Why Idealize PDF

Why Idealize?
David Enoch
I. IDEALIZATION
Roderick Firth claims that ethical truth is not determined or constituted
by what we—nonideal as we are—want or choose, but by what ideal
observers (that is, observers who are omniscient, omnipercipient, dis-
interested, dispassionate, consistent, and normal in other respects)
would want or choose, or perhaps by some other of their responses.1
This, he argues, is just what the truth of ethical claims such as “x is
right” consists in; something’s being right just is its being such as to be
chosen by the (or an) ideal observer. Peter Railton argues that what is
good for a person is not necessarily what he wants, but what his ideal
self (one who is “fully and vividly informed about himself and his cir-
cumstances, and entirely free of cognitive error or lapses of instrumental
rationality”) would want for him (nonideal as he is) to want or to
pursue.2 And Railton seems to believe that this is what it is for something
to be good for a person;3 it is to be such that one’s ideal advisor would
choose it for one. Bernard Williams believes that we have reason to do
not necessarily what we desire, but (perhaps roughly speaking) what we
would desire under suitable conditions (after rational deliberation,
which includes the elimination of false beliefs, the introduction of some
true ones, and some exercises of imagination).4 Furthermore, Williams
* For helpful conversations and comments on earlier versions of this article, I thank
Stephanie Beardsman, Hagit Benbaji, Bruce Brower, Alon Harel, Andrei Marmor, Derek
Parfit, Josh Schechter, Mark Schroeder, and anonymous referees and editors for Ethics. I
am especially grateful to David Copp, whose many comments and suggestions regarding
evolving versions of this article improved it considerably. I presented an earlier version of
this article at the colloquium of the Ben Gurion University Philosophy Department, and
I thank the audience there for the helpful discussion that followed.
1. Roderick Firth, “Ethical Absolutism and the Ideal Observer,” Philosophy and Phe-
nomenological Research 12 (1952): 317–45.
2. Peter Railton, “Facts and Values,” Philosophical Topics 14 (1986): 5–31, 16.
3. But see n. 17 below.
4. Bernard Williams, “Internal and External Reasons,” in his Moral Luck (Cambridge:
Cambridge University Press, 1981), 101–13.
Ethics 115 (July 2005): 759–787

䉷 2005 by The University of Chicago. All rights reserved. 0014-1704/2005/11504-
0002$10.00
759
760 Ethics July 2005
claims that this is what it is to have a reason to F (an internal reason,

that is, but this is the only kind Williams believes in); it is to have a
“sound deliberative route” from one’s actual motivational set to the
desire to F.5 David Lewis thinks that values are not necessarily things
we value, but rather things we would value under ideal conditions (con-
ditions of full imaginative acquaintance).6 More than that, Lewis thinks
that something’s being of value just is being such that we are disposed
to value it under ideal conditions.
All these views—and many, many others—essentially involve some
idealization.7 They tie the relevant normative fact (about the right, one’s
good, one’s reasons, or values) to hypothetical, idealized responses, or
to the responses of hypothetical, idealized agents or thinkers, or to the
responses of actual agents after going through some hypothetical, ide-
alizing procedure. As is clear from the examples above, idealizing views
differ among themselves along several dimensions: different idealizers
choose different normative terms or concepts as their analysandum,
focus on different responses, and treat different conditions as ideal.
Furthermore, idealizers often differ in their philosophical motivation
and in how they view their philosophical project (a point I return to
below). But despite the great variety of such views,8 and with certain
qualifications shortly to emerge, I think that they can all be shown to
fall prey to a single objection.
In what follows I argue that idealizing views are not likely to be able
to motivate the very idealization they employ. If so, such idealization is
likely to remain—given the philosophical concerns underlying idealizing
views—objectionably ad hoc. And this, I conclude, makes response-
dependence theories9 much less plausible, for they must either embrace
the ad-hoc-ness of the idealization or else settle for actual, nonidealized
responses. If actual-response-dependence views are as implausible as they
seem, perhaps, then, this should be taken not so much as a reason to
5. Bernard Williams, “Internal Reasons and the Obscurity of Blame,” in his Making
Sense of Humanity (Cambridge: Cambridge University Press, 1995), 35–45. Though Williams
only argues for the claim that his condition is necessary for something being a reason,
he expresses confidence that it is also sufficient. See ibid., 35–36.
6. David Lewis, “Dispositional Theories of Value,” in his Papers in Ethics and Social
Philosophy (Cambridge: Cambridge University Press, 2000), 68–94, originally published in
Proceedings of the Aristotelian Society 63, suppl. (1989): 113–37.
7. Notice that this group of views is wider than what is usually thought of as ideal
observer theories. The example of Williams’s theory of reasons should suffice to show
that.
8. It is perhaps worth mentioning already at this early stage that nothing in what
follows will depend on whether the relevant idealizing analysis is considered to be a priori
or a posteriori.
9. I take this term from Michael Johnston, “Dispositional Theories of Value,” Pro-
ceedings of the Aristotelian Society 63, suppl. (1989): 139–74.
Enoch Why Idealize? 761
idealize, but rather as a reason to reject response-dependence theories

of normativity altogether.
Evidently, then, I will not here argue against the relevant exten-
sional equivalences defended by different idealizers. For all I am about
to say here, it is possible that, say, nothing is of value without us being
disposed to value it under conditions of full imaginative acquaintance,
and nothing is such that we are disposed so to value it without also
being of value. Though idealizing views face considerable difficulties of
detail in achieving such extensional adequacy,10 I will for the sake of
argument assume that such difficulties can be satisfactorily dealt with.
I will argue that even assuming extensional adequacy, idealizing views
only achieve extensional adequacy at a hefty price, because the very
idealization they invoke does not sit well with the philosophical concerns
underlying the relevant response-dependence view. This means, of
course, that I have to say more about these philosophical concerns, but
it will prove convenient to postpone this discussion to a later stage.
The discussion proceeds as follows. In Section II I present what I
think is the—or at least a—natural reply to the request for a rationale
for idealization. I then proceed to argue that this natural line is one
typical idealizers cannot employ, given the reasons that drive them to
their view in the first place. Then, in Section III, I proceed to discuss two
replies to the request for a rationale for idealization that idealizers can,
and typically do, employ, but which I find inadequate. In Section IV I
conjecture that there are no other ways idealizers may face the challenge
of motivating their idealization, giving some reason to believe that the
conjecture is true. In Section IV I also address a general methodological
objection to the challenge I put forward to idealizing views. In Section
V I comment on the generality of my argument. I then—in Section VI—
discuss one variation on the theme of idealized-response-dependence
views, a variation that escapes my criticism. A brief conclusion follows.
II. WHY IDEALIZE—THE NATURAL ANSWER

The challenge, then, is fairly simple. Suppose you want, for whatever
reason, to ground the reasons agents have in their motivational sets, or
values in the state of valuing, or something of that sort. Why not settle
for what I actually want, or for what I actually value? What justifies the
move to hypothetical, ideal conditions, however exactly these are
characterized?
A natural answer would be to claim that the relevantly ideal condi-
10. See, e.g., Hubin’s discussion of dispositions and deviant causal chains (Donald
Hubin, “Hypothetical Motivation,” Noûs 30 [1996]: 31–54, 36–39), though Hubin thinks
the more serious problem with such views is one not of extensional adequacy but rather
of adequacy as a philosophical analysis.
tions are the conditions needed for a reliable tracking of the relevant
facts. Suppose that you want to know the time. Looking at a watch seems
like a good idea. But, of course, looking at your watch may not be such
a good idea. This depends on whether your watch keeps reasonably ac-
curate time. What you want, then, is to have a look at a good watch. An
ideal watch would be great, but we can settle for one that is less than
ideal, so long as it is close enough. So we require, say, that the batteries
in your watch be at least almost fully charged. Or consider this: you want
to know who is taller, myself or my wife. Having a look seems like a good
idea, but of course not just any look will do. What you want is to have a
look from a proper angle, from up close, when my wife is not wearing
heels.
In these cases, some (very moderate) idealization is called for be-
cause otherwise an epistemic procedure—a way of forming beliefs—may
very well lead us astray. If the watch is not reasonably accurate, or if you
are much closer to me than to my wife, the suggested epistemic pro-
cedure will fail; it will not be a reliable indicator of the relevant fact
(the time, or my and my wife’s relevant height). And this, of course, is
one good rationale for idealization: idealization (or its approximation)
is called for whenever an actual procedure is fallible in ways (partly)
corrected for by the idealization.
Some idealizers sometimes speak as if this rationale holds in the
case of their idealization as well. Here, for instance, is Lewis:
If someone has little notion what it would be like to live as a free

