be applicable only to goodness. For example, Frankena claims
that an intuitionist’s accusation that a naturalistic theorist has committed the
naturalistic fallacy begs the question against the naturalist because it starts from
the premise that no such definition is possible. That is, Frankena takes
intuitionists to mean that the indefinability of goodness is an integral part of the
naturalistic fallacy. But once we see that the naturalistic fallacy is supposed to be
a completely general sort of inference error, we should no longer feel compelled
to entertain interpretations like the one Frankena asserts of intuitionists. A
philosopher might commit the naturalistic fallacy in offering a definition of a
perfectly definable property. If, based on the fact that I have a receding hairline, I
determined that definition of ‘forehead’ is that patch of hairless skin extending
from ones eyebrows to the middle of the scalp, I will have committed the
naturalistic fallacy, in form NF3. I think the blame for part of the confusion in
this regard belongs squarely on Moore’s doorstep, because in illustrating the
113
fallacy, he chooses only those properties he has determined to be indefinable. But
in fact, the generic fallacy he highlights would apply equally to definable
properties. Take renate and cordate, for example. These are definable properties,
but to conclude, based on their coextension, that they were identical properties
would be to commit the fallacy. It would be an invalid inference. It is likely that
Moore only chooses indefinable properties because he wants to keep the analogy
to goodness fairly close, and ultimately, he will argue that goodness is
indefinable, but that is a different argument, not to be confused with the
naturalistic fallacy.
3. Bentham
The next place we encounter the naturalistic fallacy is in Moore’s
treatment of Bentham’s hedonism. Moore writes that Bentham’s theories about
rightness, at least as they are interpreted by Sidgwick, commit the naturalistic
fallacy.39
This material is fairly dense—perhaps the most difficult passage in
Principia Ethica—and it is hard to derive an interpretation of the naturalistic
fallacy from Moore’s statements about Bentham. But I think it is possible to
show that my claim—that to commit the naturalistic fallacy is to infer the identity
39
Principia Ethica, § 14, p. 69. Note that Sidgwick, himself, does not commit the naturalistic
fallacy, and is unique in Moore’s mind for avoiding that error.
114
of two properties from their coextension or some weaker relation—is at least
consistent with Moore’s criticisms here.
According to Moore, Sidgwick has claimed40
that Bentham takes the word
‘right’ to mean ‘conducive to general happiness.’ This definition of the word
‘right’ is permissible, says Moore, if it is asserting that ‘right’ refers to those
actions that are good as a means to achieving some particular end (i.e., they are
instrumentally good). Then, asserts Moore, Bentham could choose general
happiness as the good, either by proof or axiomatic assertion. This would be a
fine theory, and Bentham’s reasons for accepting it would be as good as his
reasons for asserting that general happiness was the good. By ‘the good’ we
mean the collection of those things that are morally good—that instantiate
goodness. If general happiness were the good, this would mean that general
happiness was the only thing that instantiated the property goodness. But, claims
Moore, Bentham has not proved or asserted axiomatically what the good is. What
he has done instead is to assert the definition of ‘right’ as conducive to general
happiness, and then to leverage that definition into a definition of the property
moral rightness. He takes the greatest happiness to be the right and proper end of
human action. Moore writes:
[Bentham’s] fundamental principle is . . . that the greatest
happiness of all concerned is the right and proper end of human
40
In The Methods of Ethics (1874), Bk. 1, Chap. iii, § 1.
115
action. He applies the word ‘right,’ therefore, to the end, as such,
not only to the means which are conducive to it; and that being so,
right can no longer be defined as ‘conducive to the general
happiness,’ without involving the fallacy in question.41
There is simply not enough information here to determine where the mistake is.
But if we tease the argument apart, it appears that Bentham may have committed
the naturalistic fallacy as represented by NF2, above.
Bentham wishes to defend the principle that ‘right’ means ‘conducive to
general happiness. Moore says that this is a perfectly acceptable theory, provided
Bentham meant by it that rightness is good as a means to the appropriate end, and
provided that he arrived at that conclusion in a particular way. If Bentham had
started from the principle that general happiness was the good, by which Moore
means that general happiness alone is good (thus it is the end—that for which we
ought to strive), and then inferred from this principle and the principle that right
means good as a means to that end that rightness is conducive to the general good,
Bentham would have committed no fallacy. In other terms, this is an acceptable
argument:
Permissible Benthamic argument
1) General happiness is the good (axiomatic or proved by other argument)
2) ‘Right’ denotes what is conducive to the end (to the good, that is)
41
Principia Ethica, p. 70.
116
3) Therefore right means conducive to the general happiness.
But, writes Moore, this is not the argument Bentham makes. Bentham
wants to show that the greatest happiness is the right and proper end of human
action. That is, he wants to prove (1) from the Benthamic Argument. How he
goes about it involves the naturalistic fallacy. Moore says the mistake he makes
is that “the definition of right as conducive to general happiness proves general
happiness to be the right end—a perfectly invalid procedure.”42
Moore accuses
Bentham of deriving the end of ethics from the definition of rightness. This does
not sound anything like the naturalistic fallacy as interpreted by me or any other
critic we have considered. I have cast the naturalistic fallacy as the inference of a
definition of a moral property based on its co-instantiation with some other
property. What Bentham appears to be doing is deriving the co-instantiation of
properties based upon their identity. “If right, by definition, means conducive to
general happiness, then it is obvious that general happiness is the right end.”43
The inference from the identity of rightness and conduciveness to general
happiness to general happiness being the right end is logically perfectly
acceptable. The question is how did Bentham arrive at that identity? This is
where the naturalistic fallacy occurs, if at all.
42
Principia Ethica, 70-71. 43
Principia Ethica, p. 70.
117
Moore takes Bentham’s claim that ‘right’ means ‘conducive to general
happiness’ in two ways—right as a means to an end, and right as an end. Since
Bentham takes ‘right’ to mean ‘conducive to general happiness,’ then every
instance of rightness will be an instance of conduciveness to general happiness,
thus:
Letting R symbolize rightness and H symbolize conduciveness to general
happiness,
∀x(Rx → Hx).
Moore is not certain that Bentham makes this claim. I take this as the
reason he is uncertain whether Bentham actually committed the naturalistic
fallacy.44
Bentham might not make this claim, and still be on the road to the
naturalistic fallacy, however. He might proceed from particular instances of
rightness he has identified as being conducive to the general happiness. He could
generalize to all instances of rightness, thereby satisfying the antecedent of NF2,
or he could fail to do so, and attempt the derive the identity of rightness and
conduciveness to general happiness from the more modest claim ∃x(Rx & Hx),
thereby satisfying the antecedent of NF3.45
44
“ . . . I am inclined to think that Bentham may really have been one of those who committed it.”
Principia Ethica, p. 69. 45
I take Moore to be saddling Bentham with the stronger claim, but the evidence is not clear.
118
From either starting point, the inference to the identity of rightness and
conduciveness to general happiness would be an instance of committing the
naturalistic fallacy in one of its configurations:
NF2: ∀x(Rx → Hx) → ∀x(Rx = Hx)46
NF3: ∃x(Rx & Hx) → ∀x(Rx = Hx).
Now, I am not convinced that the principle that the general happiness is
the right end is deducible from the identity between rightness and conduciveness
to general happiness. But that is not central to my argument. That Moore claims
this is what Bentham does is enough for my purposes.
To recap, Bentham attempts to derive the principle that general happiness
is the proper end of ethics. He derives this from the definition of rightness as
conduciveness to general happiness. This definition must come from somewhere.
Bentham (might) derive it from particular instances of co-instantiation between
rightness and conduciveness to general happiness, which he would believe existed
given that he took ‘right’ to mean ‘good as a means to general happiness.’ If
Bentham has so derived the definition, then he has committed the naturalistic
fallacy.
This schematic might prove useful:
46
Given the definition of ‘right’ in the sense of ‘good as a means,’ the left side might actually be a
bi-conditional. That is, the fallacious Benthamic argument might start from the coextension of
rightness and conduciveness to general happiness, and infer their identity.
119
Particular instances of rightness that are conducive to general happiness
(Bentham would likely think they all were)
⇓
Rightness = Conduciveness to general happiness
⇓
General happiness is the proper end of ethics (the good)
While this analysis of Bentham’s argument and his possible commission
of the naturalistic fallacy is somewhat speculative, no other account of the
naturalistic fallacy that I have seen is capable of accounting for any commission
of the fallacy at all.
4. Confusing two questions
In the chapter entitled “Naturalistic Ethics” Moore examines several
theories all of which are said to either commit the naturalistic fallacy or are
accepted readily because of the commission of the naturalistic fallacy. He writes:
“This discussion will be designed both (1) further to illustrate the fact that
the naturalistic fallacy is a fallacy, or, in other words, that we are all aware
of a certain simple quality, which (and not anything else) is what we
Naturalistic
Fallacy
occurs here
120
mainly mean by the term ‘good’; and (2) to shew that not one but many
different things possess this property.”47
Moore takes it as proof that the naturalistic fallacy is a fallacy that we have a
shared conception of goodness. I suppose if we did have such a shared
conception and it is what we meant by uses of the word ‘good,’ then it would be a
mistake of some sort to claim that what we meant was some other property. The
fact that we have a shared conception of this simple property is evidence that the
bad inference Moore has pointed out really is a fallacy. Insofar as an argument
with a necessarily false conclusion can be said to be fallacious, I suppose Moore
is correct. Moore hopes to convince us that the naturalistic fallacy is a fallacy
based, not on our recognition of the invalidity of the inference, but also based on
the fact that we take the conclusion to be false. Once again, we see that he does
not make the claim that naturalists commit the naturalistic fallacy because they
define goodness. Instead, he aims to show us that the fallacy really involves a
mistake by independently demonstrating that the naturalists’ definitions are false.
The order of operations is key here—failure to notice this is likely the reason that
the naturalistic fallacy has been taken to be a definist fallacy.
The second point of his discussion—that many different things possess
this property—is an attempt to show that the good (not goodness) is multifaceted,
in that goodness can be instantiated in many things, not only in instances of
47
Principia Ethica, p. 90
121
pleasure or any other single thing. The naturalists Moore takes as his primary
target in this and succeeding discussions are those that claim that goodness is
instantiated in one and only one type of thing. We can call these theories naïve
naturalistic theories.
More important than either of these claims, however, is Moore’s attempt
to give yet another explanation of the naturalistic fallacy. Discussing the naïve
naturalistic theories, Moore writes:
They all hold that there is only one kind of fact, of which the
existence has any value at all . . . . [T]he main reason why the
single kind of fact they name has been held to define the sole good,
is that it has been held to define what is meant by good itself. In
other words they are all theories of the end or ideal, the adoption of
which has been chiefly caused by the commission of what I have
called the naturalistic fallacy: they all confuse the first and second
of the three possible questions which Ethics can ask. It is, indeed,
this fact which explains their contention that only a single kind of
thing is good.48
Reading this quote makes it quite clear what error Moore focused on in
Bentham. Bentham posited a theory of the ideal, or end, wherein it consisted of
only one thing, general happiness. Moreover, he arrived at his conclusion about
48
Principia Ethica, 89-90.
122
the end based on defining a fundamental moral property, rightness, as identical
with some other property, one which instantiates rightness. That definition was,
itself, fallaciously deduced. To be “chiefly caused” by commission of the
naturalistic fallacy is to be deduced from an identity statement that was itself the
result of commission of the naturalistic fallacy.
Symbolically, using P for being pleasant and G for being good,
1) ∃x(Px & Gx) Something is pleasant and good
2) ∀x(Px → Gx) All pleasant things are good
3) ∀x(Px ↔ Gx) Pleasantness is the good (pleasure is the sole good)
4) P=G Being pleasant is being good (or pleasure is
goodness)
The naturalistic fallacy is the inference from (1) or (2) or (3) to (4). Naïve
naturalists compound the error, deriving (4) from (1) or (2) and then deriving (3)
from (4).
This becomes still more clear when we consider Moore’s claim that the
naïve naturalists, in committing the naturalistic fallacy, all confuse the first and
second of three questions ethics asks. The three questions are (1) What is
goodness (the property)? (2) What things are good in themselves (what is the
123
good)? (3) What ought we to do?49
By virtue of committing the naturalistic
fallacy, naturalists believe themselves to have defined goodness when, perhaps,
they merely noticed that something was both good and in possession of their
favorite natural property. And by virtue of their definitions, they believe
themselves to have identified the good. “That a thing should be good, it has been
thought, means that it possesses this single property: and hence (it is thought) only
what possesses this property is good.”50
Again, nothing in this passage proves that NF1, NF2, or NF3 is the correct
interpretation of the naturalistic fallacy, but they are consistent with this
passage.51
This passage also suggests that my interpretation of Moore’s critique
of Bentham is right.
5. Spencer’s Evolutionary Ethics
Moore accuses Spencer of committing the naturalistic fallacy in
developing his theory of evolutionary ethics. Moore accuses Spencer of making a
number of different errors, and he is not clear about which, if any, is the
naturalistic fallacy. I have found two distinct errors which can be taken to be
instances of NF2 or NF3. In this section, Moore also gives another explanation of
49
Principia Ethica, p. 90. 50
Principia Ethica, p. 90. 51
In fact, it is the first passage that fits well with NF1. I had been on the verge of abandoning it
until re-reading this passage.
124
the naturalistic fallacy—one that appears to make it a definist fallacy after all.
Given the surrounding context and previous explanations, I argue that Moore’s
words should not be taken strictly literally in this instance, and that my
interpretations still hold.
The first error Moore accuses Spencer of committing is possibly
identifying being more evolved with being better or higher. Moore is unsure
whether Spencer actually this identification because he uses the words ‘better’
and ‘more evolved’ inconsistently. But Moore claims that Spencer is likely
influenced by commission of the naturalistic fallacy. Though we do not find an
explicit definition of being better, we can infer that Spencer relies on one.
Moore writes:
It may . . . be true that what is more evolved is also higher and
better. But Mr. Spencer does not seem aware that to assert the one
is in any case not the same thing as to assert the other. He argues
at length that certain kinds of conduct are ‘more evolved,’ and then
informs us that he has proved them to gain ethical sanction in
proportion, without any warning that he as omitted the most
essential step in the proof.52
If this is an accurate description of Spencer’s approach, he has inferred from the
co-instantiation of being more evolved and being (morally) better that these are
52
Principia Ethica, p. 101.
125
the very same properties. Once this identity is presumed, he can claim that he has
proved that conduct gains ethical sanction in proportion to its being more evolved.
This would be analogous to claiming that pleasantness is the good based on the
identity of the properties pleasantness and goodness. The inference to the claim
about the good is valid, but relies on a bad inference to the identity claim. Here,
the claim that conduct is better in proportion to its being more evolved is a claim
about the proper end of ethics. It is derived from a claim that being better is
identical with being more evolved. And this identity is the result of the
naturalistic fallacy.
The fallacy, in this case, is in the inference from instances of conduct that
are more evolved53
and better (presumably, better than instances that are less
evolved) to the conclusion that being better simply is being more evolved.
Spencer may have made the intermediate step of presuming that all instances of
being better were instances of being more evolved. If so, he may be characterized
as committing the fallacy I have labeled NF2. Otherwise, it would be NF3.
