. 5
( 10)


each class, can be made to seem signi¬cant using Ross™s method. If other
sorts of value are otherwise equal, why wouldn™t it be better for there to
be a healthy rhododendron, a satis¬ed whim, a ¬‚ourishing corporation,
or better intonation in the woodwinds? Of course, these new sorts of
value will con¬‚ict in real cases, which set the health of one organ or the
success of one institution against another. But equalized cases of the sort
that Ross deploys suggest that they have some sort of value nonetheless.
However, my point is that it need not be basic normative value, or even
value of any ethical or political signi¬cance.
We cannot tell using the method of equalized cases the nature of the
value in question “ whether, for instance, the value has basic signi¬cance
for ethics, or whether indeed it re¬‚ects forms of assessment that can be
justi¬ed at all. And yet we saw that concrete cases alone are deceptive.
So what are we to do? We might retreat to the claim that commonsense
intuition can tell us nothing of signi¬cance. But I believe that we can do
somewhat better. We did somewhat better even in the last two sections “
for instance, in our treatment of truth and unrecognized satisfactions,
where we saw enough pattern in our intuitions about relatively concrete
cases to suggest some intuitive advantage for hedonism. But clearly some
further development of the method of abstract cases is desirable.
As a ¬rst step, we should develop Ross™s abstract intuition tests so that
they treat a wider range of cases than merely simple equalized pairs. If
we consider abstract trade-off cases, in which we force our intuition to
balance the signi¬cance of more of one good and less of another, we may
still be unable to fully assess the intuitive nature of the values involved, but
we can get a better sense of their intuitive signi¬cance and depth, of their
overall importance when the chips are down. Important and dominant
good at least points in the direction of the sort of basic value that we seek.

53 Indeed, he seems to explicitly discount the signi¬cance of disease on page 134.

Ross himself takes some steps in this direction.54 But, since his treat-
ment is the best intuitionist treatment of these issues that we possess, it is
unfortunate that his development of the topic is incomplete and in some
particulars not very convincing. It would indeed be a very large task to
treat abstract trade-off cases properly. Still, we can gain some insight from
Ross™s attempt.
What I am calling trade-off cases are pairs of situations in which it
is speci¬ed that, holding all other sorts of value constant (and all atten-
dant counterfactuals and probabilities relevantly identical, and hence their
contribution to intuitive value identical), one situation is clearly worse in
respect to one sort of value but better in respect to a second, while the
second situation is clearly better in respect to the ¬rst sort of value and
worse in respect to the second. To prevent various concrete irrelevancies
from in¬‚uencing our intuition, and so that these pairs capture at once
a range of instances, we should at least as a ¬rst stab articulate the pairs
in something like the pure and abstract manner that Ross deploys. We
should develop abstract trade-off cases.
While it is important that the cases we consider be abstract and pure, it
may be that there are some additional concrete conditions that are useful
in avoiding obviously misleading currents in our intuition when dealing
with the signi¬cance of certain goods like pleasure. For instance, perhaps
it is important that the pleasure in question not be our own, because
our normative intuitions are not cleanly engaged in cases where our own
pleasure is at issue, in part because of natural sel¬shness and in part because
of the desire to avoid obvious sel¬shness. But such concrete additions may
themselves be dangerous and misleading, and perhaps suf¬cient abstraction
can also guard against most misleading elements in our responses. So let™s
attempt this method.
To narrow our treatment to the cases that seem most signi¬cant for
our purposes, I will consider trade-off cases that range across the goods
deployed in the alleged counterexamples to hedonism discussed in the last
two sections and also the goods that Ross™s own discussion suggests are
intrinsic goods. Ross fails to isolate the intuitive intrinsic value of health
that some objectivists may favor, and also that of unrecognized desire
satisfaction. But, as I said, he does isolate virtuous action, knowledge
(implying true belief), and the concordance of pleasure and virtue. He
also suggests, in a manner apparently supported by his test, that pleasure is

54 Ross (1930: “VI: Degrees of Goodness,” 142“154).

an intrinsic good somewhat unlike the others. It is an intrinsic good only
when it has appropriate causes, when it isn™t vicious pleasure. We should
consider this quali¬cation on hedonism as well.
There are two relevant kinds of trade-off cases, those that compare
goods within the abstract categories just noted and those that make cross-
category comparisons. Consideration of the ¬rst sort of pairs may some-
times help put the second sort in the proper light. Consider health. Any
genuine foundational role for health requires some account of how the
health of one cell or organism or organ or ecosystem trades against another.
Nevertheless, let me focus on the crucial cross-category comparisons “
in the ¬rst instance, the trading of pleasure for health.
Here we can be guided by our discussion in the last section, which was
at least implicitly abstract. Imagine the relevantly abstract pair of situations.
The kind of health that isn™t eventually positively re¬‚ected in the hedonic
tone of the creatures involved, the kind our pair introduces, seems for that
very reason of lesser signi¬cance. Intuitively, the health of the sentient is
the health that intuitively matters, and its worth is at least re¬‚ected in its
effects on that sentience, mediated perhaps by preference. Intuition cedes,
it seems, the dominance of pleasure over health in this abstract trade-off
case. So Ross was right to ignore these trade-offs. The method we are
attempting works for these cases, but doesn™t reveal an interesting new
basic good.
What of unrecognized desire satisfaction? Here, as we saw at least
implicitly in section II, there are no ¬rm commonsense intuitions that
resolve these issues abstractly and at once in favor of competitors to hedo-
nism. Rather, there are con¬‚icting intuitions, which, for instance, can
slide up or down the slippery slope between unrecognized satisfactions
received after death and those received during life. There isn™t a uni-
form intuition abstractly rooted. Concrete details matter in con¬‚icting
ways. I believe that we also saw on balance in section II that cases of this
general sort, indeed of a more general sort that includes the signi¬cance
of true belief to well-being, favor hedonism. But the methodology of
abstract cases itself apparently fails to achieve adequate resolution in this
So far, abstract trade-off cases have yielded little. But there is a still more
re¬ned version of the method that we might attempt. It might be that
relatively general factors found in the concrete cases that favor hedonism
regarding the signi¬cance of true belief and unrecognized satisfaction in
the preceding sections can somehow be parlayed into specially relevant
general conditions, which allow us to specify extra but still fairly general

constraints on what we might call “relatively abstract” trade-off cases.
The conditions might allow our intuition some determinate purchase,
yet be conditions that we can see to be appropriately revealing and not
inappropriately misleading.
Still, we shouldn™t be con¬dent of that possibility for trade-offs involv-
ing unrecognized satisfaction, since the factors to which I appealed in
section II don™t seem to be of the proper form to allow it. We instead
relied, for instance, on the ¬rmness of intuitions about well-being after
death and on abstract similarities with other cases where intuitions were
less ¬rm. So again, the fact that Ross ignored these sorts of cases may
seem hence reasonable. They don™t seem good instances for his method
or anything in its rough vicinity.
Yet the issue of unrecognized satisfaction is closely related to that of
true belief, and this we can treat by reference to Ross™s explicit discussion
of that issue. So perhaps this provides an opening for something like Ross™s
method. Consider knowledge.
Here the ¬rst sort of abstract trade-off tests, which probe trade-offs
within a single category of good, are quite interesting. Ross attempts to
develop an account of how various forms of knowledge and true belief
are properly to be traded against other forms, and while it is perhaps
not convincing, we should be sensitive to the obvious dif¬culties that he
faced. To work things out properly we would need, for instance, to con-
sider the relative weight of truth as opposed to the third condition beyond
belief required for knowledge. I ¬nd it hard to believe that the presence
or absence of false barns in one™s vicinity, which consensus now holds
relevant to the third condition, could plausibly be held relevant to one™s
well-being, but still perhaps something beyond the truth of one™s belief
might intuitively be so relevant. Ross considers such factors as the degree
of match of conviction and evidence and also the generality of the content
of the beliefs in question. Issues about content are especially complex. It
may seem that deception on some matters is more intuitively harmful
to individual well-being than deception on other matters, whether an
individual cares about that difference or not. But yet it also seems unsup-
portably paternalistic to ignore individual tastes and preferences about the
relative seriousness of different forms of deception. There are many dif-
¬culties, and I commend Ross™s discussion to your attention, though it
clearly requires further development.
But, however that development goes, the crucial cases for our purposes
are relatively abstract trade-offs across categories of goods: of pleasure, on
the one hand, and life in the truth on the other. How much suffering