spirit . . . or what a world of complete harmony and constant
agreement would be like; then whether or not he blindly values
these things must have little to do with whether or not they are
truly values. . . . If only he would think harder, and imagine vividly
and thoroughly how it would be if these putative values were real-
ised . . . that would make his valuing a more reliable indicator of
genuine value. And if he could gain the fullest possible imaginative
acquaintance that is humanly possible, then, I suggest, his valuing
would be an infallible indicator.11
Lewis here seems to suggest that just as a bad angle undermines the
reliability of having a look as a way of determining relative height, so
too lack of imaginative acquaintance undermines the reliability of the
procedure he suggests for finding out whether something is of value.
11. Lewis, “Dispositional Theories of Value,” 77 (footnote omitted and italics added).
See also Firth, “Ethical Absolutism and the Ideal Observer,” 326.
But, as is often noted, this just cannot be right.12 Lewis—in his more
careful moments—does not think of our valuing under ideal conditions
as an indicator (fallible or otherwise) of what is of value. He thinks of
such valuings as the truth-makers of value claims. For something to be
of value just is, so he wants us to believe, for us to be disposed to value
it under ideal conditions. Similarly, Williams does not think of our desires
after rational deliberation as (perhaps necessarily) tracking an indepen-
dent order of reasons. Rather, he thinks that having a desire to F after
rational deliberation is what having a reason to F consists in. And if what
it is for me to have a reason to F is for there to be a sound deliberative
route from my motivational set to a desire to F, then there being such
a deliberative route is not an indicator of my having a reason. (Are facts
about mean kinetic molecular energy [perhaps infallible] indicators of
facts about heat?) And similarly, of course, for other idealizing views.
But with the collapse of the reliability and indication rhetoric col-
lapses also the natural rationale for idealization. For when we think of
one thing (the watch reading) as a reliable indicator of another (the
time), we think of the latter as independent of the former. It is, say, 9:43
p.m. anyway, regardless of the reading of my watch. If my watch keeps
accurate time, it will read “9:43.” If it does not, it may not. The time,
though, does not change together with the reading of my watch. If all
goes well, in other words, my watch’s reading tracks the time-facts that
are independent of it. Similarly, the relative height of me and my wife
does not change along with the angle from which you view us. If, say,
she is slightly taller than me, she is taller than me independently of
how—and even whether—you choose to look at us. When all goes well,
your visual impression tracks the truth about relative height that is in-
dependent of it.
My point is not the (possible) failure of extensional equivalence,
which I am here granting for the sake of argument. Rather, my point is
best thought of in terms of priority or dependence relations and best
brought out by thinking in terms of the Euthyphro Contrast:13 a good
12. For similar criticisms of Lewis, see Robert Noggle, “Integrity, the Self, and Desire-
Based Accounts of the Good,” Philosophical Studies 96 (1999): 303–31, 309; and Andrei
Marmor, Positive Law and Objective Values (Oxford: Oxford University Press, 2001), 180.
And see also a similar point made by Sosa as a criticism of Wiggins (David Sosa, “Pathetic
Ethics,” in Objectivity in Law and Morals, ed. Brian Leiter, 287–329 [Cambridge: Cambridge
University Press, 2001], 306). For the general point (that idealization of the type discussed
cannot be grounded in epistemological considerations), see also Arthur Ripstein, “Pref-
erence,” in Value, Welfare and Morality, ed. Richard Frey and Christopher Morris, 93–111
(Cambridge: Cambridge University Press, 1993), 104–5; Hubin, “Hypothetical Motivation,”
38; and Hallvard Lillehammer, “Revisionary Dispositionalism and Practical Reason,” Journal
of Ethics 4 (2000): 173–190, 177.
13. See, e.g., Johnston, “Dispositional Theories of Value,” 172–73.
watch reads “9:43” because it is 9:43. It is not the case that the time is
9:43 because of the watch’s reading. In proper conditions you see that,
say, my wife is taller than me because she is, not the other way around.
And this direction-of-because issue captures an intuitive sense of priority
or dependence:14 even given extensional equivalence, my watch’s read-
ing depends on the time, but the time is independent of my watch’s
reading.
This is why there is, in the time and relative-height examples, room
for some idealization. The reading of the watch tracks the time—which
is independent of it—only when all goes well, the perceptual impression
tracks relative height—which is independent of this perception—only
when all goes well. So there is reason to make sure—by idealizing—that
all does go well. But had we taken the other Euthyphronic alternative
regarding these matters things would have been very different. Had the
time depended on the reading of my watch, had the reading of my
watch made certain time-facts true, there would have been no reason
(not this reason, anyway) to “idealize” my watch and see to it that the
batteries are fully charged. In such a case, whatever the reading would
be, that would be the right reading, because that this is the reading
would make it right.
The natural rationale for idealization, the one exemplified by the
time and relative-height examples, thus only applies to cases where the
relevant procedure or response is thought of as tracking a truth inde-
pendent of it. This does not necessarily rule out extensional equivalences
between normative truths and our relevant responses. One may, for
instance, hold a view that is an instance of “tracking internalism,”15
according to which, necessarily, one cannot have a (normative) reason
without being motivated accordingly, not because motivations are part
and parcel of (normative) reasons, but rather because our motivations
necessarily track the independent truths about (normative) reasons.16
But typical idealizers do not think of their view in this way; they do not
think of the relevant response as (necessarily) tracking an independent
14. Which is not to say that there aren’t any problems in understanding or giving a
fully explicit philosophical account of this intuitive idea, and indeed of the philosophical
problems surrounding the Euthyphro Contrast. See Crispin Wright, Truth and Objectivity
(Cambridge, MA: Harvard University Press, 1992), 139.
15. See David Sobel, “Explanation, Internalism, and Reasons for Action,” Social Phi-
losophy and Policy 18 (2001): 218–35, 233, and “Subjective Accounts of Reasons for Action,”
Ethics 111 (2001): 461–92, 473. For a related discussion, see Connie Rosati, “Agency and
the Open Question Argument,” Ethics 113 (2003): 490–527, 525–26.
16. Perhaps Plato’s view of the relation between the good and our motivations can
serve as an example of tracking internalism.
order of normative facts. As emphasized above, they think of the relevant

response as constituting the relevant normative fact.17
This is no coincidence. Idealizers like to think of themselves as
naturalists of sorts, indeed often as suggesting a naturalist reduction of
the relevant normative facts. And they always want to think of themselves
as putting forward an alternative to more robustly, Euthyphronically
realist views, such as views that postulate the existence of purportedly
queer normative entities that are just as independent of us and our
responses as electrons and planets presumably are. The point I want to
make, then, is this: if this is your reason for endorsing a response-
dependence view of the normative, you cannot consistently employ the
natural rationale for idealization.18
III. WHY IDEALIZE—POSSIBLE ANSWERS