It is only when we view Spencer’s argument in light of a particular model
of ethical reasoning that Moore’s claims that the theory rests on the naturalistic
fallacy become intelligible. This model begins with observations of the moral
status of particular acts and the other properties those acts instantiate:
1) These instances of natural property N are good.
53
Which may or may not mean that they are more conducive to life. This is unclear to me.
126
From these observations, general principles about the relation between the moral
and natural (or metaphysical) properties may be derived:
2) All instances of N are good.
If the naturalist then intuits that N is the good (or the end), as Sidgwick does,
there is no naturalistic fallacy. But, if instead of that permissible intuition, the
naturalist first defines the moral property and then works back up to the claim
about the good, then the fallacy is committed:
3) N is identical with goodness.
Once the naturalist has taken this step in furtherance of his claim about the good,
he has committed the fallacy. Nonetheless, most naturalists make the further
claim about the good
4) Since N is identical with goodness, N is the good.
The inference from N=G to ∀x(N↔G) is valid. But the derivation of the identity
is not, so the reasons for adopting the theorist’s views on the good are bad ones.
The second error Moore accuses Spencer of committing is to argue in
favor of his evolutionary ethical theory by appeal to hedonism. In order to prove
his thesis that ‘Conduct is better in proportion as it is more evolved’ he relies
upon the thesis that ‘Life is good or bad, according as it does, or does not, bring
about a surplus of agreeable feeling.’54
Here, Moore claims that Spencer is not
embracing the naturalistic identification of being better with being more evolved,
54
Principia Ethica, p. 102 quoting Spencer, Data of Ethics, Ch. II, § 10.
127
but, possibly, identifying goodness with pleasantness. That he does embrace such
a view becomes evident when Spencer claims that “No school can avoid taking
for the ultimate moral aim a desirable state of feeling called by whatever name—
gratification, enjoyment, happiness. Pleasure . . . is an inexpugnable element of
the conception.”55
Here, we see Spencer explicitly taking a natural property to be
the proper end of ethics. We know, based on the analysis in sub-section (4),
above, that Moore takes most theories that identify the good with some natural
property commit the naturalistic fallacy. Recall, only Sidgwick seems not to
have, and he avoids the naturalistic fallacy by arriving at his conclusion about the
good by avoiding defining goodness as identical with a natural property. Whether
Spencer can be accused, then, of committing the naturalistic fallacy in his
embrace of hedonism will depend on how it is that he comes to accept his thesis
about the good. Moore does not specifically accuse him of committing the fallacy
here because Spencer’s loose use of language prevents us from knowing how he
arrived at his hedonic thesis.56
If, however, he did so from inferring from
particular instances of pleasantness being good to the identification of
pleasantness and goodness, then he would have committed the naturalistic fallacy.
55
Principia Ethica, p 103, quoting Spencer, DE, § 16. 56
Principia Ethica, p. 100-101, p.105.
128
Once the identity is presumed, it is a short step to the conclusion that pleasure is
the sole good.57,58
There is a passage in Moore’s discussion of Spencer, however, that casts
some doubt upon my interpretation of the naturalistic fallacy. In giving a capsule
summary of his criticisms of naturalistic ethics, Moore writes that certain ethical
views—naïve naturalisms—owe their influence to the naturalistic fallacy. This
fallacy, he writes, “consists in identifying the simple notion we mean by ‘good’
with some other notion.”59
I addressed this particular quotation in Chapter 2, in
my discussion of Baldwin’s interpretations of the naturalistic fallacy. I argued
that “consists in” does not mean “is identical to.” In short, defining goodness is
necessary for committing the naturalistic fallacy, but it is not sufficient for doing
so. For example, if someone failed to recognize that goodness was indefinable,
and simply intuited a definition, there would be no fallacy. Of course, we would
have little reason to believe their views about the good. Their views about the
57
One of the reasons Moore does not accuse Spencer of definitely having committed the
naturalistic fallacy in embracing hedonism is that, sometimes, he asserts that being productive of
more life is also an end. If this is the case, he does not take pleasure to be the sole end, and so, he
might not claim pleasure is the good. If he does not make this claim, it would be unfair for Moore
to accuse him of deriving that conclusion from a definition of goodness arrived at via the
naturalistic fallacy. 58
Moore claims Spencer has made another error in supposing that being a law of nature makes a
thesis respectable. Here, again, we see that Spencer’s claim might rest on the naturalistic fallacy.
The reasoning is exactly analogous to the cases of hedonism and being more evolved. A particular
thing (so called ‘directional evolution’) is respected. It is also presumed to be a law of nature.
From this is deduced that being a law of nature is to be respected (clearly false, and an instance of
the naturalistic fallacy). Then from this, the conclusion that laws of nature are all respectable
follows. This example, though a mere side-note to Moore’s discussion of Spencer, might be a best
proof that to commit the naturalistic fallacy is to make an inference error of the sort I have
characterized in NF1-NF3. 59
Principia Ethica, p. 109.
129
good are based on a definition of good that (1) we have reason to believe is false
(based on the open question argument) and (2) that, itself, is the result of an
invalid inference. I take my interpretation of the naturalistic fallacy to be
consistent with Moore’s claim that the fallacy consists is identifying goodness
with some other property because my view is that the naturalistic fallacy is to
make that identification as the result of a particular pattern of inference.60
I do not claim that my interpretation of the naturalistic fallacy is the only
one consistent with the errors Moore accuses Spencer of making. But it is
consistent with it, and it is also consistent with each additional instance of the
naturalistic fallacy’s uses and explanations in Principia Ethica. No other
interpretation I have found can claim this virtue.61
60
Moore takes all naturalistic theories that make such an identification to do so on grounds of this
particular argument form. He does not recognize any other way to support a property
identification for goodness. I suggested that someone might intuit one, but then Moore could
claim that she would have provided no reasons to believe the identification. In short, you can
intuit the good, but not goodness. This may provide a legitimate source of the criticism that
Moore begs the question against the naturalist. 61
With the exception of Moore’s comments about Mill, the only other discussions of the
naturalistic fallacy appear in his discussion of metaphysical theories of ethics (pp. 165 and 169)
and in his discussion of Beauty (p. 249), in his chapter on the Ideal. Each of these instances is
compatible with my view. With respect to Beauty, the fallacy appears to occur in the inference
that since particular instances of beauty have particular subjective effects upon people, beauty is
identical with such effects. Then, having committed the fallacy, the subjective aestheticist
compounds his error by deriving that the beautiful is all and only that which produces such effects.
130
6. Hedonism and Mill
The last Moorean claims about philosophers who have committed the
naturalistic fallacy that I wish to examine come from his comments about
hedonism and John Stuart Mill’s arguments in favor of hedonism. The first
passage is:
Moreover, the very commission of the naturalistic fallacy involves
that those who commit it should not recognize clearly the meaning
of the proposition ‘This is good’—that they should not be able to
distinguish this from other propositions which seem to resemble it
. . . .62
Just as Moore criticized Spencer for having engaged in loose writing, we may
criticize Moore on the same ground here. On any interpretation of the naturalistic
fallacy that could be derived from Principia Ethica, whether my own view or
views I dismissed previously, the commission of the fallacy need not involve not
recognizing the meaning of the proposition ‘This is good.’ It could very well be
that you commit the fallacy while clearly recognizing the meaning. It seems
likely that the confusion is because of failing to express properly the difference
between a proposition and the sentence that expresses it. But by looking at the
further claim that naturalists fail to distinguish that proposition—I can only
62
Principia Ethica, p 113
131
assume he means the proposition that something is good, by which he means
something has the property goodness—from others which seem to resemble it, we
may find a place for the naturalistic fallacy. His claim is that someone who
commits the naturalistic fallacy, say by claiming pleasure is identical with
goodness, does so because he cannot distinguish between the propositions that
pleasure is good (meaning pleasure has the property goodness) and pleasure is
good (meaning pleasure is identical with goodness). Earlier, I showed how this,
though looking like confusing the ‘is’ of predication with the ‘is’ of identity, can
be captured by NF2.
Finally, we come to the last significant discussion of the naturalistic
fallacy in Principia Ethica. Moore accuses Mill of having committed the
naturalistic fallacy in his proof that the desired is the good. The argument pattern
is the same here that we have seen in Bentham and in Spencer, but it is more
difficult to spot. 1) Mill claims that good means desirable, then (2) asserts that
what is desirable is what is desired. From this, we get, via the naturalistic fallacy,
(3) the definition that being good is identical with being desired. Therefore, (4)
the desired is the good. It is hard to spot this pattern, in part because Moore
assaults Mill for making the famously bad analogy between desirable and
132
visible.63
Most of Moore’s demonstration that Mill commits the naturalistic
fallacy is mixed up with his proof that it is a mistake to take step (2).
Step (1) looks like it would be problematic, but is permissible for Mill to
define good as desirable, if the definition is of good as a means to the end of
ethics rather than good as an end. This is analogous to what Moore would have
allowed for Bentham with respect to right.64
He may also have begun from mere
observation of things that are both good and desirable. Had he supposed that all
good things were desirable, he would have an adequate first step toward the
fallacious proof that goodness is identical with being desired.
At step (2), Mill makes the claim that what is desirable simply is what is
desired. “[T]he sole evidence it is possible to produce that anything is desirable,
is that people do actually desire it.”65
Moore seems to assert that Mill commits
the naturalistic fallacy in this step. “Well, the fallacy in this step is so obvious,
that it is quit wonderful how Mill failed to see it.”66
If he has, I take this to be an
instance of the fallacy occurring within another instance of the fallacy. Moore
accuses Mill of having argued from the fact what is desirable is actually desired to
the claim that desirable means desired. This is to infer from particular instances
of things being desirable and being desired that the properties are identical—a
63
I do not reproduce this here, in part because I am not convinced that Mill really makes this error. 64
See text accompanying fn 41 and 42, supra. 65
Mill, Utilitarianism (1861), quoted by Moore, Principia Ethica, p. 118 and reprinted in John
straightforward instance of NF3. Alternately, since Mill takes the only proof that
a thing is desirable to be that it is desired, we might make the stronger claim that
all things that are desirable are desired.67
To infer from this that they are identical
would be to commit the naturalistic fallacy in its NF2 configuration. Either way,
it is clear that to be desirable in the sense of being worthy of desire is a very
different thing than being desired. Moore could have left his criticism there, but
finding an instance of the naturalistic fallacy as support for a premise in another
argument that also relies upon the naturalistic fallacy is rather clever.68
If Mill’s identity claim between desirable and desired holds, then we can
substitute the property being desired for being desirable in the original
proposition, yielding that good means being desired. (∀x (Gx → Dx)). By
committing a second naturalistic fallacy, in NF2 form, Mill can (3) infer the
identity between goodness and being desired. Moore writes “The important step
for Ethics is . . . the step which pretends to prove that ‘good’ means ‘desired.’
Moore is adamant that this step be avoided.
67
In truth, I am not sure what direction the arrow would go here—the implication might go either
direction. 68
I suppose this nested naturalistic fallacy claim only holds if my claim about the naturalistic
fallacy is true.
134
§ 3) Conclusion
In this chapter, I argued that Moore’s critics make an error when they
accuse Moore of using the naturalistic fallacy as proof that naturalistic theories
are false. I demonstrated, by text and with examples, that Moore does not take
accusations that a philosopher has committed the naturalistic fallacy to do
anything more than undermine our confidence in her reasons for her conclusions
about the proper end of ethics. My claim relies on a model of moral
argumentation that Moore takes naturalists to make—one that involves proving
(fallaciously) that goodness is identical with some natural property, and then
inferring from that identification that their chosen property is the sole good.
I have also offered an interpretation of Moore’s naturalistic fallacy that
more closely fits the text of Principia Ethica than those interpretations I
considered in Chapter 2. On this view to commit the naturalistic fallacy is to
make an invalid inference from the co-instantiation of two properties, to their
identity. It is a general purpose fallacy, affecting any property identification, not
just definitions of goodness. Most significantly, the interpretation presented is
consistent with each of Moore’s various descriptions of the fallacy, and with each
of his demonstrations that other theorists have committed it. When conjoined
with the model of reasoning I highlighted, the view accounts for his claims that
even if we reject the definitions of good naturalists offer, we still need
135
independent argumentation to prove that they are wrong about the proper end of
ethics. The naturalistic fallacy, as I have described it, is well suited to play the
role Moore assigned to it. It neither proves too much—that a theory is false for
having committed it—nor too little.
It is not, however, immune to criticism. While the view I have described
is the most comprehensive view of the naturalistic fallacy, it may still be subject
to claims of question begging. In particular, when naturalists do not explicitly
assert that they are starting from any particular observations about good things,
Moore presumes that they are. This is the starting point of each instance of the
naturalistic fallacy. But whether this is a fair presumption is unclear. Can a
coherent story can be told that avoids assuming that goodness is indefinable when
we accuse naturalists of committing the naturalistic fallacy? This will be one
subject of the next chapter.
136
Chapter 5
Moore’s Proof that ‘Good’ Refers to a Simple, Non-natural Property1
I. Introduction
I have devoted a significant portion of this dissertation to trying to show
that to commit the naturalistic fallacy is not simply to define an indefinable
property. Moreover, I have claimed Moore does not wield the naturalistic fallacy
as a weapon, cutting down theories that do define goodness on grounds of having
committed the fallacy. Even if he does, however, we might find that he was
justified in doing so—that is, that he does not beg the question against the
naturalists—if he had a reason to believe all naturalistic definitions were false. I
believe that Moore had such a reason. Disambiguating the open question
argument from the naturalistic fallacy shows us that Moore has an argument that
is sufficiently robust to ground even the claims that Frankena attributes to him.
Most commentators take the open question argument to be Moore’s proof
that every naturalistic definition fails. Moore’s actual claim about goodness is
that it is a simple and indefinable property. In Chapter 2, I provided an
examination into what property it is that Moore claims is simple and indefinable.
1 Throughout this chapter, I use the words ‘denotation,’ ‘denoting,’ or ‘denote’ in the non-Millian
sense of the words. By ‘x denotes y’ I mean the word ‘x’ refers to the property or object ‘y.’ In
this regard, I use the words ‘denote’ and ‘connote’ in the reverse of the Millian fashion. This is to
avoid having readers thinking of ‘connotation’ or ‘meaning in the head’ when I mean to draw
attention to the referent, or meaning in the world.
137
This was followed by an examination into what it means to be simple and
indefinable. This chapter builds on that material, demonstrating that there are
several different arguments supporting Moore’s claim, some of which are not
easily refuted. I begin in section II with a close examination of Moore’s text, and
several defective interpretations of it. In section III, I offer a reconstruction of
Moore’s arguments that maintains high fidelity to the text. It should be noted at
the outset that I do not find anything that fills the role the open question argument
is usually held to perform. While I do offer an argument that might show that any
definition of ‘good’ in natural terms fails, even those that purport to identify
goodness with a simple natural property,2 I find that it was not Moore’s intention
to show this using the open question argument. Interestingly, the words “open
question” never even appear in the section wherein Moore presents his proof that
goodness is simple and unanalyzable. The first use of this phrase is in the final
sentence of § 14, and it is not taken to be a proof of the indefinability of goodness,
but as an example of defective reasoning about Ethics.
2 So not true definitions, according to Moore. Recall that definitions enumerate the parts of
complex properties, so a simple statement of property identity is not a definition.