is more true belief worth? My own even fully abstract intuition is that
true belief bought for suffering is of no deep and intrinsic worth to
well-being, and is only good in an obvious epistemic sense that is of no
intrinsic moral signi¬cance. I have a ¬rm intuitive reaction to even a fully
abstract trade-off case of this sort. But perhaps that is idiosyncratic. Still,
clearly the fully abstract trade-off does not count against hedonism in any
de¬nite way. We saw at least that in section II. Ross himself ¬nds this
general issue too dif¬cult, and so retreats to the claim that the desire for
knowledge is a virtue and that is hence why knowledge is of real, but
hence somewhat indirect, signi¬cance.55 Our central immediate concern
is whether relatively abstract trade-off cases can be developed in a manner
that provides intuitive resolution of this trade-off across categories. Still,
the situation of that test regarding true belief seems no better than that
regarding unrecognized satisfaction. It isn™t obvious that the concrete cases
that on balance support hedonism over competitors in this trade-off, as we
saw section II, have general features that can appropriately be imported
into relatively abstract trade-off cases. We seem left in this case with the
piecemeal method of the preceding sections.
Perhaps the crucial and most troubling cases for hedonism involve the
foundational signi¬cance of virtuous action, the just commensuration of
pleasure to virtue, and the related concern that vicious pleasures shouldn™t
count normatively. Some concerns about a foundational signi¬cance for
these things are in effect objections to consequentialism. Some are con-
cerns about the value of overall states of affairs that include pleasures and
pains as parts. They hence do not seem to be concerns about what I have
called the basic normative value of pleasure itself. Nevertheless, we can
get some grip on the intuitive relevance of these things to basic value by
a consideration of abstract trade-off cases, or at least by a consideration of
relatively abstract trade-off cases. It is less than ¬rst may meet the eye.
If we consider a pair of situations, with the attendant counterfactuals
and probabilities speci¬ed as relevantly identical, such that the ¬rst clearly
involves more suffering and less pleasure but also more virtuous action (and
perhaps knowledge) and also more intuitively commensurate suffering for
the intuitively guilty, my own honest but not ¬rm intuition is that the ¬rst
is worse overall than the second. Still, it does seem plausible to insist that
some of the relatively concrete details will matter if we are to establish any
¬rm and clear intuitive reaction. If we specify that it is the vicious only

55 Ibid., 151“152.

who suffer more, or that the suffering replaces a vicious pleasure, then
my hedonist reaction is undercut. But if, on the other hand, we specify
that it is the innocent who suffer more, my hedonist intuition is stronger
still. The intuition in question is even stronger if the sufferer is one of the
worst-off. And indeed, if the “guilty” or “vicious” person is the worst off,
this undercuts even my weak intuition that perhaps the guilty or vicious
should suffer more, since it seems that in the scenario in question they™ve
already suffered enough to count as appropriately innocent and should
properly grab for what they can get.
Without some relatively concrete speci¬city, fully abstract trade-off
cases are apparently indecisive in this instance. But in this situation I claim
that at least a relatively abstract trade-off case is available and effective.
There are relatively abstract features that seem relevant and can resolve
our intuition. What™s relevant is whether the extra suffering is for the
relatively worse off.
But not everyone will agree. The issue really becomes whether this
is the proper relatively concrete speci¬city to introduce into these trade-
offs. This focus on the suffering of the worst-off may seem to some an
arbitrary and misleading concretizing of the pairs, because if we focus
on extra suffering for the well-off and especially the well-off guilty or
vicious, then we™d have other intuitions. I will argue in the next part
that a focus on the suffering of the worst-off is not arbitrary here. I will
argue, following Rawls and Scanlon, that it is the appropriate way to
consider complex trade-offs of well-being and re¬‚ects the proper means
of summing value into the overall value of states of affairs. When we
cannot form an overall intuitive judgment about such a trade-off pair
described with full abstraction, then consideration of the worst-off is the
appropriate relative concretization to make. In this case, though not in
the case of life in the truth, an effective and appropriate relatively abstract
trade-off case is available, and it supports hedonism. That is because the
vicious and the worst-off may not be the same individual, so methods
of balancing the interests of the worst-off versus other moral interests
become most clearly relevant. A relative concretization of this sort may
also allow us some purchase even on the relative general importance of
truth and suffering, but not in the most characteristic and telling location
to resolve that particular issue, where the suffering and the deceived are
the same person. So in that case we are forced back to the intuitional
method of the last two sections.
My argument that the kind of relative concretization that supports
hedonism is appropriate is yet to come. But there are also other grounds

to believe that any relatively abstract trade-off case that introduces greater
suffering or less happiness for the vicious will be misleading, that there
cannot be a proper relatively abstract trade-off case that supports the view
that the pain of the vicious has positive value or that their pleasure should
be discounted. Any case that would properly suggest this would have to
be suitably pure. And that is certainly hard to assure when such intuitive
punishment for the guilty is involved, and indeed is probably impossible.
That is because the phenomenon of punishment introduces all sorts of
complications. For instance, in any relevant trade-off case, the extra pain
of the vicious could have no positive deterrent effect. So it would be
crucial that all the facts other than the extra pain be preserved in the two
situations that are compared, including all the hypothetical truth about
what would concretely occur if the vicious person did not suffer at all.
The extra pain would have to be in that way gratuitous. But even that
would not be enough to assure proper purity in the case. Remember that
MAC implies a conception of the duties of reparation and of proper pun-
ishment that introduces a kind of derived normative value for punishment
of the guilty even in the absence of deterrence. It would be important
for any relatively abstract trade-off case that is not misleading to assure
that only comparisons of basic value are in question. And when intuitive
punishment is involved, that seems impossible.
But independent of my argument in the next part that proper con-
cretization involves a focus on the worst-off, and independent of the
argument that greater suffering for the vicious is a misleading concretiza-
tion, there is also the following general and methodological point: The
proper development of the abstract method of this section requires an
understanding of the general and intuitive theoretical rationales for intu-
itive responses, and not just those particular intuitions themselves.
In the case at hand, as I™ve said, I believe that such a rationale is available
to us. Kant says that “one has never heard of anyone who was sentenced
to death for murder complaining that he was dealt with too severely and
therefore wronged; everyone would laugh in his face if he said this.”56
But let me vary the ¬gure. We might address a relatively badly off suffer-
ing person and suggest that they might have been better off, but then the
world would have needed to include more lying or ignorance or a more
striking mismatch between virtue and happiness, even though for some
reason all that wouldn™t in fact have made anyone else suffer more. If we

56 Kant (1996b: 475).

concluded, “Things really are better this way, though you must, regret-
tably, suffer more,” then I think we should reasonably and intuitively feel
very uncomfortable about that. Frankly, it would seem appropriate if they
did worse than chuckle. The next part will explain why this point isn™t
merely rhetorical, why it is appropriate to require that relatively abstract
trade-off cases assume this particular form.
Absent a consensus on such a rationale, intuitions will differ and cannot
resolve the issue. That is the most crucial point of this section. But I also
think a comparison to what Ross says puts my own intuitions, supported
by a rationale or not, in at least a relatively good light:
I think . . . that pleasure is de¬nitely inferior in value to virtue and
knowledge. . . . Most people are convinced that human life is in itself something
more valuable than animal life, though it seems highly probable that the lives of
many animals contain a greater balance of pleasure over pain than the lives of
many human beings. Most people would accept Mill™s dictum that ˜it is better to
be a human being dissatis¬ed than a pig satis¬ed™. . . . Many people whose opinion
deserves the greatest respect have undoubtedly thought that the promotion of the
general happiness was the highest possible ideal. But the happy state of the human
race which they aimed at producing was such a state as the progress of civilization
naturally leads us to look forward to, a state much of whose pleasantness would
spring from such things as the practice of virtue. . . . [I]f they thought the state of
maximum happiness would be one whose happiness sprang from such things as the
indulgence of cruelty, the light-hearted adoption of ill-grounded opinions, and
enjoyment of the ugly, they would immediately reject such an ideal. . . . [What]
amount of pleasure is precisely equal in value to a given amount of virtue? . . . [I]t
seems to me much more likely that no amount of pleasure is equal to any amount
of virtue.57

Some analytical remarks about this paragraph. From Ross™s sensitive
general method, we ¬rst get a retreat to the single and misleading con-
crete case of the contented pig, which is not supported (like relative con-
cretization involving the worst-off ) by a suitable rationale. It is so sup-
ported in Aristotle, but Ross lacks his account of a species-speci¬c human
telos. Then we get an implausible prediction about the opinions of oth-
ers, citing grounds for their rejection of pleasures that are to the contrary
embraced by many contemporary fans of juissance. And we end, worst
of all, with a wildly counterintuitive and even offensive general claim.58