But this does not mean that you cannot consistently endorse other
rationales for your idealization. In this section I consider two suggestions
that are common—if often implicit—among idealizers.
17. At one point, Railton suggests that the truth-maker of the relevant normative
claim is not the idealized response, but rather its nonnormative reduction base. See Peter
Railton, “Moral Realism,” in Moral Discourse and Practice, ed. Stephen Darwall, Allan Gib-
bard, and Peter Railton, 137–63 (Oxford: Oxford University Press, 1997), 143, originally
published in Philosophical Review 95 (1986): 163–207. And so it may seem as if Railton can
after all give the natural answer to the why-idealize question—for it is possible that advisors
in less-than-ideal conditions fail to reliably track this reduction base. But this cannot be,
I think, Railton’s view. If the dispositional fact is not to drop out of the story as an
insignificant intermediate stage between normative facts and their physicalist reduction
base, then it must still be Railton’s view that these dispositions are not thought of as
reliably tracking the dispositions-independent normative facts, but rather as grounding
the relevant normative facts. Railton’s view must be, I think, that the fact that certain
reduction-base-facts are the ones favored by an ideal advisor is what explains why it is this
(rather than any other) reduction base that does the ultimate work here. And this is
enough to make the natural answer unavailable to Railton. For a different reading of
Railton, see David Zimmerman, “Why Richard Brandt Does Not Need Cognitive Psycho-
therapy, and Other Glad News about Idealized Preference Theories in Meta-ethics,” Journal
of Value Inquiry 37 (2003): 373–94, 375. I thank a referee for Ethics for drawing my attention
to the complication discussed in this note.
18. Because the discussion has been put in terms of dependence and priority, so-
called no-priority views such as McDowell’s and Wiggins’s escape the reasoning of this
section. See John McDowell, “Projection and Truth in Ethics,” in Darwall, Gibbard, and
Railton, Moral Discourse and Practice, 215–25; and David Wiggins, “A Sensible Subjectivism?”
in his Needs, Values, Truth, 3rd ed. (Oxford: Clarendon, 1991), 185–214, reprinted in
Darwall, Gibbard, and Railton, Moral Discourse and Practice, 227–44. Nevertheless, I think
that the challenge in the text can be modified so as to pose a threat for them as well.
But so modifying it will take me too far afield, and—given my general suspicions regarding
such views—I am happy to restrict my argument here so that it does not directly apply to
them. I thank Hagit Benbaji for many relevant discussions.
A. Extensional Adequacy
Take an idealizing view and drop the idealization. The view you are left
with will be rather obviously extensionally inadequate. Counterexam-
ples—cases where the deliverances of the theory contradict intuitive,
pretheoretic judgments that we are rather confident in even upon re-
flection—are going to be too easy to come by. If you desire to drink the
liquid in the glass in front of you believing it is gin, then—endorsing
something like Williams’s view of reasons but dropping the idealization—
you have a reason to drink what is in the glass even if in fact it is petrol
(and even if, knowing that, you would not have wanted to drink the
liquid). But (pretheoretically) this seems wrong—you do not have a rea-
son to drink the petrol that is in the glass. Similarly, if I am disposed to
value lifelong solitude, then—endorsing something like Lewis’s view of
values but dropping the idealization—lifelong solitude is of value, even
if my valuing it stems merely from my never having thought through what
it would be really like. And, of course, lifelong solitude is not of value.
False negatives are also easy to come by. I do not have in my motivational
set a desire to go to the dentist, so without the idealization Williams’s
(amended) view would have it that I have no reason to go to the dentist,
even if, had I thought it through and realized what I will feel like next
week if I don’t go to the dentist, I would have had a desire to go to the
dentist. And intuitively, of course, I have a reason to go to the dentist.
The point generalizes. No view that ties normative truths to our
actual responses is likely to achieve extensional adequacy, because some
of our responses are clearly bad, not reason supported, or not in accord
with what really is of value, or with what is really good for us. Idealization
is called for in order to save even just a possibility of extensional ade-
quacy for response-dependence views.
Why not settle, then, for the following rationale for idealization?
We want a theory that is extensionally adequate; we cannot have a re-
sponse-dependence theory that is extensionally adequate without ide-
alization; presumably, we have independent reasons to go for a response-
dependence theory, rather than, say, a more robustly realist one; and
so we should idealize.19
The problem with this line of thought is that extensional adequacy
can be had for cheap. Whenever one puts forward a theory that is found
to be extensionally inadequate, one can—in principle, at least, for in
practice it may not be an easy task—patch it up and achieve extensional
adequacy (say, by using extremely long conjunctions and disjunctions
in one’s analysans). But such a way of achieving extensional adequacy
19. This line of thought is implicit, I think, in Williams, “Internal and External Rea-
sons,” 102–3, and “Internal Reasons and the Obscurity of Blame,” 36; Railton, “Facts and
Values,” 12; and Lewis, “Dispositional Theories of Value,” 77.
will be objectionably ad hoc, a case of cheating, really. What would save

the theorist is if the patched-up version of the theory still scored rea-
sonably high on the list of theoretical virtues, if it was not too ad hoc,
if it had some independent appeal, some rationale distinct from its
purported extensional adequacy. This is a perfectly general point about
theory choice: scientific theories too attempt more than mere exten-
sional adequacy. If I put forward a scientific theory (say, that water is
H2O), found purported counterexamples, and then patched the theory
up in such an ad hoc way (water is H2O unless it is Tuesday and the
water is in a wooden cup, and unless . . . ), the resulting theory would
be unacceptable even if extensionally adequate. What would save the
amended theory would be if it was still simple, explanatory, coherent,
predictive, elegant, pragmatically useful, or scored reasonably high on
whatever else is on the list of virtues for scientific theories. In the sci-
entific as well as in the philosophical case, then, what is needed for a
theory to be attractive is some rationale distinct from its purported
extensional adequacy. That such a rationale is needed can also be seen
from the obvious fact that there are many—indeed, infinitely many—
distinct ways of achieving ad-hoc-extensional adequacy, all internally
consistent but mutually inconsistent. Only one of them can be right,
and we are going to determine which on grounds independent of ex-
tensional adequacy (for ex hypothesi all of them are extensionally ad-
equate). The problem here is not the mere availability of competing
ways of achieving extensional adequacy, but rather the fact—exemplified
by their availability—that the hope of achieving extensional adequacy
cannot on its own motivate an amendment to the original theory, unless
the amended theory is still good enough as a theory.
In order to be plausible and avoid being objectionably ad hoc, then,
the idealizer must have a further rationale for her idealization, one that
is not merely the hope of avoiding obvious counterexamples and thereby
achieving extensional adequacy. But this means that we are back at
square one: remember, what we set out to find was a plausible rationale
for idealization that is consistent with the philosophical concerns un-
derlying response-dependence views. The suggestion was that exten-
sional adequacy could serve as such a rationale. We now see, though,
that it cannot do that satisfactorily without some other, independent,
rationale for idealization. We are still looking, then, for what we have
been looking for all along—a rationale for idealization.
Things would have been different had the philosophical concerns
underlying response-dependence views been themselves neutral as be-
tween actual and ideal responses.20 If you have reason to tie the relevant
20. I thank Mark Schroeder for emphasizing this point to me (in the form of an
objection).
normative facts to, say, motivations of whatever sort, be they actual or