138
II. Literature review
A. Principia Ethica § 13
Moore’s entire proof that goodness is simple and indefinable is contained
in § 13 of Principia Ethica. Since in the rest of this chapter I will rely heavily on
the text, I include it, nearly in its entirety, interspersed with some commentary
about what I perceive as the appropriate divisions and individual arguments.
Moore begins by delineating the possibilities for what sort of property
goodness might be:
[I]f it is not the case that ‘good’ denotes something simple and
indefinable, only two alternatives are possible: either it is a
complex, a given whole, about the correct analysis of which there
could be disagreement; or else it means nothing at all, and there is
no such subject as Ethics.3
Goodness is simple (and hence, indefinable), or it is complex, or it has no
meaning. By this third option, he means that there is no one thing that ‘good’
denotes. It might be that there are many things.4 It might be best to say that there
3 Principia Ethica, p 66. He makes the same point again as follows: “There are, in fact, only two
serious alternatives to be considered, in order to establish the conclusion that good does denote a
simple and indefinable notion. It might possibly denote a complex, as horse does; or it might have
no meaning at all.” p. 67. The rest of the cited passages in this chapter are from § 13 of Principia
Ethica, appearing on pp. 66-68, unless otherwise indicated. 4 Or (perhaps) none, though I do not know whether Moore would offer this as an option. In his
early works, words that have no referent in the realm of Existence always have one in the realm of
Being (more or less a Meinongian system). While this does not rule out Moore’s acceptance of
139
is no “concrete” meaning. This would explain his claim that if goodness had no
meaning there is no such subject as Ethics. After all, if there were no single thing
that is the sole simple object of thought relevant to Ethics, then the discipline
would rest on an error.5
The rest of § 13 of Principia Ethica is devoted to showing that goodness
is not complex and that goodness is not meaningless. The open question
argument has been variously identified as performing both of these tasks, though
as we shall see, it does not. First, let’s consider Moore’s proof that goodness is
not complex.
The hypothesis that disagreement about the meaning of good is
disagreement with regard to the correct analysis of a given whole,
may be most plainly seen to be incorrect by consideration of the
fact that, whatever definition may be offered, it may always, be
asked, with significance, of the complex so defined, whether it is
itself good.
This looks like what we would expect from something called an ‘open question’
argument. To prevent confusion of my analysis of this argument with others’
treatments, I will call this the Significant Question Argument. It appears that
words with no referent, without seeing any textual support for the position, I hesitate to attribute it
to him. 5 This is a bit like saying that if ‘rock’ meant both rocks and ore, there would be no such thing as
Geology (let alone Metallurgy). It is not strictly true, but the idea is clear enough. Such a
situation would undermine claims that the discipline was a science.
140
Moore proposes that if we can ask meaningfully, “with significance” whether the
definiens of a complex definition of goodness is good (whether it has the property
goodness—this seems to be a question about predication, not identity), then the
definition is wrong. I take it that only an open question can be asked with
significance. We will consider what it means to be a significant question, below.
Moore presents something strangely different from what is suggested by
this text when he gives us an example of such a complex definition and the
operation of his argument to it. The example he considers is the definition of
goodness as that which we desire to desire.
[I]t may easily be thought, at first sight, that to be good may mean
to be that which we desire to desire. Thus if we apply this
definition to a particular instance and say ‘When we think that A is
good, we are thinking that A is one of the things which we desire
to desire,’ our proposition may seem quite plausible. But, if we
carry the investigation further, and ask ourselves ‘Is it good to
desire to desire A?’ it is apparent, on a little reflection, that this
question is itself as intelligible, as the original question, ‘Is A
good?’—that we are, in fact, now asking for exactly the same
information about the desire to desire A, for which we formerly
asked with regard to A itself. But it is also apparent that the
meaning of this second question cannot be correctly analysed into
141
‘Is the desire to desire A one of the things which we desire to
desire?’: we have not before our minds anything so complicated as
the question ‘Do we desire to desire to desire to desire A?’
This passage has been recognized as the one that Virginia Woolf cited as
making her head spin.6 This passage, which is proposed as an example of the
significant question argument described in the previous passage, seems to do two
other things. First there appears to be an argument that looks very much like the
significant question argument. The question “Is it good to desire to desire A’
appears to be an open question. One might think that Moore is suggesting that if
it is an open question, then the definition of goodness as that which we desire to
desire is wrong.7 Taken this way, the passage seems to be a straightforward
example of the significant question argument. But notably absent is a conclusion
that indicates that the openness or significance of the question proves that the
properties are distinct. Instead, Moore launches into what appears to be another
argument. This argument turns on the apparent complexity of the questions we
6 Story recounted in T. Regan, Bloomsbury’s Prophet, pp 197-98.
7 We shall see that this is the interpretation given by Brian Hutchinson in G.E. Moore’s Ethical
Theory, (Cambridge 2001), at p. 30. It is also close to that presented by Stephen Darwall, Allan
Gibbard, and Peter Railton in “Toward Fin de Siecle Ethics: Some Trends,” The Philosophical
Review, Vol. 101, No. 1 (January 1992). Hutchinson takes this to be the standard interpretation.
This standard interpretation ignores Moore’s verbal cues that suggest the multitude of arguments I
have identified, and instead treats the entirety of section 13 as presenting one argument that
attempts to prove that goodness is not identical with any named or defined property. It takes the
partial argument this footnote refers to and combines it with another partial argument that Moore
takes as his evidence of a different proposition altogether. An enduring criticism of Moore’s
argument is that it is “muddled” (Hutchinson, p. 29) and “accident-prone” (Darwall, Gibbard, &
Railton, p. 115). As muddled as Moore’s presentation of his argument appears, taking parts
designed for distinct purposes and combining them into some new argument never explicitly made
by Moore does not seem a sound strategy for clarifying the text.
142
are considering. It appears that Moore is claiming that when we substitute for
‘good’ in “Is it good to desire to desire A?” the words that refer to the property
suggested as a definition (the complex), the resulting question is more
complicated than the one we started with. This is evidence that the questions are
distinct. If so given the nature of language, the words substituted for ‘good’ do
not mean the same as ‘good’. That is, they refer to a different property. Call this
the More Complicated Question Argument.
There is yet a third argument in § 13, also in support of the claim that
goodness is not identical with any complex property.
Moreover any one can easily convince himself by inspection that
the predicate of this proposition—‘good’—is positively different
from the notion of ‘desiring to desire’ which enters into its subject:
‘That we should desire to desire A is good’ is not merely
equivalent to ‘That A should be good is good.’ It may indeed be
true that what we desire to desire is always good; perhaps, even the
converse may be true: but it is very doubtful whether this is the
case, and the mere fact that we understand very well what is meant
by doubting it, shews clearly that we have two different notions
before our mind.
Here, Moore appears to claim that our ability to doubt whether it is true
that some complex property is always good and that every instance of goodness is
143
also an instantiation of that complex property demonstrates that that property is
not identical with goodness. In other words, if we can doubt a statement that
purports to tell us that the good is a complex property, then we have, when
considering that statement, two different notions in mind. Call this the Argument
From Doubt. In the lead up to this argument, found in the first sentence of this
passage, Moore suggests that mere inspection can convince us that goodness and
the complex property suggested as a definition are distinct. This may, indeed, be
a separate argument to the same conclusion. Call this the Argument From
Inspection.
Having presented at least three, and possibly four, arguments that
goodness is not a complex property, Moore turns his attention to the possibility
that it is no property, or rather, that ‘good’ has no meaning. He begins by
describing a mistake some naturalists make. It is the mistake of inferring from
what appears to be a “universal ethical principle”—like “whatever is called good
seems to be pleasant”—an identity—that ‘good’ and ‘pleasant’ refer to a single
notion (that is, that goodness is identical to pleasantness). This is the mistake I
have identified as the naturalistic fallacy. But instead of arguing that the
identification is always false, as we would expect, and as he has been consistently
interpreted to do,8 he says:
8 Consider, for example, Fred Feldman’s paper “The Open Question Argument: What it isn’t; and
What it is,” Philosophical Issues, 15, Normativity, 2005, pp. 22-43. Feldman argues that the entire
passage which Moore claims to be his argument that ‘good’ is not meaningless is really simply an
144
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 recognise 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. Every one does in fact understand the question ‘Is
this good?’ When he thinks of it, his state of mind is different from
what it would be, were he asked ‘Is this pleasant, or desired, or
approved?’ It has a distinct meaning for him, even though he may
not recognise in what respect it is distinct. Whenever he thinks of
‘intrinsic value,’ or ‘intrinsic worth,’ or says that a thing ‘ought to
exist,’ he has before his mind the unique object—the unique
property of things—that I mean by ‘good.’
This is most definitely not an argument that the identification of goodness
and simple natural properties—like pleasantness—is a mistake. It is, instead, an
argument that when we consider any proposed definition (and I supposed Moore
argument that no theory identifying goodness with a simple natural property, like pleasantness, is
true. Instead, what Moore does here is to remind us that when we consider whether any proposed
definition ‘gets it right,’ we are considering the same property as definiendum. This is proof that
we have a single thing in mind when we say ‘good.’
145
chooses simple properties as definiens to prevent the risk of confusion) and
consider the property alleged to be identical with goodness, the property we are
comparing it against is always the same property. That is, we have some
particular property in mind when we use the word ‘good’ (in its moral sense).
Call this the Uniqueness Argument.
If we have a unique property in mind when we use the word ‘good,’ from
which we may infer that there is such a property,9 and goodness (that property) is
not a complex property, then it is a simple property. Simple properties are
indefinable, hence goodness is indefinable. I take it that this is the entire
argument from § 13, with all its subsidiary arguments. It is not clear whether
Moore took himself to be making more than one subsidiary argument in his
discussion of the claim that goodness is not complex. I do not hope to resolve this
question, but rather to see what is there, regardless of what was intended. It is
altogether possible that Moore took all of the subsidiary arguments for that claim
to be different ways of expressing the same point, or that they amount to the same
argument by virtue of turning on the same features of language and metaphysics.
In section III, I explicate the overall argument that goodness is simple and
indefinable and consider the five subsidiary arguments I have identified in support
of the overarching argument. Before turning to that task, though, I consider a
9 Whether this inference is any good will be considered below.
146
number of alternative interpretations of Moore’s proof and compare them to the
text as I have just suggested it be read.
B. Some Problematic Interpretations of the Open Question Argument
The first endemic problem in commentary on the open question argument
is that philosophers are often not careful to distinguish it from the naturalistic
fallacy, or take one to be a consequence of the other.10
That the naturalistic fallacy is commonly confused with the open question
argument is proved by Baldwin, who writes in his discussion of the open question
that “[i]t is notable that the section in which [Moore] advances it is the only part
of the opening discussion in Principia Ethica of the naturalistic fallacy which
does not come straight from The Elements of Ethics.”11
Baldwin characterizes the
section of Principia Ethica wherein Moore tries to prove that goodness is simple
and indefinable as part of his discussion of the naturalistic fallacy. I have shown
that the open question argument is distinct from the naturalistic fallacy, both in
10
Another is that many are not actually discussions of Moore’s open question argument, but rather
applications of something akin to the Moore’s open question argument applied against theories
Moore did not explicitly consider. Gilbert Harman, for example, describes the OQA and claims
that it is an invalid argument. Harman, Gilbert, The Nature of Morality, (Oxford, 1977) p. 19. He
notes that someone ignorant about the identity of water and H2O would agree that it is not an open
question whether water is water, but that it is an open question whether water is H2O. So here we
have a perfectly good identity that appears to be disproved by the OQA. But is this the OQA?
The analog for ethics would be that it is not an open question whether goodness is identical with
goodness but it is an open question whether goodness is identical with X, where X names a natural
property. But it seems pretty clear that Moore does not ask whether it is a significant question
whether the identity statement holds, but rather whether an object that instantiates X also
instantiates goodness. In fairness, Harman does not claim to be explicating Moore, but merely
pointing to a kind of anti-naturalist argument. 11
Baldwin, p. 87.
147
argument structure and intended conclusion. But Baldwin here asserts that
Moore’s presentation of the open question argument (by which he means his
proof that goodness is simple and indefinable) is part of his discussion of the
naturalistic fallacy. This might explain why some take it to be a simple definist
fallacy. If we view section 13 as proving that all natural definitions of goodness
are false, including those purported definitions that suggest that goodness is
identical with a simple natural property, then surely offering such a definition
would be a mistake. This approach simply gives the attempt to define goodness
the name ‘naturalistic fallacy.’ But this approach makes Moore’s claims about
the naturalistic fallacy trivial, and makes it a complete mystery why Moore would
have written so much by way of explaining the naturalistic fallacy—more, in fact,
than he wrote in § 13.
In addition to conflating Moore’s two major arguments, critics make other
mistakes in describing the open question argument. The next few subsections
provide a cross-section of some of the mistakes typically made.
1. Prior
Before turning to Moore’s proof that goodness is indefinable, we have to
return, briefly, to Prior’s discussion of the naturalistic fallacy. His discussion
demonstrates an error common to those who take the naturalistic fallacy to be
strictly a definist fallacy. Interestingly, he makes this error even though at least
148
one of his characterizations of the naturalistic fallacy is virtually identical to one
of mine, NF1. Recall, Prior offers two interpretations of the naturalistic fallacy,
one highlighting the linguistic ramifications of the fallacy—that the naturalistic
fallacy is the assumption that because the words ‘good’ and ‘pleasant’ necessarily
describe the same objects, they attribute the same property to them—and one
focusing on the properties themselves—that the naturalistic fallacy is the
assumption that because some property is necessarily coextensive with goodness,
it is identical with goodness.12
Despite adopting a position that looks very much
like mine, he adopts a different role for the naturalistic fallacy, and in so doing,
adopts a mistaken view of the relation of the naturalistic fallacy to Moore’s proof
that goodness is indefinable.
Prior takes Moore’s proof that goodness is indefinable to be support for
his claims about the naturalistic fallacy. That is, Prior makes the same error as
Baldwin, above. Prior claims that when the naturalistic fallacy is committed, the
philosopher who has committed it fails to realize that ‘good’ and some other
adjective, say, ‘pleasant’ may be applicable to the same things and yet not denote
the same property. He takes the argument to show that two adjectives may
accurately describe the same object without denoting the same property in that
object.13
Stated this way, nothing could be more obvious. Why the open question
12
See this dissertation, Chapter 2, and Prior, Logic and the Basis of Ethics, p 1. 13
Most people take Moore’s proof that goodness is indefinable to be the open question argument.
Prior does not frame his interpretation of Moore’s argument in terms of open questions.
149
argument (or a variant of it) is needed as support for this proposition is unclear.
The mere fact that ‘black’ and ‘clicky’ accurately describe the keyboard upon
which I typed this sentence without referring to the same property of that
keyboard proves this proposition. The naturalistic fallacy, when taken in Prior’s
property-referring form, is similarly without need for support. Moore’s proof that
goodness is indefinable goes no way toward supporting the claim that inferring
the identity of properties based only on their coextension is an invalid maneuver.
Recall, the naturalistic fallacy is supposed to be a perfectly general fallacy, and is
not simply limited to identifications of goodness. We need only to think about the
coextension and identity relations to see that this is true, or reflect on a couple of
less controversial examples. If the point of the open question argument is merely
to demonstrate that coextension and identity are different relations, it makes no
sense at all for Moore to attempt to show this by reference to the property
goodness—a property about which there is so much contention. That goodness is
indefinable shows only that an argument that concludes that goodness is identical
with another property has a false conclusion, but tells us nothing about the
reasoning. Given that Prior takes the naturalistic fallacy to be about the inference,
not the definition of goodness, it is unclear why he views the open question
argument as support for the naturalistic fallacy.