57 Ross (1930: 149“150).
58 This is the guts of Ross™s discussion of these trade-offs. But he makes some other arguments,
and I commend his overall discussion to your attention. Ross (1939: 252“310) revisits the

How dare we say to the suffering that no amount of their suffering would
make the world worse when traded for a little more virtue somewhere, or
a bit more general knowledge, or a bit more suffering for the intuitively
But, as I said, Ross does provide the best and most thoughtful anti-
hedonist discussion of the good I know that is rooted in intuition and
couched at the proper level of generality and purity. It just isn™t good
We can properly conclude three things, in decreasing order of certainty.
First, there are no extant competitive conceptions of the good or well-
being that possess a de¬nite overall intuitive advantage over hedonism.
This is not grounds, of course, to stop trying for something better than all
current competitors. But the competitors span the three types deployed
in standard characterizations of possible conceptions of the good or well-
being. And until we have a coherent account of all our intuitions, we
have no reason to believe it to be possible. Hence we come to my second
conclusion. Many ethicists believe that they know, on grounds of intuition,
that hedonism is false. But they know no such thing. Third, at least if the
rationale for a speci¬c appropriate concretization of relatively abstract
trade-off cases that I have promised in this section is available, but to a
large degree even if it is not, hedonism is better supported by our overall
intuition than its competitors.

good. But the overall thrust of his remarks is defensive elaboration, rather than the
development of a systematic alternative to the account in the earlier book. One relevant
detail is on page 275, where he suggests that his earlier claim that any amount of virtue
trumps any amount of pleasure is not mistaken, but that he should have concluded from it
that pleasure and virtue are different sorts of value.

Natural Good

Hedonism is in adequate accord with our normative intuitions. But I
have also promised a direct argument, independent of appeals to normative
intuition, for the truth of that element of the Hedonic Maximin Principle.
That is the task of this chapter.
Here is the short version: Ethical discourse is justi¬catory reason giving.
And this requires hedonism, in the world we inhabit, whether we like
it or not and whatever our normative intuitions. That is because pain
and pleasure involve the only unconstituted natural normative properties
found in the world.
This is a new metaethical alternative, within the cracks between familiar
views. Justi¬catory reason giving has a metaphysical cost, but because there
is this new metaethical alternative, we can pay it.
You may reasonably worry that hedonism is antecedently more plausi-
ble than any of the controversial metaphysical and metaethical claims that
I will make here in support of it. But the nature of ethical discourse as
justi¬catory reason giving requires an objective and asymmetrical vindi-
cation of its key normative claims over possible competitors. Normative
intuitions alone are not enough. We must develop some understanding
of what conditions might make our intuitions suitably true or otherwise
asymmetrically appropriate, and see that those conditions are in fact plau-
sible. And the only obvious thing that could provide such a vindication of
a speci¬c conception of the sole basic normative value is the existence in
our world and all relevant alternatives of merely that type of basic value.
While the story I will tell here is inevitably controversial, it is a plausible
story that would deliver the truth of hedonism. And I will argue that there
is no other suitably plausible story that can deliver the truth of hedonism
or any competitor.

Section I sketches background for my metaethical proposal. Section II
presents hedonic value as unconstituted natural good. Sections III through
V develop crucial details that matter later on, including the interpersonally
comparable ordinality of that value.


Cognitivism holds that normative sentences can be true or false of the
world in the manner of ordinary declarative sentences about ice cubes in
a glass of lemonade. My hedonism involves a form of cognitivism, but not
a familiar form. It is not a kind of non-naturalism, nor is it a kind of what
I will call “constitutive naturalism”. And yet it bears certain af¬nities with
both. It holds that there are basic properties “ by which I mean properties
not constituted by others “ that are at once natural and normative. It is,
in particular, the view that physical pain and pleasure often involve such
Before I sketch this view, it is helpful to consider the range of currently
dominant metaethical alternatives. Each of those alternatives falls prey to a
variant of the same abstract argument. None can reconcile a plausible and
widely shared conception of the world, on the one hand, and the demands
of justi¬catory reason giving, on the other. This provides motivation to
pursue my metaethical proposal, even beyond its truth. Of course, I cannot
hope to disprove all standing metaethical views within a single section. But
it will be worth articulating a general worry about these familiar views
in order to provide necessary contrast and motivation for my positive
There are two key presumptions of my abstract negative argument.
First, normative practice, by which I mean our practice of ethical and
political evaluation, has as a central and indefeasible commitment some-
thing I have called “justi¬catory reason giving”. Justi¬catory reason giving
crucially involves (a) normative claims that express justi¬catory reasons for
or against things, which reasons are (b) governed by consistency and other
logical constraints, and are also (c) capable of something we might call
“deep justi¬cation”. To give a deep justi¬cation of a normative claim is
to show that no con¬‚icting claim is appropriate, that there is an objective
asymmetry that vindicates a practice of reason giving deploying the ¬rst
against a practice deploying the second.
Let me explain some of this jargon. The peculiarly normative feature
of normative discourse, of ethical and political discourse, is its capacity
to provide an articulation of reasons as justi¬cations. This is a relatively

complex feature of that discourse, and it is related closely to, but yet
distinct from, other aspects.
Reasons as justi¬cations are not simply reasons as motives. We should
grant that normative utterance expressing ethical or political evaluation has
a peculiar capacity to motivate people. Coming to believe that someone
has done wrong can invoke complicated and potentially powerful emo-
tions, which some have thought that all normative utterance expresses.
And normative utterance can be used to command people, to push them
around, because it not only expresses emotion, but also engages it. And
not only emotion but also wants and desires seem closely connected to
normative utterance or judgment. It seems at least unusual to judge some-
thing wrong and have no motivation at all to avoid it. Also, I have noted
ways in which benevolence and reciprocity are key moral motives. But
there is no simple connection between motivation and the normative. It
is possible to judge something wrong and yet do it, and indeed want to do
it. It is possible to feel very guilty about something and yet know that it is
right. And so the crucial normativity of moral utterance re¬‚ects its con-
nection not to reasons as motives to act, but to reasons as justi¬cations of acts.
If such things as reasons as justi¬cations can be made sense of independent
of motivation, and we will see that they can, then the motivational force
of normative claims might be a rather contingent matter, not essential to
their meaning in any strong sense. But in any case, we must distinguish
between justi¬catory reason giving and providing reasons as motives. We
may not want to do what we ought to do.
Ethical theories should not rest on bad puns. Words are ambiguous, and
sometimes we fail to note the ambiguities. “Reason” and “rational” are a
crucial case in point. Some theorists simply con¬‚ate reasons as motivations
with reasons as justi¬cations. But there are other, subtler con¬‚ations of
which we should also be wary. Some say that we must attribute thoughts
to persons and interpret their utterances under a constraint of “rational-
ity”, that we must assume that thinkers and speakers will rationally infer,
say, P and Q from P and from Q, and also move in a rational way from
wanting P and believing that Q is a means to P to doing Q.1 But ratio-
nality seems to involve reasons, and reasons to introduce normativity of
some sort. Still, it is important to see that this sort of rationality, which
may be crucial to interpretation, isn™t “normative” in anything like the

1 Since people sometimes fail to make obvious inferences, perhaps we should ascribe thoughts
and interpretations that make them seem as rational as possible. Only if they fail to be
minimally rational will they lack thoughts and meaningful speech.

traditional sense appropriate to the good and right. If I think that it would
have been rational, in the sense relevant to interpretation, for Hitler to
invade England before he invaded the Soviet Union, I am not normatively
endorsing his invading England in any sense. I am not claiming that he
had any justi¬cation for that. Rather than infer from his premises or act
on his desire, I think he should have shot himself sooner or gone back
to painting. The “normativity” of the right and the good, which some
people think cannot be constituted out of mundane concrete things, has
nothing obvious to do with the rationality that allegedly constrains inter-
pretation. Reasons even as coherent motivations are not obviously reasons
as justi¬cations.
Both reasons as motivations and rationality as coherence are different
phenomena than reasons as justi¬cations. They may be entwined with
reasons as justi¬cations, and they may overlap, but they are at least some-
what different. It is reasons as justi¬cations that are our concern here, the
feature of our normative practice on which my argument rests.
Justi¬catory reason giving in this sense is a deep feature of our nor-
mative practice. As I said in Chapter 1, it is deep enough that skepticism
about its possibility can generate a corrosive skepticism about ethics and
the normative in general. It is indeed arguably the most central and crucial
feature of our normative practice. But whether it is most central or not,
it is surely central enough that at least many of us would conclude, if
this commitment of ethics cannot be vindicated, that ethics is a kind of
scam, just a lot of hot air. This is in fact what many philosophers and
nonphilosophers believe, and for something recognizably like this reason.
We can see in this way that justi¬catory reason giving is quite central to
our practice of normative evaluation, that its loss would be enormous and
And in proper justi¬catory reason giving, any legitimate normative
claim must be capable of asymmetrical vindication relative to possible
competitors. Justi¬catory reasons are reasons for and against things that
themselves can be vindicated. Normative evaluation is a process of eval-
uation that yet can itself withstand re¬‚ective evaluation and demands for
legitimation. Those who come to think justi¬catory reason giving impos-
sible or nonsensical often become skeptical about the entire normative
enterprise, and people become skeptical about justi¬catory reason giv-
ing because it comes to seem arbitrary that certain things are counted
as reasons for and certain things as reasons against, because it comes to
seem impossible to provide the kind of deep re¬‚ective legitimation of our
¬rst-order standards of right and wrong that is characteristic of at least our