hypothetical, then the extensional inadequacy of actual-response-
dependence theories together with the better prospects of idealized-
response-dependence theories would give you all the reason you need
for idealization. But the philosophical concerns typically underlying re-
sponse-dependence views are not neutral as between actual and ideal-
ized responses. For instance, Williams’s internalist intuition, according
to which, roughly speaking, the reasons that apply to one must be able
to explain one’s behavior, applies to the actual behavior of actual agents,
not their idealized counterparts.21 Similarly, Railton’s intuition that what
is good for me must suit me, be made for me, or engage me is, of course,
about a connection between what is good for me and me as I actually
am, not my ideal advisor.22 The point generalizes: if only extensional
adequacy could be had on an actual-response-dependence view, it seems
no response-dependence theorist would idealize. And this alone suffices
21. Williams, “Internal Reasons and the Obscurity of Blame,” 102. It has been noticed
in the literature that idealization opens up a gap between one’s reasons (or values, one’s
good, or whatever) and one’s motivations (or other responses), thus threatening to be
inconsistent with stricter versions of internalism. See, e.g., Noggle, “Integrity, the Self, and
Desire-Based Accounts of the Good,” and Sobel, “Explanation, Internalism, and Reasons
for Action.” Noticing such a gap, Rosati has suggested a two-tier internalism, one in which
the relevant idealization is itself determined by what conditions the agent thinks of or
cares about as ideal. (A similar point, I think, is implicit in Williams [ibid., 37], where he
says that what justifies eliminating factual errors but not normative ones in the idealization
process is that we all already care about factual accuracy.) Rosati’s interesting suggestion
faces its own difficulties, which I hope to discuss elsewhere. For my purposes here it is
sufficient to note that her suggestion starts from acknowledging the tension between strong
internalist intuitions (according to which, roughly, what matters to me is what matters to
me) and any idealization. See Connie Rosati, “Internalism and the Good for a Person,”
Ethics 106 (1996): 297–326. Interestingly, putting forward her two-tier internalism as the
theory that best captures internalist intuitions but acknowledging its possible extensional
inadequacy, Rosati considers the possibility that her reasoning entails the refutation of
internalism (ibid., 309). This is where her argument is closest to mine.
22. Railton, “Facts and Values,” 9. This is even more clearly true of historicized versions
of ideal response-dependence views, such as the one recently suggested by Zimmerman.
For surely, and despite Zimmerman’s claim to the contrary (“Why Richard Brandt Does
Not Need Cognitive Psychotherapy,” 392), it is a huge stretch of internalist intuitions to
say that they are satisfied by a connection between my reasons or what is good for me
and the motives I would have had if I had a different history. David Copp suggested a
further underlying motivation for response-dependence theories: that of accommodating
the intuition that our actual responses are at least rough indicators of the relevant nor-
mative truths. But going response-dependence and idealizing is not a very good way of
accommodating this intuition, because then our actual responses are good indicators of
the normative truths only to the extent that they resemble our idealized responses, and
there is no guarantee that they resemble them to a satisfactory degree. Here, too, then,
the idealization severs the perhaps initially plausible tie between our actual responses and
the normative truths. (And, of course, there may be other ways of accommodating the
mentioned intuition.)
to show that typical response-dependence theorists are motivated by

considerations that are not initially neutral between actual and idealized
responses. As even the geography of their texts shows, idealizers start
off with actual responses, then patching up extensional inadequacies
by idealizing. And this is exactly the move characterized above as ob-
jectionably ad hoc.
B. Our Justificatory Practices

But perhaps there is another rationale for idealization, one that is con-
sistent with the philosophical motivations underlying idealizing views
and that is not disturbingly ad hoc. For the data we can use when
philosophizing about these matters include not just straightforwardly
normative beliefs (about which actions are right, what is and what is
not good for one, what reasons we have, which things are of value, etc.)
but also our justificatory practices and our beliefs about them. We think,
for instance, that if you say that lifelong solitude is of value, and we
show that you do not have a good appreciation of what such life would
be like, we have thereby discredited you as a judge of the relevant value
claim. Similarly, we think that if you say you have a reason to drink from
the glass in front of you and then find out that it is petrol and not gin,
you will take back your original judgment (“Oh, so I don’t have a reason
to drink this stuff after all!”); you will not conclude that a change in
your reasons has just been brought about (“Oh, so now I don’t have a
reason to drink this stuff; still, until just a moment ago, I did”). Some-
what more generally, we think that we are fallible in our normative
judgments and that there is room for genuine normative advice and
for coming to see that we were mistaken about our reasons.23 And so
on. The hope is, then, that from our practices of justifying normative
claims a rationale for idealization can be extracted.
The idea seems to be this.24 Characterize our practice of making
23. These features comprise the very minimal sense of normativity which Williams
(e.g., “Internal and External Reasons,” 104) argues his theory can capture. I think that
this is much too minimal a sense, but this is a point I cannot develop here.
24. I found it made most explicitly in Lewis (“Dispositional Theories of Value,” 87),
though there it is presented as an argument not so much for the truth of his idealizing
view as an argument for its status as analytic. I think that similar lines of thought are
implicit also in Firth, “Ethical Absolutism and the Ideal Observer,” 332; Stephen Darwall,
Impartial Reason (Ithaca, NY: Cornell University Press, 1983), chap. 8; Railton, “Facts and
Values,” 15; Johnston, “Dispositional Theories of Value,” 163; Stephen Darwall, Allan Gib-
bard, and Peter Railton, “Toward Fin de Siecle Ethics: Some Trends,” Philosophical Review
101 (1992): 115–89, reprinted in Darwall, Gibbard, and Railton, Moral Discourse and Practice,
3–47, 29; Ripstein, “Preference”; Michael Smith, The Moral Problem (Oxford: Blackwell,
1994), 31–32, and “In Defense of The Moral Problem: A Reply to Brink, Copp, and Sayre-
McCord,” Ethics 108 (1997): 84–119, 91 and 103–4; Valerie Tiberius, “Full Information
and Ideal Deliberation,” Journal of Value Inquiry 31 (1997): 329–38; Noggle, “Integrity, the
and justifying reason or value claims, or other relevant normative claims.

If we try to correct for false information, this must be because values
(say) are not what we value when falsely informed, but what we value
when not so mistaken; if we try to put ourselves in the relevant situation
and imagine what it is like to be in it, this must be because values are
what we value when possessing imaginative acquaintance with the thing
being valued; and so on. And this, it is important to notice, is an in-
dependent way of motivating the idealization: it does not depend on
considerations of extensional adequacy.25 It relies not on pretheoretic
judgments about what is of value, say, or what is good for one, but rather
on the standards implicit in our relevant justificatory practices (and
perhaps, to an extent, also on our beliefs about them). And notice, of
course, that a theory based on the standards of justification implicit in
a practice need not be (too) objectionably conservative. It can be critical
by exposing tensions within the practice and attempting to settle them
in accord with the deeper, more central features of the relevant practice.
Made fully explicit, I think that the argument is an instance of an
inference to the best explanation. In attempting to determine whether
something is of value we—let us assume—try to put ourselves in con-
ditions of full imaginative acquaintance. Furthermore, we think that this
is a good method of inquiry into what is of value. What explains, then,
the fact that we put ourselves in conditions of full imaginative acquain-
tance in order to determine whether something is of value? And what
explains the fact that putting ourselves in such conditions (or at least
approximating them) is a good method of determining what is of value?
Well, it is that the requirement of full imaginative acquaintance is a part
of what it takes for something to be of value. The rationale for the
idealization is, then, that it best explains important aspects of our rel-
evant justificatory practice.
I want to concede here that if this line can be made to work, it
supplies a rationale for idealization of exactly the sort needed. If such
inference to the best explanation works, there is nothing ad hoc in the
idealization, and the challenge is met. But, I now want to argue, this
inference to the best explanation cannot be made to work.
To see that, think of a view of religious obligation I will call “Ideal
Prophet Theory.” The Ideal Prophet Theorist denies the existence of
God and other supernatural entities and properties, perhaps because
Self, and Desire-Based Accounts of the Good,” 306; and Rosati, “Agency and the Open
Question Argument,” 520.
25. It may even be a way of motivating the specific idealization the relevant idealizer
puts forward (an option that both Ripstein and Lillehammer fail to take into account
when arguing that idealizers lack the resources to privilege one set of hypothetical cir-
cumstances over any other). For a related point, see also Rosati, “Agency and the Open
Question Argument,” 519.
of his (the theorist’s) naturalist leanings. But for some reason he wants
to save the truth of (some) statements about what is religiously re-
quired.26 So he comes up with the following theory: a type of action is
religiously required just in case it is sensed-as-required by a prophet in
ideal conditions, or by an ideal prophet. The theorist then comes up
with a psychological account of sensing-as-required27 and specifies—in
a naturalistically respectable way—the conditions that make a prophet
ideal. He decides which conditions to include in this specification on
the basis of our practice with religious discourse. If we treat being in a
distinct kind of trance as evidence for the reliability of the purported
prophet, he includes such conditions in the conditions that make a
prophet ideal. If we treat being drugged as undermining the religious
force of the trance, he classifies being drugged as a condition that makes
a prophet less than ideal. And so on. Of course, the Ideal Prophet
Theorist argues not just for an extensional equivalence between being
religiously required and being sensed-as-required by an ideal prophet.
He argues that being sensed-as-required by an ideal prophet is what the
truth of claims about being religiously required consists in, that being
religiously required just is being sensed-as-required by an ideal prophet.
Now suppose I put forward to the Ideal Prophet Theorist the chal-
lenge of coming up with a rationale for his idealization. Like the meta-
normative idealizer, he cannot invoke the most natural reply to the
challenge because he does not think of the responses of the ideal
prophet as tracking an independent order of facts about what is and
what is not religiously required, commanded by God, or some such (he
is a naturalist, after all). He can, again like the metanormative idealizer,
motivate the idealization by considerations of extensional adequacy, but
if this is all he can say, this will leave the idealization objectionably ad
hoc. Luckily, he can do better than that: he can refer to our religious
practice, the practice of talking about what actions are religiously re-
quired, of justifying such claims, of discrediting some who make them,
and so on. And he can present his argument as an inference to the best
explanation: what best explains, he can ask, our practice of justifying
claims about what is religiously required by consulting (or approxi-
mating) an ideal prophet? And what explains the fact that this is a good
26. What he wants to save are statements about what is (really, as it were) religiously
required, not merely statements about what participants in the relevant religious practice
consider religiously required. He wants to argue that some actions (say) really are reli-
giously required.
27. Having read Firth (“Ethical Absolutism and the Ideal Observer,” 326) and being
concerned to avoid vicious circularity, he is careful not to include in this psychological
account a belief that the relevant type of action is religiously required, or anything close
enough to such a belief.
way of coming to know which actions are religiously required? The