150
Nevertheless, looking at Prior’s version of Moore’s proof may be
instructive. Prior describes the basis of Moore’s proof that goodness is
indefinable as follows:
If the word ‘good’ and the word ‘pleasant’ apply to the same
things, but do not attribute the same quality to them, then to say
that what is pleasant is good, or what is good is pleasant, is to
make a significant statement, however obvious its truth may appear
to many people. But if the word ‘good’ and the word ‘pleasant’
not merely have the same application but the same connotation or
‘meaning’—if, that is to say, the quality of pleasantness is identical
with the quality of goodness—then to say that what is pleasant is
good, or that what is good is pleasant, is to utter an empty
tautology . . . for both statements are on this supposition merely
ways of saying that what is pleasant is pleasant.”14
This is more of an observation than an argument, but one can be made
based on what is said here. The idea is that if we take any proposed definition, a
statement to the effect that what possesses the proffered properties is good is a
significant statement. And if this is true, then the proffered definition is wrong.
Prior is careful to point out that Moore thinks this approach shows that complex
definitions fail. But then Prior goes wrong by claiming that Moore also applies
14
Prior, p. 2.
151
this argument to simple definitions (definitions in terms of a simple property
only).15
His claim, then, is that the open question argument attempts to show that
goodness is not identical with any simple or complex natural property. Rather,
the open question argument (or the significant question argument) is a mere part
of Moore’s overall proof that goodness is simple and indefinable. As I have
shown, above, it is one argument used in support of the claim that goodness is not
identical with a complex property. But it does not appear to be Moore’s entire
proof.
Moreover, Prior’s argument is not exactly Moore’s open question
argument. In fact, as I have noted, it is not clear just what really is the open
question argument.16
Note that there is no question asked and nothing that is left
“open.” It may, however, turn on the same intuitions about language at evidence
in what I have called the significant question argument. In this case, the
significant statement made can be viewed in the same light as an open question.
The openness of the question “is it good?” is a result of the meaningfulness of the
assertion. If the claim that what was desirable to desire is good was a mere
tautology, the question “is it good?” would not be an open question.
15
Prior, p. 3. 16
Consider, for example, Fred Feldman’s paper, “The Open Question Argument: What it Isn’t;
and What it is,” in Philosophical Issues, 15, Normativity, 2005, pp. 22-43, wherein Feldman finds
no fewer than four unique arguments that might claim this title. Notably, none of these actually
turns on the identification of any open questions. He also offers one more that is in the spirit of
Principia Ethica, but certainly not to be found within it that does turn on an open question.
152
2. Hutchinson
Like Prior, Hutchinson treats § 13 of Principia Ethica as though it
presented one unified argument that no natural definition (or identity statement, in
the case of simple definitions) is true, but unlike Prior, he notes that reading the
passage this way is to disregard Moore’s verbal cues.17
Nonetheless, he claims
that Moore’s proof that ‘good’ is not meaningless is better read as a proof that
‘good’ is not the name of any simple natural property, and all the arguments of §
13 preceding that passage, with the exception of the significant question
argument, are non sequitors.18
To note that there is a variety of arguments being
offered but to dismiss all but the first and last as trivial seems almost worse than
getting them wrong!
3. Sturgeon
Nicholas Sturgeon advocates something akin to the standard interpretation
of the open question argument in “Moore on Ethical Naturalism.”19
He
summarizes the open question test thusly:
[Moore] assumes that [any proposed identification of goodness
with a natural property] can be correct only if the two terms used
in stating it are synonyms . . . . Substituting synonyms for
17
Brian Hutchinson, G.E. Moore’s Ethical Theory (Cambridge 2001) pp 28-38. 18
Hutchinson, p. 29. 19
Nicholas Sturgeon, “Moore on Ethical Naturalism,” Ethics 113 (April 2003): 528–556, at 533.
153
synonyms should not change the thought that a sentence expresses.
But, as he cleverly notes, when we substitute into the very
statement of these proposed definitions, it seems obvious that in
thinking that what we desire to desire is good we are not merely
thinking that what we desire to desire is what we desire to desire,
and that in thinking that pleasure is good we are not merely
thinking that pleasure is pleasant. 20
Sturgeon makes a couple of errors in his analysis of Moore’s argument.
The first is in thinking that the open question test is one single argument. As I
have stated above, the open question argument is Moore’s argument that goodness
is simple and indefinable, not his argument that no natural definition is correct. I
contend that more is needed to reach that conclusion. My arguments to this effect
are found in the final chapter of this work. He also combines two distinct
arguments, treating the argument that goodness is indeed a property (that “good”
is not meaningless) with the argument that goodness is not complex.
The more significant error is that his explanation of the open question
argument suffers from the defect of misinterpreting the substitution Moore
engages in when he considers whether the sentences, differing only in the
substitution of purported synonyms, actually express the same thought. Here is a
proposed definition of good:
20
Ibid.
154
[1] Pleasure is good.
Sturgeon believes that when we substitute the proposed definiens for
‘good’ we get:
[2] Pleasure is pleasant.
Sturgeon is right that these two expressions do not express the same
thought. He claims that [1] is the statement of the definition. This is, of course,
an identity statement. But [2] is not an identity statement. It is a predication of
pleasantness to pleasure. The sentence that should result from a substitution of
the definiens for the definiendum is:
[2'] Pleasure is pleasure.
It is little wonder that a predication and an identity statement do not
express the same thought. After all, even with no substitution, [1] read as an
identity statement does not mean the same as [1] read as a predication.
Regardless of whether the argument turns on the substitution of synonyms in a
sentence like [1], we ought to be careful to treat the original and the modified
statements both as identity statements or both as predications.
4. Brandt
Richard Brandt describes a version of the open question argument.21
For
any two terms, P and Q, that are purported to have the same meaning, we may
21
Brandt, Richard B. Ethical Theory: the Problems of Normative and Critical Ethics (Prentice
Hall 1959), p. 164.
155
compose a question of the form: “Is everything P also Q?”22
This question asks
whether everything that instantiates P also instantiates Q. If this question is
intelligible or significant, then P does not mean the same thing as Q. Note that
this does not ask of a single P-ish object whether it is Q-ish, it asks whether every
P-ish thing is Q-ish. So on Brandt's version of the open question argument, the
test is one of coextensiveness. Since the test is phrased perfectly generally, P and
Q could be reversed—‘is everything P also Q’ and ‘is everything Q also P.’ If it
is intelligible to ask whether P and Q are coextensive, then P and Q are not
identical. This does not look much like the test Moore proposes in § 13.
Brandt suggests a criticism that applies to his version and might attach to
more ordinary versions of Moore's test too.23
While most critics emphasize that
the open question test fails because there are plenty of analyses, even good ones,
that fail the test,24
Brandt focuses on a different problem. There are presumed
failed analyses that will pass the test, at least for certain cognizers. For example,
many people will incorrectly find that there is no significant question when
'bachelor' and 'unmarried male' are substituted for P and Q, forgetting that
divorcees and widowers are unmarried males, but are not bachelors. While I find
that this version of the argument has little in common with the arguments to be
found in Moore’s text, it is worth noting that a successful version of the argument
22
So on this version, the test applies to all definitions, not just definitions of goodness. Brandt
generalizes the test in a way that it is not clear Moore would endorse. 23
Brandt, p. 165. 24
Ordinary language users might find that whether Water is H2O is a significant question . . .
156
will have to ensure both that successful analyses pass the test, and unsuccessful
ones fail.
III. Moore’s arguments explicated
We have seen that the “standard interpretation” of the open question
argument seems not to capture the full flavor of the argument Moore presented. It
combines elements of what appear to be distinct arguments, perhaps in an attempt
to create a single argument that accomplishes all that Moore’s naturalistic fallacy
has been interpreted to do. I hold that Moore’s proof is open to more nuanced
interpretations than those. By hewing closer to the text, I believe a better
interpretation of the proof that goodness is indefinable can be discerned. With
some luck, this new argument will be less susceptible to the traditionally asserted
critiques.
The first point to notice about the presentation of § 13, above, is that
Moore clearly intends his proof that goodness is simple and indefinable to be a
multi-step proof. It does not proceed by showing that every suggested definition
is wrong and to conclude from this fact that no such definition is possible, as has
been contended. Instead, Moore sets out the options—‘good’ denotes a simple
property, a complex property, or no property, i.e., it is a meaningless word—and
argues that the latter two are false. As I have pointed out, what is usually called
157
the ‘open question argument’ is Moore’s proof that ‘good’ does not denote a
complex property, or, as is more likely the case, one of the arguments to that
conclusion.
The overall argument may be reconstructed as follows:
Proof that ‘good’ denotes a simple, indefinable property
1. ‘Good’ denotes something simple and indefinable, or it denotes something
complex, or it denotes nothing at all.
2. ‘Good’ does not denote something complex.
3. ‘Good’ does not denote nothing at all.
4. Therefore, ‘good’ denotes something simple and indefinable.
This is a valid argument. The first premise presents the three possibilities
Moore recognizes for the meaning of ‘good.’ The proof is framed as about the
meaning of the word ‘good’ rather than about the identity of the property
goodness for the sake of clarity and consistency with Moorean convention. It is
particularly tricky to talk about the property, itself, if one of the possibilities under
consideration is that there is no such property. A property can be complex,
having constituent parts and thereby be subject to analysis or definition; or it can
be simple, and not admit of definition. Moore cannot engage in talk of properties
158
without or existence or being.25
He takes properties to be real things, the objects
of thought, and the bearers of meaning.26
So it is very difficult to talk of a
property having no meaning. But a word might refer to no particular property, or
no property unique to ethics. That would satisfy the third disjunct of (1).
The second premise is supported by the arguments I called the Significant
Question Argument, the More Complicated Question Argument, the Argument
from Doubt, and the Argument from Inspection. It is not clear whether Moore
intended these to amount to one argument, or as several arguments to the same
conclusion. I will treat them separately, noting their interactions when necessary.
Premise three is supported by a different argument, one that is clearly distinct
from what has commonly been called the open question argument. I have called
it the Uniqueness argument. I turn now to the individual sub-arguments.
25
This is a limitation we do not share with Moore, in part because we have adopted conventions
for talking about properties. He tries throughout the text to distinguish between existent objects
and non-existent universals, and one way he does this is by referring to universals as the meanings
of words off-set in single quotes. We could, easily enough, recast this argument in terms of
properties:
1. Either goodness is simple (and hence indefinable), or it is complex, or there is no such
property.
2. Goodness is not complex
3. It is not the case that there is no such property.
4. Therefore, goodness is simple (and hence, indefinable). 26
Technically, propositions are the bearers of meaning, but, at least in the years immediately
preceding his writing Principia Ethica, including the years when most of it was worked out as The
Elements of Ethics, he took objects to be propositions as well. When it comes to properties, he
would accept a proposition containing only a property as an element. See “Nature of Judgment,”
Mind (1899).
159
A. Arguments in Support of Premise (2)27
1. The Significant Question Argument
I have identified a number of possible arguments in § 13 for the
conclusion that ‘good’ does not denote a complex property. The first of these, the
significant question argument, claims that for any definition that identifies
goodness with a complex property, since it is always a significant question
whether “the complex so defined” is itself good, goodness is not a complex
property. This is a troubling way to describe the argument, as it is not clear just
what Moore means by “the complex so defined.”28
The usual interpretation is that
Moore means “the complex suggested as a definition.” This is supported in an
example he provides in § 26. In that section, he says of the view that goodness is
a feeling that “It will always remain pertinent to ask, whether the feeling itself is
good; and if so, then good cannot itself be identical with any feeling.”29
I
reconstruct a version of this argument below. Moore aims this argument at the
27
Don Regan claims that modern Mooreans should accept only that the open question argument
shows that goodness is normative, so different in kind than natural properties. He asks, “Why then
does a Moorean believe “good” is not complex? Just because no definition yet suggested in terms
of another normative concept produces as plausible a picture of the moral universe as the picture
which takes “good” as foundational.” Don Regan, “How to be a Moorean,” Ethics 113 (April
2003) pp. 651–677 at 658. This is to explicitly reject the view that the open question argument
proves that goodness is simple, but to tentatively accept the simplicity of goodness, pending
development of a better view. I think Moore’s arguments for the simplicity of goodness (or my
reconstructions, anyway) are worth accepting. 28
At least I am troubled by it. 29
Principia Ethica, p. 93 (§ 26 ¶ 3)
160
theory that goodness is identical with a complex property, so I frame the argument
in terms of the property being something we desire to desire.
• Significant Question Argument
1. If goodness is identical with being something we desire to desire, then the
question whether being something we desire to desire is good is not
significant.30
2. The question whether being something we desire to desire is good is
significant.
3. Therefore, it is not the case that goodness is identical with being
something we desire to desire.
Premise (1) seems a fair embodiment of Moore’s view. As we will see in
a moment, though, there are hints that Moore means something else. As stated, it
is appears to turn on the idea that if the property being something we desire to
desire is identical with goodness, then it would instantiate goodness. In other
words, it would instantiate itself. We must further suppose that when a property
self-instantiates, it is not a significant question whether it does so. But this is a
doubtful proposition. If the question of whether a property self-instantiates is not
significant, and the question of whether some complex property instantiates
30
Alternatively, the premise might read “ . . . is not pertinent” to conform with Moore’s second
example.
161
goodness is a significant question, then goodness is not identical with that
complex property.
Premise (2) asserts that it is a significant question whether the complex
property in question—being something we desire to desire—is good. This seems
right. From this, the significant question argument concludes that goodness is not
identical with being something we desire to desire.
The fault with this argument lies in premise (1). I think that even if
goodness were identical with being something we desire to desire, the question
whether being something we desire to desire is good would be significant. Moore
does not specify what it means to be a significant or a pertinent question, but there
are some hints. Roger Hancock argues that we might draw from the statements I
have suggested as belonging to alternative arguments. In the passage I called the
“argument from doubt,” Moore writes that the fact that we understand what it
means to doubt whether what we desire to desire is good suggests that goodness
and being that which we desire to desire are different properties. According to
Hancock, Moore means by this that knowing what it means to doubt whether
property F is identical with property G implies that the question whether F is G is
significant (and vice versa31
).32
This might be what Moore had in mind, but I
rather doubt it. The question whether F is G is a question of predication, not
31
After all, the argument we are trying to support is that the significance of the predication
question demonstrates that the identity under consideration is false. 32
Hancock, Twentieth Century Ethics (Columbia University Press 1974) p. 30.
162
identity. I do not know what reliable inferences we can make about the
predication of two properties that are not identical. If F were identical to G, it
might not even make sense to ask whether F instantiated G. In fact, without
having done a survey of properties, I expect that our doubts about the identity of
two properties tell us nothing about whether one can instantiate the other. In
other words, whether one property can instantiate another is always a significant
question, or at the very least, turns on something other than our knowing what it
means to doubt their identity. I do not doubt that Hesperus identical with
Phosphorus, and whether being Hesperus instantiates being Phosphorus is not a
significant question. However, my lack of doubt about identity does not ensure
that the instantiation question is not significant. I do not doubt that being a circle
is identical with being a closed plane curve every point of which is equidistant
from a fixed point within the curve, while whether the property being a circle
instantiates being a closed plane curve every point of which is equidistant from a
fixed point within the curve is a significant question.33
Goodness, however, might
be an exception.