form of justi¬catory reason giving. I grant that there might be creatures
who pursue something we might appropriately, if loosely, call “justi¬-
catory reason giving” that does not demand the sort of deep re¬‚ective
justi¬cation specially characteristic of our form. Indeed, our own norma-
tive practice seems to include certain people whose spade is turned fairly
early, who can rest content with the assurance that something is required
by our moral point of view, by our ¬rst-order practice of calling things
good and bad, without requiring any deeper legitimation of that point
of view. But if we are unable to make sense of some kind of legitimation
of our particular ¬rst-order normative practices that drives quite deep,
that rests on no starting points about which a certain arbitrariness can be
suspected, on no starting points that have genuinely credible competitors,
then the corrosive skepticism about the normative that many reasonable
and re¬‚ective people feel seems to me quite appropriate.
This ¬rst key presumption of our argument comes down to the claim
that proper normative principles must be capable of being supported by
what I have been calling a direct argument. A mere appeal to normative
intuitions that have obvious possible competitors is not enough.
But there is also a second key presumption of our argument here. I will
presume, with the consensus of current philosophers, that it is, at least at
base, a concrete world, made up of the down-to-earth and familiar things
that appear in our ordinary sensory experience, or at least via the various
sorts of experimental apparatus with which we now extend our experi-
ence. These are familiar things that we can taste, see, smell, measure, kick,
and manipulate with instruments. I presume that there are concrete trees,
tables, rocks, and humans, with spatio-temporal locations and relations,
perhaps colors and other sensory properties, and maybe causal powers.
And these, I presume, are in turn made up of concrete spatio-temporal
bits. For instance, humans are made up of complex but concrete living
cells, and these in turn perhaps of little concrete particles with causal ten-
dencies to de¬‚ect one another and move oil drops. Even some dualist and
idealist conceptions, which deploy ghostly but quasi-physical substances
and powers, or sense data with concrete sensory properties, are consistent
with this concrete notion of the world, but physicalist conceptions are
perhaps the paradigmatic such conceptions.
Somehow everything, even complex political, social, and economic
phenomena, is constituted out of relatively concrete and familiar things
and their machinations and relations. Or at least that is so to this degree:
Any two complete situations of the sort with which our analysis need
concern itself, any two “metaphysically possible worlds”, are such that if

they are qualitatively identical in all their concrete micro-detail (say, the
way in which the particles that make them up are arranged and ruled by
basic physical laws), then they are qualitatively identical in all other ways,
or at the very least in all other ways relevant to ethics. Among metaphysi-
cally possible worlds, any two that are identical in all concrete micro-detail
are identical in all details. While there are competing conceptions of the
world, they involve quite controversial, non-commonsensical, and at least
prima facie implausible resources, strange abstracta like irreducible num-
bers or propositions, or strange non-natural properties that we can only
access by bizarre faculties of Platonic intuition, strange properties that
can vary while all the spatio-temporal concrete structure and detail of
the world yet remains unchanged. That is why this second presumption,
which we might put as the claim that everything is constituted out of
the natural and hence concrete, is in accord with the rough consensus of
current analytic philosophers.
My general argument in this section is that if we survey the usual
metaethical suspects in light of the dual requirements set by justi¬catory
reason giving and our concrete world, we are quickly forced to the novel
metaethical position that supports my variant of hedonism, whether we
like hedonism or not. The ordinary alternatives cannot deliver the nec-
essary direct vindication of normative claims about basic value given the
concrete nature of our world. But to see this, we will need to consider
each class of familiar alternatives.
First candidate: non-naturalism.2 This is the view that there are nor-
mative properties of things that yet are not constituted by the basic natural
properties and relations of things in the world.3 Non-natural properties
of this type might provide a direct vindication of normative claims, but
they are of course excluded by our presumption that the world is con-
crete. Any two relevant possible worlds that are qualitatively identical in
regard to natural properties are qualitatively identical tout court. A robustly
concrete sense of reality demands no less.
But it is also instructive to consider a second negative point, which is
more closely analogous to forms of our abstract negative argument that
will be deployed in the rest of this section. Despite his paradigmatic non-
naturalism, G. E. Moore believed that the moral supervened on the natural,4
that two things could not differ in goodness unless they differed in natural

2 Moore (1903).
3 If we include as a “thing in the world” the world itself.
4 Though he didn™t use the term.

properties. And this is important if non-naturalism is to seem at all plau-
sible. Certainly our normative practice presumes that there are criteria
for goodness in natural properties. Any practice that lacked this feature
would seem terribly arbitrary in its normative deliverances, probably even
to those who accept that we have a bizarre and implausible faculty of dis-
cerning bizarre and unnatural moral goodness. Under such conditions,
someone could do concretely exactly what someone acting rightly did,
and yet be correctly discerned by the moral faculty to be doing wrong.
Supervenience of the normative on the natural is a deeply indefeasible
aspect of our normative practice. A moral practice without it is too radi-
cally unlike our own to be justi¬catory reason giving in anything like the
full sense. Without supervenience, non-naturalism is absurd. But my point
is that if non-naturalism is true, supervenience is not plausibly explicable.
When constitution links properties “ for instance, when being H-O-H
constitutes being water “ then it is plausible that being water supervenes
on being H-O-H. But non-naturalism denies the constitution of the nor-
mative by the natural. Nor can causation plausibly link the natural and
the normative properties of things, given our successful consensus view
about the kinds of causal laws that govern our world. Non-naturalists may
propose sui generis synthetic a priori and necessary connections between
properties not linked by causation or constitution, but no one believes
these at all plausible but themselves. If we exclude causation and consti-
tution (and presume that identity is a degenerate form of constitution),
then the supervenience of non-natural normative on natural properties is
not plausibly explicable. And it is required for any justi¬catory reason giv-
ing worthy of the name. Even Moore and other non-naturalists, by their
insistence on the supervenience of the normative on the natural, in fact
admit this. The truly characteristic form of non-naturalism, which denies
that supervenience, is too bizarre to take seriously. But such a bizarre and
implausible situation is what we should expect if normative properties are
in fact distinct in the manner that non-naturalists propose.
Second candidate: noncognitivism. Noncognitivism holds, roughly,
that normative sentences do not function in the manner of ordinary
declarative sentences, to assert facts about the world, but rather to express
attitudes or prescribe actions. Noncognitivism is largely motivated by
a robust sense of concrete reality, so it has no problem with our sec-
ond presumption. But such views also have a notorious dif¬culty that is
immediately relevant to our concerns about justi¬catory reason giving.
For instance, according to both Ayer and Stevenson, even when all
disagreements in genuine belief about the world are removed, there may

well remain signi¬cant disagreements in attitude or emotion that under-
lie the acceptance of apparently contradictory normative utterances and
principles, and such that one such utterance or principle will be no more
justi¬ed or justi¬able than the other.5 Hence these paradigmatic noncog-
nitivisms cannot underwrite the central nature of our normative discourse
as justi¬catory reason giving. No deep and asymmetrical legitimation
of basic normative principles is possible according to these views. Later
noncognitivists work hard to develop other resources to better capture the
role of reason giving in our normative discussion.6 But even those later
and more sophisticated forms suffer analogous problems. For instance, on
Gibbard™s view, as on any reasonably plausible noncognitivist view, we are
left with the possibility of alternative practices of norm acceptance differ-
ent from our own, the practices of mythical coherent anorexics or Nazis
or actual religious fanatics, which have a sort of normative symmetry with
our own, to which ours is not objectively superior. We may choose to
shield ourselves from normative interaction with such people, to eschew
normative discussion with them so as to prevent them from in¬‚uencing
us, and to emote in various sophisticated ways against them, but there
is no fact that constitutes them as being in error. These people are not
subject to evaluation as irrational in some asymmetrical way, in some way
they cannot with equal and symmetric right apply against us.7 Hence the
essential justi¬catory reason giving nature of our normative discussion is
once again lost.
There are other sorts of metaethical projects that, while not classically
noncognitivist, bear crucial similarities to noncognitivism, and that suffer
what are from our perspective similar problems. For instance, Bernard
Williams held that we cannot hope to address a justi¬cation of an intu-
itively ethical life something like our own to those who do not share
more or less our dispositions and desires, to those outside our own tra-
dition.8 But this metaethical view is problematic for the reasons we have
just traced.9 It cannot provide the necessary asymmetrical vindication
required for the existence of genuine justi¬catory reason giving. There are
also “practical-reasoning theories” and “constructivist” accounts, which