answer is that this is (a part of) what it is to be religiously required.
Ideal Prophet Theory fails for more than one reason, and I do not
mean to suggest that metanormative idealizing views fail for all the
reasons for which Ideal Prophet Theory fails. As will become clear
shortly, I will pursue a much more specific analogy. I want to highlight
one failure of the reasoning above. The problem I want to focus on
stems from the gap between the commitments of the Ideal Prophet
Theorist and those of the participants in the relevant religious practice.
The former is a naturalist, denying the existence of God and other
supernatural facts in which religious obligation may be thought to be
grounded. But the commitments of the participants in the relevant
religious practice are not plausibly understood as consistent with nat-
uralism.28 They can only be understood as committed to exactly the kind
of metaphysics the Ideal Prophet Theorist wants to deny. This gap makes
the Ideal Prophet Theorist’s argument into a non sequitur: the best
explanation of the relevant justificatory practice is not in terms of the
theorist’s metaphysical beliefs (including the denial of the existence of
God), but in terms of the participants’ ones. Participants believe in God
(or in other supernatural facts), and so the best explanation of the
relevant part of religious practice is the obvious one: participants believe
that the relevantly ideal conditions are the conditions best suited for
the tracking of an independent order of facts regarding religious ob-
ligations (say, God’s commands).29 Perhaps they are wrong, perhaps
because there is no God. But clearly, their belief that there is a God is
very relevant to answering the question of what best explains their re-
ligious practice. The theorist’s reasons for denying the existence of
God—good as they may be—cannot support the claim that what best
explains religious practice is the Ideal Prophet Theory and the ideali-
zation it employs.
What best explains, then, the relevant part of religious practice is
the (false, we’re here assuming) belief of participants in God, and their
belief that the specified ideal conditions are well suited for finding out
about his commands and so about their religious obligations. But earlier
on I emphasized that what needs to be explained is not just the justifi-
catory practices but also the fact that they are good, or justified, methods
of forming beliefs about the relevant facts. It should now be clear that
this is not quite right. The naturalist who attempts to make sense of the
28. This is so even if some participants endorse Ideal Prophet Theory. The question
relevant here is not what (if any) explicit metareligious beliefs participants have (or think
they have), but rather what metareligious commitments are embedded (though perhaps
implicitly) in their practice.
29. Or that these ideal conditions are indicative of being in such epistemic conditions.
discourse about religious obligation must, if he is honest, concede that

the practice is best explained as involving a commitment to some su-
pernatural facts. This means, of course, on this naturalist’s assumptions,
that the relevant part of religious discourse is systematically erroneous.
And this means that the procedures of finding out about religious ob-
ligations licensed by the practice are after all not good ones, because
they are founded on false beliefs.30
Let me summarize, then, the Ideal Prophet Theory detour: an Ideal
Prophet Theorist cannot rely, as supplying a rationale for idealization,
on standards implicit in our practice of justifying claims about religious
obligation. Of the two suggested inferences to the best explanation—
one explaining our practices, the other explaining their merit—the
former fails because a much better explanation is available (in terms
of the participants’ false belief in God), and the other fails because the
purported explanandum is, in effect, denied by the Ideal Prophet The-
orist himself.
I want to suggest that the metanormative idealizer who tries to
supply a rationale for her idealization by drawing on our justificatory
practices fails in an exactly analogous way. Regardless of how good her
(metaphysical, epistemological, or whatever) reasons are for denying a
more robustly realist view of the relevant normative truths, still these
are not reasons to deny that our justificatory practices are committed
to some such realism.31 This being so, what best explains our justificatory
30. Perhaps participants are erroneous because they believe in God and that their
methods of inquiry track his commands. But perhaps their methods of inquiry are still—
as a matter of fact—good at tracking facts about what is religiously required, because such
facts are reducible to facts about the responses of ideal prophets, and the methods of
inquiry successfully track these. This may be so, and if it is, this may be a good reason to
favor a revisionary account. I return to this option in Sec. VI. Am I assuming here any
controversial theses about the nature of epistemic justification? As just noted, the partic-
ipants’ method of inquiry may in fact be reliable even though for reasons different from
the ones they have in mind. So am I assuming—when saying that the Ideal Prophet Theorist
is committed to denying the claim that the participants’ method is a good one—the denial
of reliabilism about epistemic justification? Not really. Even reliabilists about justification
typically concede—as they ought to—that reliability only gives an account of prima facie
justification. They agree that this prima facie justification may be outweighed by other
facts (say, a belief, itself justified, that the relevant belief-forming mechanism is unreliable).
The case in the text is a case where such outweighing is clearly in place: the participants
employ epistemic methods which are in fact reliable, believing that what makes them
justified is that they track an independent order of facts about God’s commands. This
belief is false, and this defeats—or perhaps undermines—the justificatory status their
methods and (other) beliefs may have had otherwise.
31. And here too what is relevant is not the explicit metanormative beliefs—much
less the explicit metanormative statements—of participants in normative discourse. What
is relevant, rather, are the deep metanormative commitments embedded (perhaps im-
plicitly) in normative discourse and practice themselves. The fact that many sophomores
practices is not that an idealized response is what the relevant normative