To illustrate, let us consider the properties goodness and pleasantness. I
know what it means to doubt whether goodness is identical with pleasantness.
33
I am told that this example was first suggested as an objection to Moore’s requirements for
successful analyses by Ralph Barton Perry in General Theory of Value, (Cambridge,
Mass.: Harvard University Press, 1926), p. 131. Once we recognize that transparency is hard to
come by in a definition, definitions of geometric properties come to mind as paradigm cases, so it
is not surprising that I have stumbled across the same counterexample.
163
The suggestion under consideration would then make the question whether
pleasantness instantiates goodness a significant question. That seems fine. But if
we consider the property goodness, alone, we can see that this suggestion for the
meaning of ‘significance’ fails. I do not know what it means to doubt whether
goodness is identical with goodness.34
We might expect from this that it is not an
open question whether goodness is good. But, in fact, this does seem like a
significant question to me.
Moore does not seem to think it is.35
I contend that it is not clear from
these passages what Moore thinks about this question. While he clearly takes it
not to be significant, the standard interpretation is that he thinks it is insignificant
because it is obviously true, and tautologous.36
This is, I think, a mistake caused
by commentators failing to recognize that he was asking questions of predication,
not identity. Of course goodness is identical with goodness. To deny this would
be a contradiction. But Moore is asking about predication here, not identity. And
he cannot believe that goodness is self-instantiating. Moore takes goodness to be
a property that things have in virtue of their intrinsic properties. But it is not,
itself, an intrinsic property. It is also simple and without parts, so it cannot have
34
Actually, Moore’s language here is unfortunate. I think I know what it means to doubt it, I just
do not actually doubt it. 35
Roger Hancock suggests that Moore thinks sentences of the form “Whatever is P is P” are
always tautologous and cannot be doubted. Hancock, p. 30. Further, Hancock suggests that this is
because the denial of such a sentence is self contradictory. While this is true when the sentence
expresses an identity, it is not necessarily true when it expresses a mere predication. After all,
happiness is not happy. 36
Hancock, p. 30.
164
any intrinsic properties by which to instantiate itself. Therefore, the question “is
goodness good?” is not significant by virtue of having an obvious answer.
Goodness is not good, itself, because goodness cannot instantiate itself. So, on
this theory of what it means to be significant, the significance of a question about
whether a given complex property is good does seem to support the conclusion
that the complex property and goodness are not identical. Unfortunately, it
suffers the defect of being question begging. We ought not rely on a notion of
significance that requires the insignificance of a question about the predication of
goodness to turn on a feature of goodness we are trying to prove, namely, its
simplicity.
Moore needs an account of significance such that whether goodness is
good is not significant and whether any complex property is good is significant.
It is unlikely that Moore can offer such an account because the first condition
cannot be met. I contend that no account of significance will do. I expect that on
any account of significance that makes the first question insignificant, the second
question will be insignificant too. And that on any account that makes the second
question significant, whether goodness is good will be significant as well. Based
on these considerations, the significant question argument fails.
165
2. The More Complicated Question Argument
The preceding argument was based on the very first sentence of Moore’s
argument that goodness is not identical with any complex property. This
interpretation seems reasonable, though it does not comport with the example he
gives in sentences immediately following, in which it is asked, not whether a
particular complex is good, but whether an object or state of affairs, A, that
instantiates that complex is good.37
If those sentences are intended as an example
of the significant question test, then I have the argument wrong, though I would
be in good company. I think, though, that the argument in this passage is
different.
As in the previous argument, the complex property under consideration is
being something we desire to desire. Using ‘A’ as a placeholder for a description
of a state of affairs or an object, Moore asks us to consider a series of questions:
(a) Is A good?
37
Moore seems to apply the argument as being based on the fact that it is an open question
whether the definiens is good and also whether an object or state of affairs that instantiates the
definiens is good. This has been noted by Connie Rosati as well. See, “Agency and the Open
Question Argument,” Ethics vol. 113 (April 2003) p. 497. Moore seems also to ask whether the
state of affairs of the object instantiating the definiens is good. So, is property P good?; Is A
(which instantiates P) good?’ and Is A instantiating P good. This last one is particularly
interesting, because the analogous question about goodness is not tautologous, but arguably
meaningless. It is not clear to me what it means to ask whether it is good that a state of affairs is
good. It is not clear to me that particular moral judgments have any moral value themselves at all.
It strikes me that Moore might be on to something here. While the more complicated question
argument is interesting in and of itself, the fact that the very complicated questions are intelligible
when the same would not be true if ‘good’ were inserted in place of the words denoting the
complex property suggests that Moore can, by asking the right question, get the result that
questions involving only ‘good’ are not significant. (in the sense of ‘making sense to ask).
166
(b) Is it good to desire to desire A?
(c) Is the desire to desire A one of the things which we desire to desire?
(d) Do we desire to desire to desire to desire A?
Moore points out first that (b) is an intelligible question. It asks the same
thing of the desire to desire A as (a) asks about A itself. The theory under
consideration is that ‘good’ means the same thing as ‘something we desire to
desire,’ so we ought to be able to substitute for ‘good’ in (b) to arrive at (d). (c) is
a paraphrase of (d) that is easier for some people to understand. On a theory of
meaning that preserves meaning when synonyms are substituted for each other,38
(b) and (d) will mean the same thing if ‘good’ and ‘that which we desire to desire’
denote the same property. But, says Moore, (b) and (d) do not mean the same
thing. When we assert (b) we do not assert something as complicated as (d).
• More Complicated Question Argument
1. (d) is more complicated than (b).
2. If (d) is more complicated than (b), then (b) does not mean the same as
(d).
3. Therefore, (b) does not mean the same as (d).
4. If (b) does not mean the same as (d), then ‘good’ does not mean
‘something we desire to desire.’
38
A perfectly reasonable view.
167
5. Therefore ‘good’ does not mean ‘something we desire to desire.’
I have phrased the argument in terms of the word ‘good’ rather than in
terms of goodness, itself, since the premise this argument supports is also in terms
of ‘good.’39
In support of (1), we might point to the words that make up the sentence,
but that would not be particularly helpful. Doing so would undermine our faith in
(2). (d) is syntactically more complex than (b), but that proves very little.
Consider the sentence that conveys the definition of a circle, above. ‘This is a
circle’ is far less complicated, syntactically, than ‘This is a closed plane curve
every point of which is equidistant from a fixed point within the curve’ but these
sentences mean the same thing. Moore does not tell us why he believes (b) is
more complicated than (d). When we consider (b), “we have not before our
minds anything so complicated as” (d).40
This seems right. When we consider
the two questions, one just seems simpler. It has fewer relations to keep straight.
Unfortunately, this is true also of the sentences about the circle. The difference is
that for a person well-versed in geometry, the meaning of the syntactically
complex sentence about a circle does not seem any more complicated than the
meaning of the second one. On the other hand, a committed naturalist who
39
This argument, like the overall argument that this one supports, can be rephrased in terms of
properties rather than words, but it is far easier to treat it, as Moore does, as an argument about
what the denotations. Since Moore is a realist about meaning, we can draw the appropriate
conclusions about the properties themselves. 40
It is not clear here whether he means for us to consider the sentence that expresses the question
or the question itself.
168
offered the definition we are considering might stand in the same position as our
geometer with respect to (b) and (d)—although (d) looks more complicated than
(b) to Moore and to me, to an expert naturalist, they might appear equally simple.
I reserve judgment on this premise for the time being. Ultimately, even if this
argument fails, I think we can construct one from its ashes that succeeds.
Premise (2) relies on the implicit claim that no two questions that differ in
complexity mean the same thing. They are different questions. I do not know
whether this is true or not, but I suspect it is. Premise (4) relies on the view that
synonyms can be substituted for each other without a resulting change in the
meaning of sentences in which they appear.41
This seems correct.
On the whole, this argument looks pretty good. Note an important feature
of this argument: it does not work for simple properties offered as definitions of
goodness. For example, if the definition under consideration was that goodness
was pleasantness we would get:
(b') Is it good that A is pleasant?
(d') Is it pleasant that A is pleasant?
Here, the argument would fail at (1) because (d') is no more complicated
than (b'). Without the leverage from their differing levels of complexity, Moore
could not show that (b') and (d') did not have the same meaning. But this is no
41
In Fred Feldman’s version of this argument, he sensibly attributes a compositional theory of
meaning to Moore that accounts for the substitutivity of synonyms. See, Feldman (2005) pp 31-
33.
169
drawback of the argument. Remember that this argument was only in support of
the premise that ‘good’ does not denote a complex property. As I have outlined
above, Moore does not present an argument in § 13 that goodness is not identical
with any simple property.
3. The Argument From Inspection
The entire argument from inspection is contained in one sentence in
Principia Ethica. Again discussing the suggestion that goodness means that
which we desire to desire, Moore writes:
Moreover any one can easily convince himself by inspection that
the predicate of this proposition—‘good’—is positively different
from the notion of ‘desiring to desire’ which enters into its subject:
‘That we should desire to desire A is good’ is not merely
equivalent to ‘That A should be good is good.’
The proposition Moore is writing about is not specified. But given the
examples with which he follows, it appears that he means the proposition
expressed by an affirmative answer to (b). On the other hand, we do not think of
words as being the subjects or predicates of propositions, but of sentences.42
We
42
This is a picky point. Moore is pretty clearly not talking about the words ‘good’ and ‘desiring to
desire,’ but the properties to which they refer, which can make up the subject and predicate of the
resulting proposition.
170
can cure this defect in one of two ways. We can take Moore to mean either that
we can tell by inspection of the properties that enter into the appropriate places in
the structured proposition that the denotations of ‘good’ and ‘being something we
desire to desire’ are different, or that we can tell by inspection of the meanings of
propositions that result from affirming (b) and substituting the definiens and the
definiendum that ‘good’ and ‘being something we desire to desire’ are different.
The first possibility ought to be rejected. If we had that kind of epistemic access
to the properties themselves, there would be no need for argumentation. Whether
goodness was identical with any complex property would be immediately
ascertainable. It seems to me that the very point of Moore’s dialectical device—
to substitute the properties in propositions and recognize that the resultant
propositions have different meanings—is to provide a way to see that the
properties are different when it is not apparent by direct inspection of the
properties themselves.43
We have, instead, to make use of the second route.
Recall (b):
(b) Is it good to desire to desire A?
The affirmative answer to the question posed might be:
43
In fact, there may be little reason to think that it is any easier to see that the meanings of whole
sentences are different than it is to see that the meaning of constituent parts of those sentences are
different. Given the compositionality of meaning that seems to underlie some of these arguments,
direct inspection of the individual properties is likely what allows us to discern a difference in the
meanings of the sentences. If so, then what I have called the argument from inspection may be
two distinct arguments. One relies on direct inspection of the properties, and the other on
inspection of the properties in the context of a whole sentence.
171
(e) It is good to desire to desire A.
If we substitute ‘good’ for ‘to desire to desire’ we get:
(f) It is good that A is good.44
If the definition under consideration is true, we would expect that (e) and
(f) have the same meaning. But we can tell “by inspection” that they do not.
From this we proceed exactly as in the More Complicated Question argument.
• Argument From Inspection
1. (e) does not mean the same as (f).
2. If (e) does not mean the same as (f), then ‘good’ does not mean
‘something we desire to desire.’
3. Therefore ‘good’ does not mean ‘something we desire to desire.’
This argument is slightly less satisfying than the More Complicated
Question argument. This is because we have so little support for (1). In the
previous argument, we had the fact that one question was more complicated than
the other to bolster our intuitions that (b) and (d) were distinct in meaning. Here,
we have inspection of (e) and (f). I am not convinced that we can discern a
difference. Moore claims that we can see that one is not merely equivalent to the
other, but I am not convinced. I expect that someone who really believed that
goodness and being that which we desire to desire were the same property would
44
Again, I think the real puzzle here is how (f) is meaningful at all.
172
not see the difference even after close inspection. To tell the difference between
the propositions expressed by (e) and (f) we would need the same privileged
epistemic access that I doubted above.
4. The Argument from Doubt
The last argument I find in § 13 in support of the claim that ‘good’ does
not denote a complex property is the “argument from doubt.” Moore claims that
the fact that we understand what is meant by doubting that goodness and that
which we desire to desire are coextensive shows that we are thinking about
distinct properties. He writes:
It may indeed be true that what we desire to desire is always good;
perhaps, even the converse may be true: but it is very doubtful
whether this is the case, and the mere fact that we understand very
well what is meant by doubting it, shews clearly that we have two
different notions before our mind.45
I have framed the argument in terms of coextension because Moore seems
to grant that it might be true that what has the complex property is always good,
45
Principia Ethica, p. 68, § 13, ¶ 2.
173
but he doubts whether the relation goes the other way. He doubts whether every
instance of goodness is an instance of something we desire to desire.46
We can construct a fairly simple argument from this.
• Argument From Doubt
1. We doubt whether goodness is coextensive with being that which we
desire to desire.
2. If (1) then goodness is not identical with being that which we desire to
desire.
3. Therefore, goodness is not identical with being that which we desire to
desire.
Premise (1) is uncontroversial. Many people doubt that the properties are
coextensive. That is, they either doubt that all instances of goodness are instances
of desiring to desire, or they doubt that all instances of desiring to desire are
instances of goodness. Premise (2) is problematic. The fact that we doubt that
two properties are coextensive does not prove that they are not identical. It does
show that we doubt that they are identical because you cannot have identity
without coextension. It might be objected that I have mischaracterized the
argument. Instead of arguing from the presence of doubt to the non-identity of
46
It is not necessary to frame the argument this way. Moore certainly seems to doubt both
propositions. But he doubts one more than the other. Framing the question as I have takes care of
both.
174
goodness and the given complex property, Moore argues from the fact that we
understand what it means to doubt the coextension that the properties are not
identical. The argument would go as follows:
1'. We understand what it means to doubt whether goodness is coextensive
with being that which we desire to desire.
2'. If (1') then goodness is not identical with being that which we desire to
desire.
3'. Therefore, goodness is not identical with being that which we desire to
desire.
Unfortunately, this argument is no better than the first, and in fact, I think
it is somewhat worse, in that it really amounts to the traditional interpretation of
the open-question argument, and fails for the same reasons. That is, given the
most likely explanation for (2'), the argument will be plagued by Frege problems.
Presumably, we cannot understand what it means to doubt that two identical
properties are coextensive. An understanding of the relations identity and
coextension seem to guarantee this. If so, to understand what it means to doubt
whether two properties are coextensive is, presumably, to be open to the
possibility that one property can be instantiated without the other, and hence, the
possibility that they are not identical. But, as we know, this willingness to
entertain the non-identity of properties does not demonstrate that they are not
identical. Many of us understand what it means to doubt that Hesperus is
175
Phosphorus. But this does not show that Hesperus and Phosphorus are non-
identical. Being Hesperus is identical with being Phosphorus. So premise (2') is
false.