5 Ayer (1936: Chapter 6); Stevenson (1937, 1944).
6 Hare (1952, 1963); Blackburn (1971, 1984, 1985, 1988); Gibbard (1990b, 1992). But see
also Geach (1958, 1965) and van Roojen (1996).
7 Hare (1952, 1963) does not constitute an exception to this claim, because he admits the
possibility of fanatics, though his later work improbably denies this.
8 Williams (1985). Alasdair MacIntyre is also in this tradition.
9 For elaboration of this point, see Mendola (1990a).

are popular competitors of traditional metaethical cognitivism.10 Such
accounts hold that proper normative judgments are those that are the
result of proper practical reasoning,11 or of some sort of proper construc-
tion process analogous to Rawls™s use of the Original Position as a choice
procedure for principles of social justice.12 Such accounts can take two
possible forms, one in which the constraints on proper reasoning or con-
struction are capable of asymmetrical and direct vindication, and another
in which they are not. I have nothing to say against the ¬rst, though
the vindication of the various constraints evidently then collapses into
the other metaethical alternatives. Indeed, I will deploy a variant of this
form in the next chapter. But famous extant forms deploy constraints that
are not capable of asymmetrical and direct vindication. This explains, for
instance, Rawls™s famous retreat in the Dewey Lectures and in Political
Liberalism from any real attempt to provide one.13 These forms fall prey to
our now-familiar argument against views analogous to noncognitivism.
Since noncognitivism and its analogues and also non-naturalism are
inadequate, we seem forced to naturalist cognitivism. But there is a variety
of relevant types. First, there are relativist forms.14 But naturalist relativisms
either fail to capture the con¬‚ict among apparently contradictory norma-
tive practices that must be preserved as a precondition of the legitimation
of one rather than another, or, even if they can preserve such con¬‚icts
of truth, fail to provide the resources for an adequate legitimation of just
one of the alternative practices. In either case, naturalist relativism does
not provide what the crucial justi¬catory reason-giving nature of our
normative practice requires, the necessary objective asymmetry. So we
must turn to substantially nonrelativist versions of cognitivism.15 There
are three relevant forms. Two are familiar, but fail. The third is the new
alternative sketched in section II.
First, there is the traditional form of naturalist cognitivism that
involves analytic reductions of normative properties to non-normative
concrete base properties. Possession of particular non-normative concrete
base properties is supposed to entail possession of particular normative

10 Darwall, Gibbard, and Railton (1992).
11 Baier (1978a, 1978b); Par¬t (1984: 1“114); Gauthier (1986). But see Mendola (1986).
12 Rawls (1971, 1980, 1985).
13 Rawls (1980, 1993).
14 Harman (1975).
15 Most normative theories imply a limited kind of relativism. But what is crucial is that there
be some basic level of (nonrelative) normative property that supervenes (in some nonrelative
way) on some particular natural (and presumably general) properties.

properties. But it would seem that Moore™s familiar open-question argu-
ment disables such accounts. Any non-normative base properties that ana-
lytically entailed the presence of a normative property could not coher-
ently be said not to involve such a property, and yet it seems that there are
equally coherent and yet distinct normative practices that differ regarding
what non-normative base is suf¬cient for the good or the right. So the
¬rst form of naturalist cognitivism seems inadequate, and for reasons that
re¬‚ect our abstract negative argument.
The second familiar form holds that while normative properties aren™t
themselves basic natural properties or relations, they are constituted by
them, and yet in a way that doesn™t require analytic reductions. This form
is perhaps paradigmatically represented by Sturgeon, Brink, and Boyd.16
Discussion of this “constitutive naturalism” requires some background.
Traditional analytic reductive naturalism presumed a traditional iden-
ti¬cation of the analytic and a priori and necessary. It is this that makes
it apparently vulnerable to the open-question argument. But we have
learned a new orthodoxy. Kripke has convinced us that the identity
of water and H-O-H is necessary but a posteriori.17 It is now gener-
ally believed that theoretical identities are necessary and yet cannot be
evaluated on a priori grounds. This seems to undercut the Moorean
worry that if normative property X is to be identical to or constituted
by non-normative properties Y, then that can be evaluated on a priori
grounds such as those deployed in the open-question argument. It is this
space opened by our recent orthodoxy that is occupied by constitutive
But even if we accept this general picture, it is important that the
constituted normative property be at once recognizably natural and rec-
ognizably normative, that we capture the crucial normativity of the moral
in a way that can properly underwrite objective asymmetries between dif-
ferent practices of normative evaluation. The constituted properties must
be recognizably justi¬catory, and not merely motivational or merely such
as to perform the causal work of the normative. The detailed constitutive
naturalist proposals seem to me insuf¬ciently convincing in this respect.18
Hence the clearly normative and natural properties on which I will shortly

16 Sturgeon (1985); Boyd (1988); Brink (1989).
17 Kripke (1980).
18 The charge about normativity can be found in Sayre-McCord (1988). For apparent asym-
metries between standard theoretical identities and the case at hand, rooted in normativity,
see Horgan and Timmons (1990“91, 1992).

focus would serve, I believe, as the best candidates for such a treatment
even if the semantic and metaphysical conception that undergirds consti-
tutive naturalism were true.
But the central motivation for my proposal lies elsewhere. The situ-
ation is more dif¬cult for naturalist cognitivism than these elements of
recent orthodoxy suggest. The semantic and metaphysical conception
that undergirds constitutive naturalism is false, and hence so is the view
itself. Or at the very least that orthodoxy, rooted in a particular and
controversial interpretation of Kripke, is breaking down. I believe justly
so. There is wide interest in the development of what has been called
“two-dimensionalism” as an alternative to the formerly standard treat-
ment of Kripke-style cases.19 This view,20 which has developed a gath-
ering con¬‚uence of support that is almost a trend, underlies the new
metaethical alternative I will propose, though I believe that the objective
hedonism proposed in the next section is also the most attractive way to
deliver natural normativity even if constitutive naturalism is true.
Two-dimensionalism in its paradigmatic forms suggests that statements
like “water is identical to H-O-H” express something that can be factored
into two parts, an a posteriori and contingent claim like “H-O-H is the
locally dominant watery stuff” and an a priori claim like “Water is the
locally dominant watery stuff.” In the strict sense, there are no necessary
a posteriori propositions involved.21 Water, if it exists, ¬lls the watery role
in our world, and that is a truth of armchair analysis, perhaps contingent
but still a priori. But it is contingent and a posteriori that H-O-H, in
particular, happens to ¬ll that role.
Nevertheless, at least according to some partisans of two-dimen-
sionalism, including David Lewis, Frank Jackson, and David Chalmers, if
H-O-H were present in a possible world then it would of necessity be
suitably watery, and this can be known a priori.22 It is this point that is
of greatest signi¬cance for us. For the effect of this semantical and meta-
physical conception is to undergird the open-question argument against
constitutive naturalism. If non-normative natural properties can suf¬ce

19 Stalnaker (1978); Davies and Humberstone (1980); Horgan (1984); D. Lewis (1994);
Chalmers (1996); Mendola (1997: 108“113); Jackson (1998).
20 Or rather something closely analogous that is sketched in Mendola (1997: 108“113).
21 If “water” is a rigid designator, that is only because it means something like the rigidi¬ed
description “the actual locally dominant watery stuff”, or more precisely dthat [locally
dominant watery stuff], where dthat [x] is read so as to incorporate the description in
brackets in its semantic value.
22 For critical discussion, see Byrne (1999) and Block and Stalnaker (1999).