fact consists in. What best explains our justificatory practice is rather
our (perhaps implicit) belief, false though it may be, that, say, conditions
of full imaginative acquaintance are conducive to the reliable tracking
of an independent order of value-facts. And the idealizer cannot con-
sistently require an explanation why it is a good justificatory method,
because she believes it is not: she believes that there is no independent
order of value-facts that our epistemic methods reliably track.
Perhaps, though, this is too quick. For the idealizer can argue that
there is an important disanalogy between her theory and that of the
Ideal Prophet Theorist. Whereas it really is highly implausible that the
relevant part of religious discourse and practice is not committed to
the existence of supernatural facts in which religious obligation is
grounded, in the case of normative discourse this is not so. Take God
away from talk about what is religiously required, the idealizer can say,
and clearly error theory is the way to go. But take a more robust realism
away from talk about what actions are right, what is good for one, what
reasons we have, or what is of value, and still error theory is not the
only—not even the most plausible—way to go. Indeed, the idealizer may
argue that assuming otherwise is already assuming that response-
dependence views are false, and so that in so assuming I have in effect
been begging the question against her.
I agree, of course, that the metanormative case is not as clear-cut
as the Ideal-Prophet-Theory case (this, after all, is why I thought it
helpful to use the latter as a heuristic device). But this line of thought
cannot save the metanormative idealizer. Let me make, then, two points
in reply.
First, in assuming that our practice is not best explained by the
idealizer’s view I have not been assuming that response-dependence views
fail. I have not been assuming that some other view—certainly not that
a specific other view, such as a robust realism of sorts—is on the whole
more plausible than the idealizer’s view. All I have been assuming is that
as far as fidelity to our practice is concerned response-dependence the-
ories are not as good as some of the alternatives. And making this latter
assumption is in no way tantamount to assuming the former one, because
(and not only them) express some subjectivist or relativist metanormative intuitions thus
has very little weight in assessing the commitments of normative discourse. Indeed the
attempt to motivate idealization by referring to our practice is, it seems, an attempt to
motivate the idealization by reference to the standards implicit in our normative practice,
not to whatever explicit metanormative beliefs participants may or may not have. In draw-
ing lessons not from the explicit metadiscourse beliefs of (some) participants but from
the commitments more deeply embedded in the relevant discourse and practice, I am
here following, of course, a philosophical tradition exceeding the interests of metanor-
mative theory. I thank Bruce Brower for pressing me on this and related points.
for all I have said there may be other considerations—ones not having
to do with fidelity to our practice—that count in favor of response-
dependence views and, indeed, that outweigh such fidelity considerations,
thereby rendering response-dependence views on the whole justified (I
return to this point shortly). Indeed, even response-dependence theo-
rists often concede that their view does not capture the commitments
of the (relevant part of) normative discourse and practice as a more
robust realism.32 Typically, they argue for their view by emphasizing some
other advantages it has over robust realism—say, by avoiding an ontology
of queer entities and a queer epistemology that accompanies it. Consider
again Ideal Prophet Theory. The Ideal Prophet Theorist may concede
that his theory is not as loyal to the practice as some competing (su-
pernaturalist) view but may argue that his view should still be preferred
on the whole because of its naturalist-friendly ontology. I have said
nothing incompatible with such a claim. But if it is true that Ideal
Prophet Theory is not as loyal to the practice as some supernaturalist
view, then the Ideal Prophet Theorist cannot rely on the practice in
order to motivate his idealization. Perhaps, in other words, the discrep-
ancy between Ideal Prophet Theory and religious practice cannot all by
itself defeat Ideal Prophet Theory (because of its many other advan-
tages), but such discrepancy does prevent the Ideal Prophet Theorist
from relying on the commitments of the practice in putting forward a
positive defense of his view, for instance, in providing a rationale for
his idealization. Analogously, perhaps the discrepancy between idealiz-
ing views and (the relevant part of) normative discourse and practice
is not sufficient to defeat such views (because of their many other ad-
vantages), but such discrepancy certainly does prevent the idealizer from
relying on the commitments of the practice in putting forward a positive
defense of her view, for instance, in providing a rationale for her ide-
alization.
The dialectical situation, then, is significantly different from the
common one where the phenomenology of moral discourse is taken to
support a rather robust realism but sensible metaphysical and episte-
mological considerations are taken to support a different, perhaps re-
sponse-dependence, view. For we are not at this stage attempting a com-
prehensive assessment of competing metanormative views.33 Rather, we
32. See, e.g., Lewis, “Dispositional Theories of Value,” 92.

33. But if you find the metaphysical and epistemological commitments of a more
robust realism utterly unacceptable, aren’t you at least then justified in accepting an
idealizing response-dependence view, learning to live with whatever theoretical shortcom-
ings it may have? No, because the metanormative field is not exhausted by these alter-
natives. Perhaps you should endorse a noncognitivist view of sorts; perhaps a naturalist
view that reduces normative properties to some objective, response-independent natural
properties; or perhaps you should go for a metanormative error theory. In this article,
are now attempting to evaluate one way in which the idealizer tries to
motivate her idealization, namely, by relying on the standards of justi-
fication implicit in normative discourse. And in order to evaluate this
attempt at addressing the why-idealize challenge, we should focus our
attention just on considerations of fidelity to the practice.
My first point, then, is that my assumption regarding fidelity to our
relevant practices does not beg the question against the idealizing re-
sponse-dependence theorist. The second point I want to make is that
this assumption is quite plausible. Of course, I cannot defend it at length
here,34 but let me make a few preliminary remarks supporting the claim
that our normative discourse as a whole does not support the idealizer’s
argument. First, the point is often made in metaethical discussions—
and rightly so, I think—that moral discourse seems to incorporate an
objective purport and that for this reason a rather robust realism is the
view that best captures the phenomenology of moral discourse.35 And
the analogous point holds, I think, for normative discourse more gen-
erally. But it is very hard—perhaps even impossible—to get this kind of
objectivity on the basis of a response-dependence view, idealized though
it may be.36 If so, this is some reason to be suspicious of the loyalty of
remember, I do not argue for a robust version of realism, only against idealizing views. I
briefly return to the possibility of motivating the idealization by considerations of overall
plausibility in the next section.
34. I have tried to do just that in my dissertation, mostly in chaps. 3 (“Deliberation”)
and 4 (“Rejecting Alternatives”). See David Enoch, “An Argument for Robust Metanor-
mative Realism” (PhD diss., New York University, May 2003), available online at http://
law.mscc.huji.ac.il/law1/newsite/segel/enoch/index.html.
35. For such a point made in the context of an introductory text, see Stephen Darwall,
Philosophical Ethics (Boulder, CO: Westview, 1998), 54.
36. This familiar point is emphasized by Jeremy Koons, “Why Response-Dependence
Theories of Morality Are False,” Ethical Theory and Moral Practice 6 (2003): 275–94. Some
idealizers—such as Williams, Brandt, and Railton—acknowledge this fact and celebrate it.
See Williams, “Internal and External Reasons”; Richard Brandt, A Theory of the Good and
the Right (New York: Prometheus, 1979); Railton, “Facts and Values.” Others acknowledge
it and concede that it draws an intuitive price from their view (see Johnston on the
Euthyphro Contrast [“Dispositional Theories of Value,” 70–73] and Lewis on the contin-
gency of value [“Dispositional Theories of Value,” starting on 82]). Yet others are optimistic
about the prospects of accommodating the relevant objective purport, perhaps by guar-
anteeing the convergence of all ideal observers or advisors, consistently with their response-
dependence view. See Roderick Firth, “Reply to Professor Brandt,” Philosophy and Phenom-
enological Research 15 (1955): 414–21, 415; Smith, “In Defense of the Moral Problem,” 89;
Bruce Brower, “Dispositional Ethical Realism,” Ethics 103 (1993): 221–49, 223 and n. 28
on 233. I think that this last strategy is hopeless: the only way to guarantee convergence
of all (possible) ideal agents is, I think, to incorporate into one’s idealization a “whatever-
it-takes clause” (see Johnston, “Dispositional Theories of Value,” 145), or some other similar
normative condition. Given the naturalist motivations of such views, this would be unac-
ceptable. For a similar point made in the context of criticizing Smith, see Donald Hubin,
“Converging on Values,” Analysis 59 (1999): 355–61; see also the exchange between Brandt
and Firth: Richard Brandt, “The Definition of an ‘Ideal Observer’ Theory in Ethics,”
response-dependence views to our relevant justificatory practices. Sec-