Perhaps it may be objected that competent reasoners cannot know what it
means to doubt that Hesperus is Phosphorus, and that, generally, when we
consider properties that are identical, we cannot know what it means to doubt
their coextension. Such a proposal amounts to amending premise (1'), removing
‘We’ and substituting ‘Competent moral reasoners’ in its place. I think this
modification fails as well. Even a competent moral reasoner may understand
what it means to doubt whether goodness is coextensive with, say, intrinsic value.
Then, the argument would conclude that goodness is not identical with intrinsic
value. But Moore takes these to be the same property.47
Finally, a Moorean
might object to the argument on grounds that the explanation offered for premise
(2') fails. She might claim that there is no general principle about doubting that
justifies the result here, but rather, the argument only holds for goodness and
proposed complex properties offered as definitions. I see no reason to limit the
reasoning for premise (2'). Without an account of goodness that explains why our
understanding what it means to doubt its coextension with a given complex
property has any bearing on whether the properties are identical, the argument
will fail. I do not see a plausible explanation forthcoming.
47
He takes ‘good’ and ‘intrinsic value’ to be names for the same property.
176
5. Conclusions regarding the arguments above
The upshot of this discussion is that in support of the claim that ‘good’
does not denote a complex property, we can construct four arguments using
Moore’s text. Of these, only one, the “significant question” argument, seems to
turn on anything that looks like an open question. Three of the arguments—the
“significant question argument,” the “argument from inspection,” and the
“argument from doubt”—suffered from defects that seem insurmountable. The
If the defective arguments could be supported, there is a minor difficulty
that I mention only as an aside. Two of the arguments, the “significant question”
argument and the “doubt” argument, conclude that goodness is not identical with
being that which we desire to desire. In order to be support for the overall
argument that ‘good’ neither denotes a complex property nor is meaningless, the
arguments would need additional premises to bring us from the claim about the
distinctness of the properties to the claim that the word ‘good’ does not denote
some complex property. But the required premise is both easy to supply and
plausible. For example, we might try “If goodness is not identical with being that
which we desire to desire, then ‘good’ does not denote the property being that
which we desire to desire.”
177
A more significant problem for all of these arguments is that none of the
arguments conclude that no complex property is identical with goodness.48
Each
one argued for the conclusion that a specific property was not identical. So, for
instance, while the “more complicated question” argument might be sound, it
does not, in its current form, support the premise that ‘good’ does not denote
anything complex. There appear to be several ways to resolve this problem.
Moore would need to represent each of these as general arguments, or as
inductive arguments, or perhaps, he could claim that after we perform the
argument on a number of different complex properties, we begin to see that no
complex property can be a satisfactory definition. I expect that the first option
holds out the most hope. Unfortunately this hope is misplaced. For example, to
generalize the “more complicated question” argument, we might start with
generalized versions of (b) and (d). Where X is the definiens of any definition
wherein ‘good’ is made to denote a complex property, we would get:
(b'') Is it good that A instantiates X?
(d'') Is it X that A instantiates X?49
My intuitions about whether (d'') is more complicated a question than (b'')
do not help in this situation. I cannot motivate the first premise of the argument
because I simply do not find (d'') to be more complicated. Of course this is
48
Or alternatively, that ‘good’ does not denote any complex property. 49
This is equivalent to, “(d''') Is A’s instantiating X a thing that instantiates X?” but this, though it
looks more complicated that (b'') does not fare any better than (d'') when (b'') is reformulated to
mirror (d''')’s structure. Consider, (b''') Is A’s instantiating X a thing that instantiates goodness?
178
because by substituting ‘X’ for the name of a complex property, I have masked
the syntactic complexity of (d''). And it turns out that my intuitions are tied to
how complicated the question looks syntactically. But this actually highlights a
real problem that afflicts this argument. The complexity of the expression used to
refer to the complex property seems to make a difference to this argument.50,51
So, while this argument is convincing to me for certain properties, if generalized,
it is less convincing.
I have little faith in my ability to make judgments about the complexity of
questions that are perfectly general. This might suggest favoring the Argument
From Inspection but I am afflicted with a similar problem there. I am not
confident in my assertions that statements of the appropriate generalized forms
have different meanings.
Moore, himself, seems to favor the third strategy I suggested for solving
the problem. We should apply the argument to a number of different complex
properties. After multiple applications, we will see that no definition wherein
50
In retrospect, I can think of at least one complex property with a simple-sounding name, and it is
such that competent geometers might find (d'') more complicated. It is the property Being
prismatic. I suggest the following analysis of this property: x is prismatic iff x is a triangular
prism, or x is a rectangular prism, or x is a pentagonal prism. Even a non-disjunctive analysis
would be quite complicated—being a three dimensional object having a polygonal cross section,
etc. The point is that properties with simple names might be extremely complex, and such
properties are the ones to look at to judge whether the More Complicated Question argument and
the Argument from Inspection fail. 51
The property named ‘plaid’ is a complex property consisting of having several colors and some
sort of recurring geometric pattern, etc. If being plaid were offered as a definition of goodness, we
might find the question “is x plaid?” no more complicated than “is x good?”
179
‘good’ denotes a complex property will be true.52
I find this suggestion
unsettling. It amounts to yet another argument with a premise I find dubious. We
might construct the following argument on Moore’s behalf.
• Moore’s Generalized Conclusion argument
1. We can apply the More Complicated Question argument against a number
of proposed complex definitions. (Go ahead. Try it. We have seen it
work with being that which we desire to desire. We can apply it also, with
good results, to the properties being in conformance with the collective
will, being such that at least 4 out of 5 people surveyed would approve,
being what Jesus would do, etc.)
2. If (1), then we can see that no complex definition will survive the More
Complicated Question argument.
3. If we can see that no complex definition will survive the More
Complicated Question argument, then ‘good’ does not denote any
complex property.
4. Therefore ‘good’ does not denote any complex property.
I see no reason to accept (2) or (3). And this concern is not limited to the
More Complicated Question argument. For any argument we have considered so
far, the Generalized Conclusion argument seems to fail. Of course, this is rather
unfair of me. Inductive arguments of many sorts, even ones we have no real
52
See Principia Ethica, the first sentence of § 13, ¶ 2.
180
doubts about, when reframed as deductive arguments suffer from the defect we
see in (2). Consider, for example, the argument that the sun will rise tomorrow
morning:
S1. The sun has risen every morning in historical memory.
S2. If (1), then the sun will rise tomorrow.
S3. Therefore, the sun will rise tomorrow.
I am quite confident of the conclusion, but not because I find the argument
convincing. In fact, I think (S2) is false. Past performance is no guarantee of
future results. However, since the conditions underlying the former dawns still
obtain, and will obtain (in all likelihood) tomorrow, the conclusion is justified.
And even without knowing about those conditions, the inductive argument would
be pretty good.
Moore likely intends some sort of induction here. We are supposed to see
that the consequent of (2) is true, but not for any deductively valid reason. The
expectation is that we will simply become aware of the fact that there is
something about complex definitions that make them fail the More Complicated
Question argument. I have suggested that the most likely reason, that our
intuitions about the complexity of the question are based on the syntax of the
question, is unsatisfactory. But perhaps there are other reasons.
181
6. Final thoughts and the Revised General Significant Question Argument
Before leaving arguments in support of the claim that ‘good’ does not
denote any complex property, I consider the comments of one of Moore’s critics,
and I offer a version of the significant question argument that has a better chance
of succeeding than the one I presented above. Roger Hancock considers what I
have called the Argument From Doubt in his book Twentieth Century Ethics.53
His explanation of the argument fails in two instructive ways.
First, he fails to distinguish this argument from the significant question
argument. Instead, he treats it as an explanation of what Moore means by ‘a
significant question.’ Hancock thinks the question “Is F good?” is significant if
the claim “F is not good” is never self-contradictory. The idea here is that for any
complex property F, it can always be asked with significance, whether F is good
because “F is not good” is never self-contradictory.54
Unfortunately, as I
mentioned in footnote 35, this suggestion for what it means to be a significant
question fails. Moore would not accept it because he thinks the question whether
goodness is good is not a significant question, but I contend, he would reject the
claim that “Goodness is not good” is self-contradictory.
53
Cited above. 54
This, in turn, is based on the claim that “Whatever is F is F” is a tautology. And this claim is
false, at least with regard to predication.
182
Hancock’s second error is that he fails to distinguish between the
following statements:
‘That we should desire to desire A is good’
and
‘Whatever is F is good,’ where F is substituted for ‘that which we should
desire to desire.’
Hancock makes this error on his way to his conclusion about what it
means to be a significant question. Moore asserts: “‘That we should desire to
desire A is good’ is not merely equivalent to ‘That A should be good is good.’”55
Hancock generalizes these as
(g) Whatever is F is good
and
(h) Whatever is F is F.56
His argument is that Moore claims that if ‘good’ were synonymous with
some non-ethical expression, F, then (g) and (h) would mean the same thing. (g)
and (h) do not mean the same thing because (h) cannot be doubted and (g) is
doubtable. (h) is not doubtable because the denial of (h) is self-contradictory.57
Why Hancock does not treat this as a distinct argument is unclear to me. He
55
Principia Ethica, pp 67-68. 56
Hancock, p. 30. 57
Alternatively, because (h) expresses a tautology. One reason to think Moore did not intend this
particular argument is that nowhere in his discussion of the arguments in § 13 does he use the
word ‘tautology’ or ‘tautologous.’ In fact, the word does not appear in the entirety of Chapters I
and II. I find it surprising that his commentators attribute to him this claim or ones like it.
183
seems to have all he needs to motivate an argument against the definition of
‘good’ as F. Notice, though, that his argument does not faithfully recreate
Moore’s.
The mistake is in not recognizing that Moore’s statement asserts that
something other than what (g) asserts. (g) asserts that something that instantiates
F is good. Moore’s statement asserts that it is good that something (A)
instantiates F. It ought to be easy to see that one asks about the goodness of A
and one asks about the goodness of the state of affairs of A having property F. I
highlight this mistake because it demonstrates a possibility for a Significant
Question-style argument that does not suffer from the defects we identified in the
original.
The question whether it is good that a state of affairs is good is not
equivalent to the question whether it is good that a state of affairs is F, where F is
a complex property. This is because it is a category mistake to ask whether it is
good that a state of affairs is good. But it is not such a mistake to ask whether it is
good that a state of affairs is F, where F is a complex property. Why do I say that
it is a category mistake? This is not the same claim as my contention that
goodness is not good, itself. Instead, this is the claim that moral value simply
does not attach to moral facts. Moral value attaches to acts and states of affairs,
but not to facts about what acts or states of affairs are or are not good. Moral
value might attach to my believing a certain moral proposition, but not to the
184
proposition (or the underlying fact) itself—it might be morally obligatory to
believe that torture is wrong. Other normative properties might attach to moral
propositions. For example, the proposition that charitable giving is good might
instantiate the property being a justified belief in Bill. Or a moral proposition
might be self-evident. These would be two different normative properties—
epistemic properties—that can rightly be predicated of moral propositions.
Consider the proposition that it is morally wrong to pay taxes in support of
war-mongering governments. This proposition may be true or false, but the
proposition itself is not morally wrong. Neither is it morally right, permissible,
nor any other moral property.58
In short, the proposition that it is good that it is
good that A is neither true nor false. As a result, the question whether it is good
that it is good that A (alternatively, whether it is good that A is good) is not a
significant question. I have, therefore, been arguing that what it means to be a
significant question is that the question can be answered by a statement that has a
truth value.59
I cannot say with any degree of certainty that this is what Moore
had in mind. But it is an explanation that utilizes the example sentences he gave
without the modifications that others, like Hancock, have introduced. It also has
the benefit of being useful for creating a version of the Significant Question
argument that succeeds without being question begging.
58
What I am objecting to is second-order moral properties. 59
Either ‘true’ or ‘false.’ I do not admit, for purposes of this discussion, the possibility of multi-
valued logics.
185
Consider the following questions:
(i) Is it good that A is a thing we desire to desire?
(j) Is it good that A is a thing that is good?
I contend that (i) is a significant question. An answer, yes or no, to that
question will be either true or false. This is because moral values can and do
attach to propositions about complex properties, including complex psychological
properties. If I assert that it is good that we desire to desire A, I am asserting
something true or false. Therefore, (i) is significant. (j) is not significant. This is
because, as I have argued, moral value does not attach to moral propositions.
Note that since the feature of goodness that I am exploiting in this argument is not
one we are trying to prove, the resulting argument will not be subject to claims of
question begging. Note also, that goodness is unique in not being able to be
truthfully predicated of a state of affairs being good. I know of no complex
property that shares this feature. Consequently, for any complex property, the
question whether it is good that A instantiates that complex property is always
significant. Now we are in a position to construct a version of the Significant
Question argument that is general and not obviously unsound.
186
• Revised General Significant Question Argument
1. For any complex property F, we may ask with significance
‘Is it good that A is F?’
2. We may not ask with significance
‘Is it good that A is good?’
3. If (1) and (2), then ‘good’ does not denote any complex property.
4. Therefore, ‘good’ does not denote any complex property.
This argument has a conclusion written in terms of the word ‘good,’ as
required by the premise of the overarching argument that this argument is
intended to support. It is also a general conclusion, applicable to any definition
that attempts to define goodness in terms of a complex (non-simple) property.
And finally, it does not turn on an account of significance that requires us to
accept that goodness is simple. The unique feature of goodness that I have
pointed out does not require any particular metaphysical commitments, at least
about the sort of definitions that are acceptable. Finally, premise (3) is supported
by considerations of meaning discussed above.60
This argument is also sufficient to show that many simple properties are
not identical with goodness as well simply by removing the word ‘complex’ from
(1), (3) and (4). There are instances, of course, where such a modification would
60
Preservation of meaning through substitution of synonymous terms.
187
prove problematic. If ‘good’ were substituted for ‘F,’ the argument would fail,
because (1) and (2) would be contradictory.
Even without that modification, we might doubt (1) on grounds that there
might be a complex property about which the question whether it is good that A
instantiates the property is not significant. While I suppose this may be true, I am
completely at a loss to think of one, and until this situation changes, I am inclined
to accept (1).
Based on this argument, primarily, and perhaps on particular instances of
the More Complicated Question argument, I think the claim that ‘good’ does not
denote any complex property is well supported. This leaves the possibility that
‘good’ is meaningless, or that it is simple and indefinable. Premise (3) of
Moore’s § 13 argument is that ‘good’ is not meaningless.
A revised version of my Revised General Significant Question Argument
might suffice to show that most simple properties are not identical with goodness,
but I do not take this to be Moore’s purpose in § 13. That is, if successful, the
argument I offered on Moore’s behalf shows a good bit more than he intended,
and maybe as much as most commentators think he intended. I do not pursue this
argument further at this time, as it generally undermines any motivation for
construing the naturalistic fallacy in the way I find most consistent with the text.
188
B. Argument in Support of Premise (3)
Here, Moore is more economical and offers only one argument. This
argument, like the ones already discussed, has been fairly well misconstrued. I
very briefly describe three bad approaches to the argument before presenting the
one that best comports with the text. All three approaches share a common
error—that of not treating the argument as what it claims to be, that is, not
treating the argument as supporting the view that ‘good’ is not meaningless.