to constitute a normative property, than that must be analytic, more or
less. In light of the asymmetrical vindication of normative claims that
justi¬catory reason giving requires, basic natural properties must imply
analytically the crucial normativity of the ethical. But if all basic natural
properties are non-normative, then they do not.
Ironically, the two applications of two-dimensionalism to ethics that
have been developed by well-known partisans, by Jackson and Lewis, are
reminiscent in various ways of proposals by constitutive naturalists and
even old analytic naturalists, and hence apparently fail in familiar ways.23
It seems to me that we are instead forced by two-dimensionalism and
its ilk to a treatment of the ethical case that is analogous to Chalmer™s
dualist treatment of phenomenal experience “ of qualia “ in light of two-
dimensionalism. The normativity of the ethical is such a strange feature
of things that it is reminiscent of the oddity of physically unconstituted
phenomenal qualia. The normativity of concrete properties can be deliv-
ered consistently with the strictures of two-dimensionalism and naturalism
only if there are unconstituted natural properties that are themselves yet
That indeed is my position.24 It is a form of non-constitutive natu-
ralist cognitivism, which identi¬es some normative properties with some
basic natural properties. Such a position has no more problem with the
open-question argument than non-naturalism, and it likewise delivers the
necessary asymmetrical vindication of normative claims. And yet it assures
the (degenerate) supervenience of the normative on the natural, and avoids
non-natural metaphysical extravagance. And it doesn™t rest on shaky and
controversial necessary a posteriori propositions. And I stress again that it
helpfully delivers the essential normativity of normative properties even if
two-dimensionalism is a confusion and constitutive naturalism is indeed
Before I sketch this view, there is another class of current metaethi-
cal competitors that are worth mentioning by way of contrast, and that

23 D. Lewis (1989). Jackson (1998: 113“162) develops one possible strategy against the open-
question argument.
24 I sometimes think that there is a complication: While two-dimensionalism or something
closely analogous provides the correct analysis of the identity of water and H-O-H, still
it may be that there are genuine a posteriori necessities linking our physical states to our
phenomenal experience. Since the natural normative properties that I will eventually note
here involve our experience of pain, that may seem to provide room for the truth of
constitutive naturalism. But it is the experience of pain and not painfulness itself that might
be reasonably thought to be constituted by the physical. And this would not be suf¬cient
to deliver natural goodness. We will return to this point in section V.

also fall into the crack between traditional cognitivist and noncognitivist
accounts.25 These accounts are the so-called sensibility theories, predom-
inantly those of Wiggins and McDowell.26 Such accounts hold that nor-
mative properties are akin to secondary properties like color as conceived
by Boyle, properties that are present in our experience of things only
because of our peculiar sensibilities. And yet, these accounts continue,
such properties are not for that reason seriously defective. Perhaps the
property of being funny is an even more revealing analogy than color.
Those who ¬nd things funny are motivated to laugh, and hence norma-
tive properties might on this model provide reasons as motivations. Maybe
X is funny iff X is such as to appropriately cause the comic sentiment. It
is not that all people ¬nd truly funny things humorous, since not all have
a sense of humor. Nor will anything that anyone ¬nds funny be so, since
not all humor is appropriate. But people with the right comic sensibilities
will ¬nd the funny so. The humor of something will be a matter of its
noncomic properties, which engage the comic sensibilities of those with
a good sense of humor, and yet there are unlikely to be precise noncomic
criteria for being funny. The true humor of things will be present in the
experience of those with the appropriate comic sensibilities, and their
sense of humor will allow us to ¬gure out to a very rough degree what
in the world constitutes humorous things. Some sorts of humor will be
merely in the eye of those who have poor comic sensibilities, and humor
will be absent in the experience of those with no comic sensibility at all,
and yet humor might be an objective and genuine property of things.
Perhaps this is because, with McDowell, we™ve softened up our sense of
the mind-independence and objectivity of everything else, or perhaps it
is because there is a speci¬c set of natural properties in the world that
engage proper humor.
It is important, in order to understand the contrast between such views
and my proposal, to remember that there are other traditions of thought
about, for instance, phenomenal color than those on which McDowell
and Wiggins rely. Since at least Galileo, some have held that color state-
ments are ¬‚atly false “ for instance, because colors lack the kind of mind-
independence and objectivity that they seem to have in our experience.
These are views analogous to my proposal, though inverted into skepti-
cism. On my account, normative properties are like colors as conceived
by Aristotle, like concrete colors of a kind that Galileo denied.

25 Blackburn™s noncognitivist view is closely allied to these quasi-cognitivist alternatives.
26 McDowell (1978, 1979, 1981, 1985); Wiggins (1987a, 1987b, 1987c).

It is important to distinguish two theses: that while objects out in the
world lack colors, still sense data have them, and that while we have
experience of things with color, literally there is nothing, not even a
part of the mind, that has color. Even sense-data theorists agree with
Aristotelians in the relevant sense. But sensibility theorists are in a tradition
that holds a third thing, that there is in the material world itself color, as
a re¬‚ection of our sensibility (and also perhaps because that experience is
ultimately responsive to some genuine and mind-independent features of
the world).27
A form of our abstract negative argument tells against sensibility the-
ories. Even if we grant the conception of color on which McDowell
and Wiggins rely, it is undeniable that different animals and people have
different sorts of color experience. Think of the color-blind. And of
course at least true humor is there only in the experience of those with
the appropriate humor sensibility. But this invokes our familiar problem
about vindication. Clearly there can be those who lack the sentiments
that constitute our particular moral responses. There may be those who
have a kind of overarching aesthetic or religious sentiment that allows
them to see the sacred or beautiful where those with our moral sensibil-
ity see merely great injustice, and this may seem to them to trump our
evanescent moral concerns, just as we see our moral concerns to trump
theirs. They of course lack a sensibility. But so do we. We are back to
a situation much like that which the noncognitivists and Williams faced.
There are alternative sensibilities than our own, just as there are alterna-
tive practices of normative evaluation, and no means to privilege one as
correct or appropriate. But then, if the sensibility theorists are correct, we
cannot provide the ultimate legitimation that justi¬catory reason-giving
practice demands.
Nevertheless, the sensibility theories seem in one way close to the truth.
What is required if we are to make sense of the reason-giving nature of our
normative practice is something like color, but color as it was conceived
not by Boyle, but rather by his Aristotelian opponents, or even by sense
data theorists, a property really there in our experience and in the world
itself, mind-independent at least in the way that the color of sense data
would be mind-independent, a truly and fully objective component of
the world, there for all (at least on some people™s sense data) even if not
all see it, a property not dependent on variable sensibility in the way that
humorousness is. That more or less Aristotelian property would be at once

27 McDowell would probably not like the last bit.

a natural and a normative property. It would be a normative property that
was unconstituted by other natural properties, as the non-naturalists held,
while yet being itself a natural concrete property of the sort favored by
naturalist cognitivists.
In the cracks between previous cognitivist theories we can glimpse the
possibility of nonconstitutive naturalist cognitivism. It deploys no suspect
non-natural properties whose supervenience on the natural cannot be
plausibly explained, and yet it deploys natural properties that are in fact
normative, properties that don™t merely constitute the normative in the
controversial, and I believe ultimately implausible, manner proposed by
contemporary constitutive naturalists. Indeed, such properties are like
colors, though not quite as the sensibility theorists conceive them.
Nice work if you can get it, some may say. But in fact such unconsti-
tuted normative and yet natural properties are plausibly involved in our
concrete experience of pain and pleasure. That is the position that I will
occupy over the rest of this chapter. Because this cognitivism ¬ts into the
cracks between the various views that we have surveyed, it avoids their
characteristic failures to provide a proper account of justi¬catory reason
giving in our concrete world.


Justi¬catory reason giving apparently requires objective normative prop-
erties and facts, which are there in the world for all.28 And such properties
in our concrete world must plausibly be unconstituted concrete proper-
ties, or at the very least must be constituted by other natural properties
and yet still be obviously normative. Are there any such properties?
There plausibly are. Consider intense physical pain. And consider it
in this light: Assume that there are phenomenal states or qualia or sense
data or raw feels that are a part of it. And further assume, in accord with
one controversial but live current contender, that someone™s phenomenal
experience is not constituted solely by the physical facts about them, by
the machinations of their molecules or those of their environment.29 The
phenomenal nastiness of pain is still a natural property. But, I claim, it is
a normative property.
Let me go slowly at this crucial point. Consider violent murder “
say, of the red knight by the gold knight, on the point of a lance in