ond, it seems plausible to assume that the history of our practices is
relevant to the question of what best explains them. Assuming—as again
seems plausible—that our justificatory practices were shaped in eras in
which a fairly robust realism was much more in fashion than it is today,
this counts against an attempt to argue that the best explanation of
these practices is in terms of idealized responses rather than in terms
of (the attribution of a belief in) a more robust realism. Third, several
idealizers prefer to think of their views as revisionary accounts, not
sufficiently loyal to our practices to count as an adequate descriptive
theory of the normative (I return to such revisionary accounts in Sec.
VI). This, I think, is some—though certainly not conclusive—evidence
that it is not going to be easy to convince us that the best explanation
of our justificatory practices is going to yield a rationale for idealization
that is consistent with the philosophical concerns typically underlying
idealizing views.37
These considerations are, of course, very preliminary. Let me em-
phasize, then, that the conclusion they are meant to establish in our
context is rather weak, in the following ways: first, as already emphasized,
in our context such considerations are not meant to undermine—all
on their own—idealizing response-dependence views of the normative.
Rather, they are supposed to show merely that the idealizer cannot
appeal to our relevant practices in her attempt to motivate her ideali-
zation. Second, remember that my targets here are views that opt for
the non-Socratic direction of the Euthyphro Contrast, views according
to which, for instance, something is of value because we would value it
under ideal conditions (and not the other way around). So for the
argument here to work it is sufficient that the considerations from the
previous paragraph count against such views. In particular, even if no-
priority theorists such as McDowell and Wiggins—who attempt to steer
a middle course of sorts between the two contrasting Euthyphronic
alternatives—capture the commitments of our practice better than a
more robust realism, this is still sufficient to show that the idealizers
who are my target here—those who believe in the priority of our re-
sponses over the relevant normative facts—cannot rely on our relevant
practices in motivating their idealization. Third, I do not here argue
against intranormative reductions that make use of some idealization.
Even if, for instance, the best account of the rightness of actions is in
terms of the responses of the virtuous (or the practically wise) person,
Philosophy and Phenomenological Research 15 (1955): 407–13; and Firth, “Reply to Professor
Brandt.” But this matter merits a more detailed discussion than I can supply here.
37. Perhaps Lewis’s mistake—discussed in Sec. II above—strengthens this point, for
it shows that even such a prominent idealizer is tempted by our practice to say things that
sound more robustly realist.
and indeed even if such a view best captures the commitments implicit
in our practice of making and justifying claims about moral rightness,
still this is not in tension with anything so far said, because of the
normativity of the notion of the virtuous (or practically wise) person.38
The crucial point is that no nonnormative response-dependence ac-
count of the normative can plausibly claim to be sufficiently loyal to
our relevant practice—loyal enough, that is, in order to motivate her
idealization by referring to the commitments implicit in it.39
The considerations from two paragraphs ago, then, are supposed
to establish a very modest point. Even with this in mind, these quick
considerations are perhaps not conclusive. Furthermore, in order fully
to flesh out these and related considerations, it will be necessary to
examine the details of the relevant part of normative discourse. Here,
in other words, it will no longer be possible to reach conclusions whose
scope includes all metanormative idealizing views, regardless of their
more specific subject matter, because a detailed examination of our
justificatory practices may uncover important and relevant differences
between different parts of our normative commitments.40 Nevertheless,
the challenge of coming up with a rationale for idealization applies
across the metanormative board, and it should be clear that—regardless
of the more specific subject matter—in the context of coping with this
challenge it is here that the battle is to be fought: in attempting to
characterize and explain our relevant practice and not, for instance, in
the more general realms of the metaphysics and epistemology of the
normative. The less-than-conclusive considerations above should suffice
to show that for the idealizer it is not going to be an easy battle to win.
IV. OTHER OPTIONS?

I have only discussed three possible ways of motivating idealization: the
natural reply (which presupposes a commitment to a more robust real-
ism), considerations of extensional adequacy (which would not alleviate
the air of ad-hoc-ness about the idealization), and arguments from the
standards of justification implicit in our practice (which fail because even
if the idealizer’s theory is overall the best one out there, still it is not loyal
enough to our practice to rely on arguments from its characterization).
38. McDowell is thus out of the picture twice over: for not only does he offer a no-
priority view, but he also includes an ineliminable normative element in his response-
dependence account. See, e.g., McDowell, “Projection and Truth in Ethics,” 221, and
“Values and Secondary Qualities,” in Morality and Objectivity, ed. Ted Honderich, 110–29
(London: Routledge & Kegan Paul, 1985), reprinted in Darwall, Gibbard, and Railton,
Moral Discourse and Practice, 201–13, 207.
39. I briefly return to intranormative idealizing views in n. 47.
40. I thank David Copp for pressing me on this point, to which I return also in Sec.
V.
Notice that if the argument of the previous sections goes through,

combined strategies likewise fail. An attempt to motivate idealization by
drawing, say, both on the standards of justification implicit in our prac-
tice and on considerations of extensional adequacy will suffer from the
problems of both replies and will therefore fail. The same goes for
somewhat different attempts to motivate idealization that are, at bottom,
parasitic on one or more of the types of considerations already discussed.
For instance, an attempt to motivate idealization by the need to reach
a wide reflective equilibrium,41 one that incorporates (as much as pos-
sible) our pretheoretical judgments about (say) what we do and what
we do not have reason to do, some more general reflective judgments
(say, about how different circumstances affect the reliability of normative
judgments), and some relevant theories we hold (such as, perhaps, a
theory of knowledge or of epistemic justification)—such a strategy too
will not save the idealizer. This is so because once unpacked, this strategy
is really a combination of previous ones (of achieving extensional ad-
equacy and respecting the standards implicit in our relevant practice),
or of considerations parasitic on them.
All this does not amount, of course, to an argument establishing
that the motivations for idealization already discussed (and those some-
how based on them) are the only possible alternatives. And though I
do think it is highly plausible that this is so (otherwise, wouldn’t it be
surprising, after all, that after decades of rather intense philosophical
interest in such views, there is no hint in the literature of any other
rationale for the idealization?), I do not have an argument establishing
this fact. So think of my argument as a challenge directed at the idealizer,
showing that she has some unfinished business to attend to. She has to
come up with a rationale for her idealization that is consistent with the
philosophical motivations underlying her idealizing view. The fact that
all attempts at such rationales I can reconstruct from the literature fail
41. An anonymous referee suggested a related motivation for idealization: achieving

extensional adequacy in a way that is acceptable on reflection. This suggestion too fails,
it seems to me. First, it seems largely to comprise a combination of the second and third
replies discussed above: achieving extensional adequacy, and doing so in a way that respects
the standards of justification entrenched in our relevant practices. And second, it is not
clear what “acceptable on reflection” exactly means here. It seems to me that to avoid
just being an instance of the third strategy, this restriction must incorporate something
like a requirement that extensional adequacy be achieved consistently with the philo-
sophical motivations underlying the relevant idealizing view (it will clearly not do, for
instance, if a way of achieving extensional adequacy is acceptable on reflection simply
because our reflection is [implicitly] guided by hidden robustly realist commitments). But
then this is the requirement I have been emphasizing all along, and assuming without
further argument that it can be met would be begging the question against me.
gives reason to think, at the very least, that adequately addressing this
challenge is not going to be easy.42
But can the idealizer refuse to play this game altogether? Can’t she
point to the other advantages of her theory—say, that it renders nor-
mativity consistent with a naturalistically respectable ontology and epis-
temology and that it is (roughly) extensionally adequate—and argue
that even without a separate rationale for the idealization, still her theory
is the most plausible theory of normativity overall?43 Of course, this claim
has to be argued for by assessing for overall plausibility all competing
theories of normativity, an endeavor I cannot embark on here. And
given just how disturbingly ad hoc the idealization is, I find it hard to
believe that idealizing views are going to end up being the most plausible
ones overall.44 But let me concede the following point: nothing need
be wrong with such a rejoinder from the idealizer. If, as seems likely,
the debate over normativity must be decided on grounds of plausibility,
then nothing I have said proves that idealizing views will not after all
emerge victorious. Perhaps, for instance, all theories of normativity suf-
fer from serious shortcomings, and perhaps the absence of a rationale
for the idealization is not among the most severe of those.
Even if this is not a fatal flaw, however, it is nevertheless a flaw. To
see this, assume two idealizing theories that are just alike in all aspects
relevant for theory evaluation except for the following difference: the
first theory has no independent motivation for the idealization available
to it, and the second one does. Isn’t it clear, then, that given that other
(relevant) things are equal, the second theory is to be preferred to the
first (because, somewhat roughly, it is less ad hoc, or more coherent in
some important sense of this word)? In other words, the fact (if it is
a fact) that idealizers cannot motivate the idealization they employ in
a way consistent with the considerations underlying their response-
dependence views makes their theory less attractive—because more ad
42. The point is sometimes made that idealizing views with reductive, naturalist as-
pirations have no nonarbitrary way of choosing among all possible ways of idealizing (see
Johnston, “Dispositional Theories of Value,” 155; Ripstein, “Preference”; Lillehammer,
“Revisionary Dispositionalism and Practical Reason”). I think that this critique is not with-
out power (though see n. 25 above), and it may be related to my criticism of idealizing
views. But I want to note that it is nevertheless distinct from it. Johnston, Ripstein, and
Lillehammer argue that the idealizer has no way of motivating one idealization rather
than another. I argue that she has no way of motivating any idealization at all. In fact, my
argument can be seen as a reply (at least to Lillehammer): there are very good reasons,
if I am right, to privilege one set of circumstances in the context of typical idealizing
views—namely, the actual circumstances.
43. I thank Dalia Drai and David Copp for this and related suggestions.
44. I am not a naturalist. But even if you are, still you should not dismiss the point
in the text. Remember—idealizing views are not the only naturalist game in the meta-
normative town.
hoc—than it otherwise would have been. And this—to paraphrase David