Hancock uses the first two sentences of the passage in which Moore
makes his argument61
to argue for a view of significance, in support of his version
of the open question argument. He writes
“. . . Moore is attacking the view that ethical words are
synonymous with non-ethical expressions. And his argument is
that when we ask “Are F’s good?” we can recognize on reflection
that there are two distinct things before our mind . . . . The
question “Is S P?” is significant, then, if S and P designate two
distinct things . . . .”62
61
Reproduced here: “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
recognise 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.” Principia Ethica, p. 68. 62
Hancock, p. 31.
189
It is understandable how Hancock might take this passage to make this
argument. But doing so requires ignoring the second sentence which makes it
fairly explicit that Moore is focusing on recognizing that the thing we are
comparing against many different properties is the same property in each case.
Instead of emphasizing the distinction between goodness and any other property
under consideration, Moore is emphasizing the sameness of the property we
referred to by ‘good’ in each case.
Thomas Baldwin claims that Moore’s argument that ‘good’ is not
meaningless really shows something else altogether. He writes that Moore’s
argument is really “that because we can distinguish in thought between the
meaning of ‘good’ and that of any putative non-ethical analysis, all such analyses
must be incorrect.”63
He claims that this argument, like the other Moorean
argument—by which he means the traditional interpretation of the open question
argument64
—hinges on the failure of “reflective substitutivity” of proposed
definitions, which is evidenced by our ability to doubt their identity. In other
words, since we can doubt that ‘good’ and ‘P’ where P refers to any (natural)
property, goodness and P are not identical.
This is certainly not what Moore has done here. He never ever asserted
that the proof shows that we can distinguish in thought between the meaning of
63
Baldwin, T., G.E. Moore, Routledge, 1990 (London), p. 89 64
Baldwin ignores the division of arguments I have advocated, arguing instead that there are only
two arguments, both having equal claim to the title “open question argument” and both to the
conclusion that no definition of ‘good’ is correct.
190
‘good’ and that of any putative non-ethical or natural analysis. What he did assert
is that if we keep at it, repeatedly considering the question ‘Is P after all good?’,
after a while, we may see that the word on the right side of the sentence—
‘good’—always refers to the same property, in case after case. And this, supposes
Moore, proves that ‘good’ is not meaningless. We will consider whether he is
right in a moment.
Finally, Fred Feldman, who does a very good job of distinguishing the
various arguments in § 13, fails to see the point of the argument. Rather than
taking it to show that ‘good’ is not meaningless, Feldman takes the argument to
be an attempt to refute the view that goodness is identical with some simple,
natural property.65
Perhaps this is because Moore offers only “Is pleasure . . .
after all good?” and alludes to other questions of the same form about
pleasantness, being desired, and being approved as an examples of the sort of
question one should consider. But he does two things that clarify his true
purpose. First, he writes that the question we consider is “Is pleasure (or
whatever it may be) after all good?” The phrase “or whatever it may be” does not
imply that the question can only be asked of simple, natural properties. It can be
asked of any property, simple or complex, natural or non-natural. I take it that the
example questions were all about simple definitions of goodness because they are
easier to ponder, and we are less likely to get bogged down in considering the
65
Feldman, p. 35.
191
question so that we can focus on the appropriate thing—that ‘good’ means the
same thing in each question. And this is because the point of asking the question
is not answering it. It is recognizing that we are asking, every time, whether the
property under consideration instantiates the same property, goodness. The
second clue Moore provides is in the first sentence of the paragraph from which
the example comes. Moore writes “And the same consideration (our ability to
doubt whether those things that instantiate the property offered as a definition
instantiate goodness—WP) is sufficient to dismiss the hypothesis that ‘good’ has
no meaning whatsoever.”66
The very point of the argument that follows is to
show that ‘good’ is not meaningless. It is not to show that proposed simple
definitions fail.67
I have already described the idea behind the argument in my comments
about Hancock, Baldwin, and Feldman. All that remains is to present the
argument. Consider the following questions:
(k) Is pleasure after all good?
(l) Is desire after all good?
(m) Is being approved after all good?
66
Principia Ethica, p. 68. 67
Prof. Feldman does not describe the material where Moore tells us the point of this passage.
Instead he writes “In any case, after a bit of introductory commentary . . . .” and then launches into
the argument. In this case, that introductory commentary demonstrates the real point of the
passage. I suspect Feldman had a different axe to grind. He presents an argument that is not
Moore’s but is an interesting argument nonetheless. It goes toward his overall thesis that
commentators have done a miserable job of presenting the open question argument.
192
(n) Is being that which we desire to desire after all good?
Uniqueness Argument
1. (k) – (n) ask the same unique question of each property.
2. If (1) then ‘good’ denotes the same property in every context.
3. Therefore ‘good’ denotes the same property in every context.
4. If (3) then ‘good’ is not meaningless.
5. Therefore, ‘good’ is not meaningless.
Premise (1) is supported by our basic intuitions about what we are asking
when we entertain questions like (k) through (n). I phrased each of these
questions as questions about properties, but we could instead ask whether
particular states of affairs were good and still achieve Moore’s desired conclusion.
That is, we do not have to consider questions based on attempted definitions of
goodness.
Premise (2) is a stretch. Moore writes that when we ask enough of these
questions, we “may become expert enough” to recognize that ‘good’ denotes the
same property in each question. The idea is simply that we always have the same
idea in mind when we ask such questions. And the only way this would be the
case is if ‘good’ meant a single thing. So, in fact, (2) is somewhat stronger than
Moore’s claim.
193
This is a modest argument.68
Note what this argument does not do. It
does not prove that goodness is not identical with any simple natural property. In
fact, there is no explicit argument for that proposition in this section. This
argument rules out nothing as a definition of ‘good.’ It merely shows that there is
some one thing to which ‘good’ refers in each of these moral uses.
IV. Conclusion
I have argued that rather than being one unified argument that goodness is not
identical with any non-natural property, Moore’s open question argument really
consists of two arguments. The first is that goodness is not a complex property.
The second is that goodness is not “meaningless.” The first argument is
supported by a number of distinct sub-arguments. These are not all sound,
especially when we consider versions of the argument with general conclusions
(as opposed to arguments that particular complex definitions fail). However,
problems with Hancock’s notion of significance suggest an account of that results
in a very robust argument, the Revised General Significant Question Argument. I
believe this argument poses a significant challenge to definitional naturalists.
68
There might be a sub-argument in here. The primary argument shows that you always have a
unique property in mind when you wonder whether pleasure is good. The sub-argument notes that
everyone understands the question “is X good?” And their state of mind is different when they ask
that question than it would be when they asked “is X pleasant?” or “is X desired?” But this seems
to invite us to do something like the Argument From Inspection but to the conclusion that ‘Good’
is not meaningless. I resist this impulse because it is not clear to me that distinctness of state of
mind yields distinctness of meaning, and further whether this implies that ‘good’ has a meaning at
all.
194
Chapter 6
Conclusions
I. Concluding remarks about the relation between the open question argument
and the naturalistic fallacy
In the previous chapter, I argued that rather than being one unified
argument that goodness is not identical with any non-natural property, Moore’s
open question argument actually consists of two arguments. The first is that
goodness is not a complex property. This claim is supported by at least five sub-
arguments, most of which turn on our intuitions about language and meaning.
The second is that goodness is not “meaningless.” This claim amounts to the
claim that there is a unique property which goes by the name “goodness.”
It appears that at least one of the sub-arguments that can be reconstructed
from Moore’s writing is sound and may prove significantly more than Moore
intended.1 Rather than merely proving that goodness is not identical with any
complex property, it may be sufficient to show that goodness is not identical with
any simple property that can be referred to by any natural property referring
name. If this is the case, then Frankena’s question begging claim can be
1 Though it is not clear that it is exactly what he had in mind. My notion of what it means to be
significant is not what most commentators attribute to Moore.
195
dispensed with because Moore will have shown that goodness is indefinable.
Recall that Frankena claimed the naturalistic fallacy was question begging
because when conceived as a definist fallacy, it seems to presuppose that
goodness is indefinable. While I strenuously object to this conception of the
naturalistic fallacy, if Moore can show that goodness is indeed indefinable by a
method that does not invoke the naturalistic fallacy, he will have avoided begging
the question. The version of the open question argument I called the “Revised
General Significant Question Argument,” while turning on a conception of
significance that I am not sure Moore shared, seems to show that goodness is
indefinable without invoking the naturalistic fallacy. Hence, I believe that even if
we construe the naturalistic fallacy as Frankena urges, Moore can avoid the
criticism that he has begged the question.
My interpretation of Moore’s open question argument has consequences
for Prior’s argument as well. Prior correctly believes that Moore takes the open
question argument2 to show that goodness is both simple and incapable of
definition. This is a reasonable claim because, for Moore, being incapable of
(analytic) definition is a consequence of being simple. As I pointed out in
Chapter 4, Prior invokes something akin to the traditional interpretation of the
open question argument. His complaint with the argument he reconstructs on
2 When I refer to the open question argument in this section, I am referring to the entire argument
of Section 13 in Principia Ethica, despite the fact that the arguments in that section do not turn on
the “openness” of any particular question or questions, but rather on some other features of
language, such as the presumptive complexity of questions containing property names.
196
Moore’s behalf is well founded, although his point can be made clearer. The
central problem for the view is that it turns on a feature of language that is simply
untrue. In order for the standard version of the open question argument, and
Prior’s version as well, to succeed, language would have to be transparent. That
is, the argument requires denying referential opacity—a feature of language that
contemporary philosophers by and large accept. When we consider analyses such
as water is H2O, or a brother is a male sibling, we find that it is possible to create
questions that at least some people would take to be open ones.
We might argue, though, that the fact that the open question argument
seems to fail in these contexts is irrelevant because moral language is unique—
that we have privileged epistemic access to moral properties, and as a result, the
sort of transparency that does not exist with non-moral properties and objects does
not plague our moral language. Whereas ordinary language is referentially
opaque, moral language is transparent. In other words, if two expressions of
moral language, ML1 and ML2, mean the same thing, and an individual
understands the expressions,3 then she will know that any two sentences (or
questions) which differ only in the substitution for ML2 for ML1 (or vice versa)
have the same meaning.
3 As I have phrased this transparency thesis, the privileged epistemic access implies that
understanding an expression means knowing its referent and recognizing that it is the same as that
of any of that expression’s synonyms. For example, if I had such privileged epistemic access to
astronomical objects, I would not make the ‘Hesperus/Phosphorus’ error because I would know,
by virtue understanding the names ‘Hesperus’ and ‘Phosphorus’ that each referred to the planet
Venus.
197
This response does not succeed. Two problems suggest themselves. First,
and less decisively, Moore might be loathe to accept the principle underlying this
response. Moore would not want to claim that moral language is unique in this
way. He was, after all, attempting to put Ethics on similar footing to other
sciences, in part by employing rigorous thought and process to moral reasoning.
Carving out a special epistemic niche for moral properties and special status to
moral language would run counter to this project. Second and more significantly,
the transparency problem plagues even moral language. Consider the words
‘goodness’ and ‘intrinsic value.’ Moore takes these words to be synonyms that
each refer to the property moral goodness. But if we perform the open question
argument on these words, which do refer to the same property, they would ‘fail’
the test. That is, most people would find the question, “x is intrinsically valuable,
but is it good?” an open question. If the openness of this question is an indication
that ‘goodness’ and ‘intrinsic value’ are not names for the same property, then
presumably, they do not name the same property. Moore cannot accept this. This
suggests that we do not have special epistemic access to moral properties that
would defeat the lack of transparency concerns. This is a point recognized by
Prior.4 So, if the open question argument is given its standard interpretation,
Prior’s criticisms are well-founded.
4 Prior, p. 3.
198
Once we place the open question in its proper context, and properly
disambiguate the many arguments contained within it, we can see that Prior’s
objection fails. The open question argument is Moore’s proof that goodness is
simple and indefinable. And although one of the arguments I reconstructed on
Moore’s behalf seems to have as a consequence that goodness is not identical
with any natural property, Moore did not take this to be the point of the argument.
The point of the open question argument really seems to be to show that goodness
is simple. But a simple property, while incapable of analytic definition, is
nonetheless still susceptible to successful property identifications. For example,
the property yellowness is simple, and it is identical to the property being the
color that when mixed with blue yields green.
Moore is concerned to limit this kind of definition (definition as
identification with a simple property) of goodness as well. After all, his thesis
that goodness is a non-natural property would be false if it turned out that a
successful property identification between goodness and some simple natural
property could be made. This is where Moore’s claims about the naturalistic
fallacy become important. And it is the reason I devote two chapters to justifying
an alternative to the traditional definist interpretation. Moore’s claims about the
naturalistic fallacy do not show that any such property identification is incorrect,
only that a particular form of argument, one which might be the only one
available, does not provide adequate support for any such identity claim. The
199
picture that emerges on my view of the open question argument and the
naturalistic fallacy is very unlike that conceived by Prior.
That the naturalistic fallacy is commonly confused with the open question
argument is proved by Baldwin, who writes, in his discussion of the open
question argument, that “It is notable that the section in which he advances it is
the only part of the opening discussion in Principia Ethica of the naturalistic
fallacy which does not come straight from The Elements of Ethics.”5 I have
shown that the open question argument is distinct from the naturalistic fallacy.
But Baldwin here asserts that Moore’s presentation of the open question argument
(by which he means his proof that goodness is simple and indefinable) is part of
his discussion of the naturalistic fallacy. This might explain why some take it to
be a simple definist fallacy.
Rather than taking the open question argument to be the proof that it is a
fallacy to try to define goodness, I reverse the concepts. The open question
argument is Moore’s proof that goodness is simple, and hence indefinable, and
Moore’s claims about the naturalistic fallacy show that identifying goodness with
any simple natural property is epistemically problematic.
Recall that for Moore, the way that philosophers who commit the
naturalistic fallacy proceed to define goodness is to recognize that goodness and a
given natural property are instantiated in the same objects or states of affairs, and
5 Baldwin, p. 87.
200
to infer from this that they are identical. Moore’s complaint with this approach is
that it gives you no reason to accept the identity of goodness and the natural
property in question. So analytical definitions are inapplicable, by virtue of the
success of the open question argument (in one or more of its variations), and
simple definitions are of dubious value by virtue of the commission of the
naturalistic fallacy by philosophers who offer such definitions. Recall also that
Moore recognizes that we might engage in reasoning about the good without
defining goodness at all. This was the tactic employed by Sidgwick, for instance.
This suggests a question: May we arrive at a property identification, that is, a
definition in terms of a simple property, that does not commit the naturalistic
fallacy?
I suggested in Chapter 3 that we might avoid the naturalistic fallacy by
relying on intuition to arrive at our definition. This might not be particularly
satisfying either, given that we may not share much commonality in our
intuitions. Moore has no objection to identifying the good with natural properties,
and doing so by intuition. But he does not suggest it as a method for arriving at
property identities or definitions. Consider his discussion of Sidgwick’s view that
“pleasure alone is good as an end.” This proposition—that pleasure is the good—
is held as “an object of intuition” by Sidgwick.6 Moore writes that his own
intuition arrives at a different result. “It may always be true notwithstanding;
6 Principia Ethica, p 126, § 45.
201
neither intuition can prove whether it is true or not . . . .”7 Moore is left trying to
offer reasons “capable of determining the intellect” rather than to offer conclusive
refutation or even indubitable intuitions in opposition to Sidgwick’s intuitively
derived thesis. If intuition is not a reliable way to arrive at the coextension
between goodness and some natural (or metaphysical) property, then it is surely
inadequate as a route to certainty with respect the identity of goodness and a
natural (or metaphysical) property.