28 This section and parts of the next are based on Mendola (1990b).
29 See, for instance, Chalmers (1996).

his golden arms. Perhaps some murders are, in all their context and
with all their consequences, for the best. Perhaps in this case the gold
knight was not unreasonable. But most ethicists would like to believe
that there is something about at least most violent murders, considered
at least in isolation from their context and consequences, that is bad. Is
there any suitably objective intrinsic disvalue present in such cases? Since
it seems likely that the worst part of murder is that it hurts the mur-
dered, focus attention on intrinsically bad things that might be thought
to happen to the victim of such a violent murder. And put it this way: Is
there anything in what happened to the red knight that cannot be ade-
quately and completely described without admitting its disvalue? I believe
there is.
Ignore for a moment the knight™s psychological states, his pain and
dismay. Consider the nonpsychological side of what happened to him.
Some sharp metal cut some live human ¬‚esh, some blood surged out of
its customary place and, we shall assume, choked the organism. Respi-
ration ceased, and then the red knight died. This is something that we
are likely to feel squeamish about, that we may regret, but is there any-
thing in the machination of molecules that we believe this piercing and
surging and choking to be that involves disvalue? I think not. One can
certainly imagine this ¬‚eshy part of the murder described quite completely
and adequately by Martians, with the detachment of a biologist studying
yeast metabolism. The Martians might provide a complete and adequate
description of the phenomenon that admitted no normative properties
at all. Perhaps the red knight was too proud or surprised to scream, but
even had he thrashed and screamed, limbs scrambling and the emission of
screaming sounds can be adequately described in a way colorless of value.
If there is more to screaming and thrashing than this sort of movement,
it should be found in the victim™s psychology.
Let us widen our consideration of the murder to include the red knight™s
psychological states. First, he no doubt wanted not to be murdered and,
more generally, not to be cut or pierced. No doubt he thought such
things were bad. But propositional attitudes like beliefs and desires seem,
at least on the surface, not generally to require commitment to normative
properties in their adequate and complete descriptions.30 Martians could

30 Not all agree. Some hold that there are normative properties involved whenever there is a
state with meaning or content, because such states are constrained by principles of rationality
or are governed by standards of correctness. But this rests on the con¬‚ation of two senses
of “rationality” or of “correctness” noted earlier.

describe the knight™s wanting not to be murdered, as we might describe
an amoeba™s ¬‚eeing our pipette, without talking about disvalue at all. And
since they can describe his belief in witches without committing them-
selves to the existence of witches, or describe his disbelief in microbes
without denying that there are microbes, it seems they can also adequately
characterize his belief in disvalue without committing themselves to the
existence of disvalue. Our Martians are describing, not evaluating, the red
knight™s desires. And in the concrete world we presume, there is noth-
ing more to his wanting and believing than the concrete “ for instance,
behavior and neural machinations and maybe qualia. At least there are no
irreducible psychological relations to irreducible Platonic forms of evil.
That the red knight did not value the murder and that the gold knight
did implies nothing about its having objective value or disvalue.
Are there other of the red knight™s psychological states whose descrip-
tion might require commitment to disvalue? Since he died, his conscious-
ness ceased, though this too seems characterizable without commitment
to disvalue. But now consider what he experienced, what he sensed and
felt. Assume that there were phenomenal states, sense data or raw feels, that
he experienced in being murdered. Assume that, in Nagel™s phrase, it was
like something to be him then.31 And further assume, in accord with one
controversial but live current contender, that his phenomenal experience
is not constituted by the physical facts about him, by the machinations of
his molecules or those of his environment.
Is the red knight™s phenomenal experience in being murdered charac-
terizable by the Martians without commitment to value or disvalue?
No doubt we have already considered a part of his agony, his ¬‚inch-
ing or running or screaming behavior and his negative conative attitude
toward being stuck and killed. Such things can, it seems, be described
without commitment to disvalue. But such things are not all there is to
at least some pain. Consider the phenomenal component of agony, the
sensations and feelings and raw experience that accompany being stabbed.
No one, not even a Martian, I claim, could give a complete and adequate
description of the red knight™s psychological states without characterizing
those raw feels, which we are presuming for the moment to be physically
unconstituted and hence to involve the presentation to him of charac-
teristic natural phenomenal properties not constituted by other natural

31 Nagel (1979).

We may be able to imagine a very strange man who, when in a phe-
nomenal state like that of the red knight, whistled and snapped his ¬ngers
and did math sums.32 Perhaps we can imagine someone being stabbed and
writhing like him while having a phenomenal state like that of intense
sexual pleasure centered on the wound. But neither of these are our pro-
tagonist. There are phenomenal and not merely behavioral differences,
differences in raw experience and not merely in propositional attitudes,
between those in bliss and those in agony. Any complete and adequate
characterization of the red knight™s murder must do these justice.
Some will hold that complete and adequate descriptions of situations
involving phenomenal experience do not necessarily involve commit-
ment to phenomenal properties. Perhaps phenomenal states are consti-
tuted in such a way that someone™s physical characterization entails the
presence of the relevant phenomenal states “ for instance, the experi-
encing of the phenomenal properties “ without anything in fact having
those phenomenal properties. Perhaps feeling a phenomenal property is
like believing in witches. Or perhaps phenomenal properties are real,
but constituted by other natural properties. Still, one live contemporary
alternative, consistent with our metaphysical presumptions so far, is that
adequate characterizations of situations involving phenomenal states must
involve commitment to various irreducible and otherwise unconstituted
phenomenal properties.
It is not implausible to claim that unconstituted phenomenal properties
must be captured by any complete and adequate description of the red
knight™s murder. And indeed, the case for concrete pleasure and pain
involving certain concrete phenomenal properties is strong even if the
metaphysical picture that undergirds constitutive naturalism is correct and
such properties are constituted by physical properties. But let me be very
clear about the asymmetry I am claiming here. We can desire to be a
witch when there are no witches, but if we have phenomenal experience
of yellow, then something is yellow. In fact, I also claim, there is something
irreducibly and unconstitutedly yellow, either something really out in the
world, or at least a sense datum under our hats.33
While this view is of course controversial in our historical situation,
in which many hold that sensory experience is as of yellow though there
is nothing in the world that is so, not even a sense datum, or at the very

32 D. Lewis (1980).
33 If yellow is out in the world, then perhaps a hallucination of yellow, like one of pain, requires
no such postulation. We will return to this.

least that the yellow we experience is a natural property constituted by
physical properties like a certain range of surface spectral re¬‚ectance, still
the view in question is, as I™ve said, one live competitor. Indeed, it is often
motivated by arguments that are structurally similar to the open-question
argument: You look at a gold bar and have a certain sort of phenomenal
experience. But it seems to some that it might well be an open question
whether your physical twin in a physically identical environment has the
same phenomenal experience, or any at all. He might be a zombie or a
qualia invert. And the openness of that question suggests to some that the
physical cannot constitute your phenomenal experience.
At least such qualia dualism is relatively concrete and robust. Even
though it involves physically unconstituted qualia, it involves nothing that
is non-natural in Moore™s sense. It is at least concretely comprehensible.
And that gives it a great advantage over alternative forms of normative
That is my main point, that this so far familiar qualia dualism unex-
pectedly but very plausibly implies a form of normative realism. Painfulness
“ or, more accurately, the phenomenal property present in certain sorts
of extreme and paradigmatic physical pain “ is a kind of disvalue. That is
my new idea.34 The phenomenal difference between those in bliss and
those in agony includes a difference in a sort of felt phenomenal value.
The phenomenal difference between pain and pleasure seems (at least in
part and sometimes) to be that the phenomenal component of the former
is nastier, intrinsically worse than that of the second.
The red knight was stabbed to death. Just as no one can adequately
describe what it was like to be him without capturing his sensation of his
red and ¬‚owing blood and hence the property of phenomenal redness,
so no one can describe what it was like to be him without capturing the
nasty sensations he felt and hence the property of phenomenal nastiness
or disvalue. And no one can understand what his phenomenal state was
without knowing that it was intrinsically bad, worse than pleasure. No
one, not even a Martian, can give a complete and adequate characteriza-
tion of the red knight™s murder while ignoring the phenomenal state that
was a part of that situation. And no one, not even a Martian, can give a
complete and adequate characterization of that phenomenal state with-
out capturing its nastiness, its intrinsic disvalue. The red knight™s murder
possessed what we might call objective intrinsic disvalue.