Is there another way to support such a property identification? One
obvious one would be to mount a Leibniz’s Law style argument in favor of the
identity of goodness with a given property.8 This sort of argument would not
suffer the defect of an inference from co-instantiation of goodness and a natural
property to identity of those properties. Instead, it turns on the Identity of
Indiscernibles.9 An example follows:
1. Goodness has every property that pleasantness has.
2. Goodness has no property that pleasantness does not have.
3. If (1) and (2) then goodness has exactly the same properties as
pleasantness.
7 Ibid. Note also that Moore claims that, “propositions about the good are all of them synthetic
and never analytic.” Principia Ethica, p 58, § 6. If this is so, then Sidgwick’s claim, a proposition
about the good, cannot be true by definition. If it is true at all, it could only be contingently so. 8 This, it seems to me, might exhaust the possibilities. Coextension, even necessary coextension,
does not give sufficient reason to accept an identity. Intuition seems inadequate. Inferring from a
relationship weaker than coextension surely will not do. 9 Symbolized as ∀F(Fx � Fy) � x=y.
202
4. Therefore, goodness has exactly the same properties as pleasantness.
5. If goodness has exactly the same properties as pleasantness, then
goodness is identical with pleasantness.
6. Therefore goodness is identical with pleasantness.
The only unobvious premise in this argument is premise (5).10
(5) is an instance
of the principle of the Identity of Indiscernibles. This principle states that for any
two objects, if they share all of their properties, that is, if each has exactly the
same properties, they really are simply one single object. In this argument, the
principle is applied to properties. If two properties have exactly the same
properties, then, we might think, they really are one and the same property.11
While this argument runs afoul of one of the arguments I offered as a
possible reconstruction of Moore’s open question argument,12
if my claims about
Moore’s real intentions with that argument are correct, since it identifies goodness
with a simple property there can be no complaint that the argument’s conclusion
is barred by the open question argument. And it does not present a problem for
the naturalistic fallacy, because it does not involve any inference from the co-
instantiation or coextension of goodness and pleasantness.
10
(1) and (2) might not be obviously true, but their meaning is obvious enough for our discussion. 11
There are some reasons to doubt whether this principle is applicable to individual properties
and, in fact, whether the principle is true at all. Max Black has argued that a universe containing
only two exactly similar objects would prove the Identity of Indiscernibles false because while the
two objects would be indiscernible, they would clearly not be identical. “The Identity of
Indiscernibles,” Mind, New Series, Vol. 61, No. 242 (Apr., 1952), pp. 153-164 at 156-161. 12
The revised general significant question argument.
203
Of course, Moore would not accept this argument. First, Moore would
just reject premises (1) and/or (2). One easy way to show that (1) or (2) is false is
to provide a counterexample. If Moore could offer an example of a state of affairs
or an object that was pleasant, but not good, or vice versa, then that would show
that goodness and pleasantness are not identical. If such an example could be
found, then one property would have a property that the other did not possess.
For example, if donating blood was good, but not pleasant, then goodness would
have the property Being instantiated by instances of blood donation, but
pleasantness would not have this property. This would show that (2) is false, and
hence, the argument is would be unsound.13
Moore would be more inclined to take a different approach to refuting this
argument. He would deny both (1) and (2) but not on grounds that there is a
counterexample, but rather on the ground that goodness simply has a quality that
no natural property can possess. That is, goodness is a normative property and
pleasantness is not. This is, according to Darwall, Gibbard and Railton, the
enduring lesson of the open question argument.14
They believe that even if the
13
One must be careful when utilizing this strategy. This is, in essence, to turn the argument on its
head and rely on the non-identity of discernibles principle. The danger becomes apparent when
we consider another property that goodness has that pleasantness does not, namely, being believed
by Moore to be non-natural. It is well established that this sort of property causes trouble for
arguments utilizing the non-identity of discernibles principle. We need only think of the Lois
Lane, Superman/Clark Kent arguments to remind ourselves of this. 14
Stephen Darwall, Allan Gibbard, and Peter Railton, “Toward Fin de Siecle Ethics: Some
Trends,” The Philosophical Review, Vol. 101, No. 1 (January 1992), pp 116-117. Note that
Frankena arrived at this conclusion fifty years earlier in his contribution to P. A. Schilpp (ed.) The
Philosophy of G. E. Moore. See, W. J. Frankena “Obligation and Value in the Ethics of G.E.
204
argument is ultimately unsuccessful in exposing a fallacy,15
Moore has shown that
goodness is irreducibly normative.16
Since goodness is supposed to be unique in
this respect, we might conclude that goodness does not share all of its properties
with any natural property, so again (2) is false.
I expect that this is how Moore would reply, but it does expose him to yet
another claim of question begging. Nicholas Sturgeon has pointed out that even if
we grant that the open question argument successfully shows that ‘good’ refers to
a different property than any natural term does (a proposition that Sturgeon is not
willing to accept), this does not prove that goodness is non-natural. ‘Good’ is
certainly co-referential with ‘good,’ and if ‘good’ happens to refer to a natural
property, then Moore’s argument goes no way toward proving that goodness is
non-natural.17
Sturgeon argues that Moore has begged the question by assuming
that no natural properties have the normativity associated with goodness. Without
an argument that natural properties do not have normative force, Moore is
Moore,” in Schilpp, p. 102. And Prior reached the same conclusion in 1949. See Logic and the
Basis of Ethics (1949) p. 5. Note that Frankena also argues that the fact that goodness can be
defined in terms of “ought” means that it is not a simple property. I believe this is a mistake.
Frankena seems to be offering the wrong kind of definition. A definition in terms of “ought”
would be a conceptual analysis—it tells us a way to think about the property goodness—but it is
not a definition in the Moorean sense. Moorean definition requires an enumeration of the
component properties that make up the definiendum. I can give a definition of yellowness in
terms of greenness, but this does not prove that yellowness is not simple. 15
They, too, treat the open question argument as part and parcel with Moore’s claims about the
naturalistic fallacy. 16
See also, Connie Rosati, “Agency and the Open Question Argument,” Ethics 113 (April 2003):
490–527 at 495. 17
Nicholas Sturgeon, “Moore on Naturalism,” Ethics 113 (April 2003): 528–556 at 536.
205
foreclosed from arguing against the Identity of Indiscernibles argument on the
grounds set forth in the preceding paragraph.18
Finally, Moore might object to the argument on the ground that he does
not accept the method as a way to define goodness. In all of Principia Ethica, we
never see an instance of identifying the properties that goodness has. There are,
however, certain features asserted of goodness. For instance, Moore asserts that
goodness is the property that is the proper subject matter of Ethics and that it is
that property whose instantiation in a state of affairs gives us a reason to act in
such a way as to bring about that state of affairs. But these are not intrinsic
properties of goodness. These are relational properties of goodness. They tell us
things about how goodness is related to our philosophizing, or about its causal
relation to our motivations,19
but they do not tell us about the nature of the
property itself. On Moore’s scheme, this is a consequence of the simplicity of
goodness. If the property had intrinsic properties which we could enumerate, it
would have parts, and be susceptible to the sort of analytic definition Moore
claims is impossible. So, in fact, the sort of enumeration of properties required to
motivate this Identity of Indiscernibles argument is ruled out as a consequence of
the open question argument.
18
This is not quite true. Moore could easily make such an argument for any particular natural
property. Sturgeon’s argument only forecloses claiming that (2) is false on the ground that no
natural property is normative. 19
Assuming for purposes of this example only that we are motivated to do that which we ought to
do.
206
The same is true for natural properties. Consider, again, the simple
property yellowness. We can assert of yellowness that it is the color of lemons
and sunflowers, or that paint instantiating this color, when mixed with blue paint,
yields green paint. These properties are not intrinsic properties of yellowness, but
rather, they tell us how yellowness is related to other things. More importantly,
they are not necessary to yellowness, and hence, do not uniquely identify that
property.
In all, these are rather potent reasons to reject this sort of approach to
defending a property identity for goodness. This suggests that the only way open
to naturalists20
who wish to define goodness is to proceed from the co-
instantiation of goodness and their preferred natural property, and in so doing,
commit the naturalistic fallacy. If my interpretation of the naturalistic fallacy is
correct, committing the naturalistic fallacy does not guarantee that the proffered
view is false. It merely means that it cannot be supported by a well-justified,
deductive argument.
II. Overall Conclusions
The purpose of this dissertation was to offer an interpretation of what it
means to commit the naturalistic fallacy that is maximally consistent with Moore's
text and examples, to distinguish Moore's claims about the naturalistic fallacy
20
Again, we are discussing reductive naturalists here, those who attempt to define goodness in
terms of natural properties.
207
from the open question argument, and to correct mistaken interpretations of the
open question argument. In pursuing these goals, I found that the naturalistic
fallacy and the open question argument are complementary, rather than redundant,
arguments. Moore uses the open question argument to show that goodness is not
identical with any complex natural or metaphysical property. But this still leaves
the possibility that goodness is identical with a simple natural or metaphysical
property. His claim that philosophers who attempt to define goodness commit the
naturalistic fallacy undermines support for any such identification, even if it does
not conclusively refute such definitions.
My first chapter provided justification for this dissertation. I noted that
Principia Ethica has had an enduring legacy, though there is little agreement on
what Moore claimed or how he attempted to support his claims. I identified
Moore’s discussion of the naturalistic fallacy as the most celebrated yet least well-
understood portion of the text, and so began my examination of his moral
philosophy with this concept.
In Chapter 2, I explained the rudiments of Moore’s view on the nature of
goodness, and considered some implications of his view. Goodness is a simple
property, meaning it has no constituent parts, and hence it is indefinable. I also
explained what a successful definition might look like for Moore.
In Chapter 3, I examined rival interpretations of the naturalistic fallacy
provided by Baldwin, Prior and Frankena, and interpretations provided by Moore,
208
himself, finding that none of them maintains high fidelity to the text of Principia
Ethica. Baldwin offers three possible interpretations, none of which were
particularly appealing. Each relied on taking a particular bit of Moorean text out
of context. Where the interpretations were consistent with the text, they did not
offer a fallacy, that is, there was no bad inference being made.
Prior offered two interpretations similar to the view I defend. One of
Prior’s interpretations—that to commit the naturalistic fallacy is to infer from the
necessary coextension of goodness and some natural property that they are
identical—seems to me to be a correct statement of one version of the naturalistic
fallacy. However, Prior attributes far too much to Moore, claiming that the
naturalistic fallacy is Moore’s entire argument that all forms of naturalism are
false. Naturally, Prior finds that the naturalistic fallacy is not well suited to this
task. What Prior did, we saw, was to conflate the naturalistic fallacy and the open
question argument, which together, make some significant strides (yet, fall short
as we have seen) toward proving non-naturalism.
Finally, we looked at Frankena’s influential views on the naturalistic
fallacy. Frankena takes the most methodical approach of Moore’s critics, but
ultimately attributes too much to Moore’s claims about the naturalistic fallacy as
well. Foreshadowing Prior, Frankena takes the naturalistic fallacy to be Moore’s
argument against all forms of naturalism, and finds that such an argument cannot
be sustained without assuming non-naturalism from the start. I argued that Moore
209
need not provide an argument that naturalism is false before he accuses a
philosopher of committing the naturalistic fallacy, because committing the fallacy
does not entail the falsity of the view. The naturalistic fallacy was not intended to
refute any theory, but rather to show that any attempt to define goodness is based
on a bad inference.
In Chapter 4, I offered a close read of Moore’s statements wherein he
describes the naturalistic fallacy and of his examples of philosophers who commit
the naturalistic fallacy. I found that in each case, to commit the naturalistic
fallacy is to infer from either the coextension (necessary or otherwise) of two
properties or the mere supposition that all instances of one property are also
instances of another, that the properties are identical. My interpretation of the
naturalistic fallacy suggests that while it is a common enough error, it is not well-
suited to play the role that many of Moore’s commentators assigned to it. As we
have seen above, it does have a role, and the contours of that role are shaped by
the open question argument.
In Chapter 5, and concluding in this chapter, I offered a reading of
Moore’s open question argument that emphasizes fidelity to the text (despite the
fact that I retain the ill-deserved name) and rejects the traditional interpretation.
Rather than proving that all naturalistic definitions of goodness are wrong, I find
that the argument really only demonstrates that no complex definition is correct.
Where Moore does deploy an argument that can rightfully be called an open
210
question argument21
against a simple definition, it fails, and for the usual reasons
cited.
We are left, then, with an array of arguments, some of which are fairly
robust, in support of the proposition that goodness is a simple property. It seems
likely that one of the arguments that is effective against complex definitions will
be effective against some simple ones as well, but it is not clear that this was
Moore’s intention. Nevertheless, we are now in position to deploy the naturalistic
fallacy against the naturalist. Any naturalist who offers a simple definition of
goodness, i.e., a property identity in terms of a simple property, is likely to
commit the naturalistic fallacy, thereby undermining support for her position.
This does not disprove all versions of reductive naturalism, but it suggests that no
version can cite as support for its definition a robust deductively valid argument.
There will always be some reason to doubt the inference from the co-instantiation
of goodness and the proffered natural property to their identity.
21
In sections 14 and 26.
211
Bibliography
Auyang, S. Y., How is Quantum Field Theory Possible? (New York: Oxford
University Press 1995) cited in Michael Esfeld, “Holism in Cartesianism
and in Today’s Philosophy of Physics,” Journal for General Philosophy of
Science / Zeitschrift für allgemeine Wissenschaftstheorie 30 (1999), pp. 17-
36.
Baldwin, Thomas. G.E. Moore (London: Routledge, 1990).
Baldwin, Thomas. “Editor’s introduction” Principia Ethica Thomas Baldwin, ed.
(Cambridge: Cambridge University Press, 1993), pp. ix-xxxvii.
Black, Max. “The Identity of Indiscernibles,” Mind, New Series, Vol. 61, No. 242
(Apr., 1952), pp. 153-164 at 156-161.
Brandt, Richard B. Ethical Theory: The Problems of Normative and Critical
Ethics (New York: Prentice Hall, 1959), p. 164-65.
Clarke, Mary E. “Cognition and Affection in the Experience of Value,” Journal of
Philosophy Vol. 35, No. 1 (Jan., 1938), pp. 5-18, cited in Frankena (1939),
p. 465.
Darwall, Stephen; Gibbard, Allan; & Railton, Peter. “Toward Fin de siecle Ethics:
Some Trends” The Philosophical Review, Vol. 101, No. 1 (Jan., 1992), pp.
115-189.
Feldman, Fred. “The Open Question Argument: What it isn’t; and What it is”
Philosophical Issues, 15, Normativity, 2005, pp. 22-43.
Fischer, John. “On Defining Good,” The Journal of Philosophy Vol. 51, No. 23
(Nov., 1954), pp. 730-736.
Frankena, William K. “The Naturalistic Fallacy” Mind Vol. 48, No. 192 (Oct.,
1939), pp. 464-77.
Frankena, William K. “Obligation and Value in the Ethics of G.E. Moore,” in
Schilpp (1942), pp. 93-110.
Goodenough, Ursula. “The Holes in Gould's Semipermeable Membrane Between
Science and Religion,” American Scientist, (May 1999): Online source,