34 Not very new. See the discussion of antecedents in Mendola (1990b), and especially Lewis

If someone feels bad, then there is something bad, at least in cases of
extreme physical pain. My further claim, to which constitutive natural-
ists dissent, is that this involves unconstituted but natural disvalue. Like
other phenomenal properties, the disvalue present in agony is uncon-
stituted by physical properties, though it is itself concrete and natural.
It is just like phenomenal yellow. The objective but unconstituted phe-
nomenal component of agony involves a correspondingly objective and
unconstituted phenomenal property that is usually present in cases of at
least extreme physical pain, a painfulness or “unpleasant hedonic tone”,
as it was once called.35 And such objective phenomenal properties are, at
least in part, a sort of intrinsic disvalue or badness. Something analogous
is true of certain paradigmatic physical pleasures. They involve objective
intrinsic value.
All of my claims here are controversial. Some deny that there are phe-
nomenal experiences. Some deny that phenomenal experience cannot
be constituted solely out of other natural properties. That the experience
of pain at least often involves a phenomenal “hedonic tone” was classi-
cally held by Broad, Duncker, and Edwards.36 But there were also classic
objections to pleasure™s being like a sensation or feeling.37 And I sus-
pect that the claim that vivid physical painfulness is a normative property
will not be attractive even to some who accept unconstituted hedonic
But while all of these claims are somewhat controversial, they all have
at least some plausibility. Together, these claims represent a plausible con-
crete case for an objective and suitably normative fact, which is some-
thing we have regrettably lacked. If the most plausible forms of two-
dimensionalism38 are correct, then there is little hope for another. And
the phenomenal disvalue of agony is a lonely plausible case for a normative
but natural fact even if constitutive naturalism is true, and the phenomenal
and normative are constituted by the physical.
While it turns out that the kind of unconstituted natural normative
property in our world is a certain kind of value and disvalue, which I call
objective intrinsic value or goodness, this claim should not be misunder-
stood. I don™t think this exhausts the meaning of “good” in English, nor
even any intuitive notion of intrinsic good. It is not possible, I think, to

35 Broad (1930: 229“233).
36 Duncker (1940“41); Edwards (1979: 46“47).
37 Ryle (1954: 54“67); Broad (1930: 231); Perry (1967).
38 Or analogous views like my own.

“meaningfully” question whether it is good, but it is possible to question
whether it is all the good. In fact, it isn™t. Nor do I think that some sort
of dominance of value words over other normative terms “ say, those that
deontologists prefer “ is available by meaning analysis. It is just that the
kind of normative property that there happens to be in our world ¬ts
certain value terms. Perhaps it ¬ts certain deontic terms also.
Still, even if you grant the general plausibility of my key metaphysical
presumptions here, there remain natural objections to the view I have
Objection One: Pain is supposed here to be bad tout court, from the
point of view of the universe. But isn™t my experience of my pain bad for
me in particular?
Reply: Just as in the case of phenomenal yellow, such an analysis is too
complex for the content of a simple phenomenal state. My physical pain
presents itself as bad, not speci¬cally as bad for me or bad for all people,
but bad. Perhaps one way to see this is to consider on the alternative
supposition exactly to whom my pain presents itself as bad. To me at this
moment, to me as a continuing entity, or to my genetic line?
Objection Two: But if the disvalue we ¬nd in pain is disvalue from the
point of view of the universe in this way, in what sense does this support
a hedonistic conception of our well-being? How is our pain bad for us?
Reply: The value I am speaking of need not be something that specif-
ically constitutes individual well-being. So I am defending hedonism in
what is perhaps only an unusual and attenuated sense. Recall the ¬rst
section of the last chapter.
Objection Three: It is implausible that all pains involve this sort of
objective nastiness. Consider painful anxiety. It has little in common phe-
nomenally with the physical pain of being stabbed.
Reply: I do not claim that all pains involve this negative hedonic tone.
I have only so far claimed that a few paradigmatic and vivid concrete
physical pains involve it.
Objection Four: If the painfulness of paradigmatic physical pains is
a kind of badness, if to judge that a pain of that sort is getting worse
is to make a value judgment, then hedonism is in fact endangered. We
cannot then argue that such a change should be avoided because it involves
more pain. There isn™t the space between the painfulness and the to-be-
avoidedness that that “because” invokes.
Reply: We can argue that such a change should be, in itself and irrespec-
tive of context, avoided because it would involve more painful hedonic
tone. But the “because” links two properties that are not as ¬rmly distinct

as in some more familiar cases. There are other cases like this. There is
water in the lake because it is ¬lled with H-O-H.
Objection Five: It is a central claim here that some natural facts about
pain are irreducibly normative, in that they provide justifying reasons
without being anything other than natural facts themselves. This view is
contrasted with the view of non-naturalists, on the one hand, who hold
that there are normative facts that cannot be reduced to natural facts, and
the view of noncognitivists, on the other hand, who hold that there are
no normative facts because the only facts are natural ones. But the view
that there is a middle position is problematic. Let us grant that natural
facts, such facts as that someone is in pain, often function as reasons, and
so have normative force. No non-naturalist need deny this. What he or
she denies is that the true proposition that a natural fact is a reason is itself
a natural fact. And noncognitivists agree with the non-naturalist about
that. This is because Moore™s open-question argument is taken by both
to show that it is a mistake to think that the proposition that something
is a reason is itself a natural fact, or that the predicate “is a reason” refers
to a natural property of natural objects. If this is correct, however, it
seems to leave only two positions possible. The ¬rst is non-naturalism,
according to which, while the predicate “is a reason” does not refer to a
natural property, it can still be predicated truly or falsely of natural facts.
The second is noncognitivism, according to which, since the predicate
“is a reason” does not refer to any natural property, it cannot, strictly, be
predicated truly or falsely of anything at all. There is no middle ground to
occupy, unless, contrary to what Moore™s argument suggests, “is a reason”
does refer to a particular natural property. In light of this, the claim that
there are irreducibly normative natural facts is either trivial or obscure.
Either it means only that some natural facts, such as facts about pain, are
reasons, which no one denies. Or it means that some natural facts are
themselves facts about what natural facts are reasons, which has not been
Reply: My claim is that there is in fact a third position that we have
generally overlooked, that there are properties that are at once paradig-
matically natural and paradigmatically normative, as those words are cus-
tomarily used. And this possibility cannot be undercut by Moore™s open-
question argument, even though the two-dimensionalist analysis of the
identity of water and H-O-H is correct.
To see this properly, it is important to ¬rst recall that I do not claim of
our word “value”, no more than of our word “reason”, that its meaning
is such that it applies by de¬nition to just and only the objective value

discerned here. Rather, whatever determinate meanings such words have
speci¬es conditions that are met by the objective property I™ve noted.
The property I™ve suggested is, as it were, one kind of possible value or
reason, the objective and unconstituted kind there happens to be in the
world. So I agree in a sense that “is a reason” does not refer to a natural
property, or indeed to any single basic property, non-natural or natural.
My argument for the existence of the objective value we have noted isn™t
really a semantic argument at all.
Can Moore™s argument show that there isn™t this sort of objective value?
Of course it can in some weak sense be meaningfully asked whether
hedonic value of the sort I note is bad, just as it can in the same sense be
meaningfully asked if bachelors are unmarried. But that is presumably not
the kind of meaningfulness that Moore™s question properly isolates. My
central claim is that to fail to note the badness of phenomenal painfulness
is to miss something that is necessary to a complete description of what
some pains involve. So in the relevant sense I claim to have shown that
Moore™s question is not open for this one sort of case. Still, there might be
other kinds of things properly called value for which the question is open.
The objection is phrased in terms of a paradigmatically deontological
property rather than in terms of value. We will turn to the issue of the exact
nature of phenomenal value and disvalue in the next section, but it is, I
think, of a nature that might also be captured by talk of reasons in a suitably
broad sense rather than of value. The central point for now is the general
nature of the phenomenal property itself and its normativity. The detailed
nature of the property will dictate how it is best characterized in detail,
and indeed it may have more than one proper normative characteriza-
tion, drawn from different traditions of normative vocabulary. Perhaps we
might call the natural normative property of pain “to-be-eliminatedness”.
Now focus on the contrast between my view and non-naturalism
and noncognitivism. For a non-naturalist or noncognitivist, the claim
that the natural phenomenal painfulness we have discerned is intrin-
sically bad would have to be open in the Moorean sense. That is the
difference between my view and an analogous sort of non-naturalism or
noncognitivism. It is, I claim, a difference that is to my advantage.
Let me put the elements of this long reply together explicitly: In fact
there is middle ground, which closes Moore™s open question for hedonic
value despite the fact that “is a reason” is not itself a single natural property.
Objection Six: The suggestion is that negative hedonic tone is disvalue,
even though “disvalue” doesn™t mean the same as “negative hedonic
tone”. Why isn™t this proposal vulnerable to the two-dimensionalist

objection that was proposed against constitutive naturalism in section I?
That objection denies the possibility of genuine necessary a posteriori
propositions “ for instance, property identities.
Reply: Because there is an a priori entailment linking negative
hedonic tone with disvalue. But notice that the entailment runs only in
one direction. No identity of meaning is required.
Objection Seven: There are various rival accounts of pain. For
instance, there is the view that there is no phenomenal experience
characteristic of pain, but merely that aversion is. Second, there is the
view that there is something phenomenal characteristic of pain but that
it is not badness. Third, there is the view that there are different kinds of


. 5
( 10)