and Stephanie Lewis—at the very least raises the stakes.45 It shows just
how significant the advantages of idealizing views must be if we are to
accept them despite their being ad hoc in this way.
True, my challenge doesn’t conclusively refute idealizing views (how
often have you come across an argument of that sort?). But at the very
least it exerts a price from idealizers, it shows that idealizing views are
less plausible (much less plausible, I would say) than may otherwise
seem. It is rare in philosophy that we can do much more.
V. GENERALIZING
Idealizing response-dependence views come up in many contexts. As
already mentioned, they come up in the metanormative contexts, with
different normative concepts as the relevant analysandum (the four I
have been using as my examples do not exhaust the list). And they come
up elsewhere as well—most notably, perhaps, in philosophical accounts
of secondary qualities such as colors and in aesthetics.46 Can my argu-
ment against idealizing views be generalized so as to apply to such other
instances of idealized response-dependence views?
This is, of course, a question of independent interest. But it can
also be the beginning of an objection. For if my argument seems to
generalize to contexts where response-dependence views seem highly
plausible, this would justify a strong suspicion that somewhere my ar-
gument has gone wrong. For this reason too, then, the question of the
generalizability of my argument deserves discussion here.
Now the first thing I want to note here is that the challenge—that
of coming up with a rationale for the idealization that is consistent with
the philosophical concerns underlying the relevant response-dependence
view and that is not objectionably ad hoc—applies across the board.47
45. David Lewis and Stephanie Lewis, “Holes,” in David Lewis, Philosophical Papers,
vol. 1 (Oxford: Oxford University Press, 1983), 3–9, 8–9.
46. And perhaps anywhere (else) where an epistemic understanding of (the relevant)
truth is suggested. Indeed, an epistemic theory of truth itself can be thought of as a
response-dependence view, one to which my challenge prima facie applies.
47. What about theories that make use of hypothetical choices or preferences on the
way to presenting an intranormative reduction, that is, a reduction of some normative con-
cepts or properties to other, still normative, ones? Consider, for example, the Rawlsian project
of giving an account of political justice in terms of a hypothetical choice in fair conditions.
It seems to me the challenge of coming up with a rationale for idealization applies here too
but is easily met. The rationale for the idealization—for taking choice in the original position
rather than in actual circumstances to constitute the principles of justice—is precisely that
the original position is (relevantly) fair and actual circumstances are not. But this way of
addressing the why-idealize challenge crucially depends on the occurrence of a normative
concept or property in the analysans (namely, that of fairness). It is not available to meta-
normative idealizing views, views that, perhaps somewhat roughly, attempt a reduction of
That an idealized response-dependence view cannot address this chal-

lenge satisfactorily is a reason for rejecting it, be it a theory of moral
value, prudential value, epistemic justification, colors, or the beautiful.
And this completely general result may already be of some interest, be-
cause it draws attention to some unfinished business idealizers need to
address—they need to show that their view can motivate the idealization
it incorporates.
Furthermore, the unavailability of the first, natural answer to the
why-idealize challenge is also independent of subject matter. So long as
the relevant idealizer thinks of her view in the way characterized in
Section II above—where the relevant truth depends on our relevant
responses, and not the other way around—she cannot motivate the
idealization by considerations regarding the accurate tracking of an
independent truth.
Where the subject matter does make a difference to the argument
of the previous sections is not, then, in the seriousness of the initial
challenge, or in the unavailability of the natural way of addressing it,
but rather in the availability of the other two. Let me take them in turn.
In Section III.A I argued that considerations of extensional ade-
quacy cannot serve as a non-ad-hoc rationale for idealization so long as
the philosophical concerns underlying the relevant idealizing view are
not neutral between actual and ideal responses; so long, in other words,
as idealizers start off thinking about actual responses and then go on
to idealize as a way of patching up their view in order to achieve (rea-
sonable) extensional adequacy. There I also argued that at least in the
more influential versions of metanormative idealizing views, this is ex-
actly the case. But this may not be the case with some other metanor-
mative idealizing views, and it may very well not be the case in other
contexts. Perhaps, for instance, there is good reason to think of colors
as somehow strongly related to our—actual or somewhat idealized—
responses. And perhaps this reason is not initially a reason to believe
that such a connection exists between colors and our actual responses
in particular. If so, if there is reason to go response-dependent regarding
colors that is neutral between actual and ideal responses, then consid-
erations of extensional adequacy may be all the rationale needed for
idealization.
Second, the conclusion of Section III.B—that typical metanorma-
tive idealizers cannot motivate their idealization by drawing on the best
explanation of our relevant justificatory practices—turned on a claim
the normative to the nonnormative. Relatedly, Lillehammer (“Revisionary Dispositionalism

and Practical Reason,” 177–79) argues that response-dependence theories of color (e.g.)—
unlike such theories of reasons—can privilege the relevant set of circumstances, because so
privileging them is a matter of normative commitment, and so in the normative case but
not the color case a vicious circularity will result.
that is specific to the relevant subject matter. There I argued that the
best explanation of our relevant justificatory practices is in terms of a—
perhaps false—belief that they are the best procedures for tracking an
independent order of normative facts. But whether in other contexts
the analogous point holds depends on the specific features of the rel-
evant practices. It is quite possible, and consistent with everything I have
said, that, say, our practices of justifying claims about the beautiful are
not plausibly explained by attributing to participants a (perhaps false)
belief that these are good procedures for tracking the independent
aesthetic facts. Whether this is so depends on the characteristics of
aesthetic discourse and practice, an issue I cannot, of course, discuss
here.
A distinction between idealizing views of normativity and of other
views may emerge in another way as well: in the previous section I
conceded that it is in principle possible that we may have to learn to
live with an ad hoc idealization, if idealizing views of normativity are—
this flaw notwithstanding—still overall the most plausible alternative.
This is true for idealizing views elsewhere as well. And whether idealizing
views are the best alternative depends, of course, on the other available
alternatives. For all I’ve said, then, it is quite possible that even without
an independent rationale for idealization we should go for an idealizing
view of the beautiful, say, because alternative views of the beautiful suffer
from even more serious flaws, whereas in the case of the normative
there are better, nonidealizing alternatives. (For all I’ve said, it is also
possible that the truth goes the other way around.)
To conclude, then: the challenge applies across the board, and—
regardless of subject matter—the natural answer to it is unavailable to
response-dependence idealizers. But, depending on the specific philo-
sophical concerns underlying the relevant idealizing view and on the
specific characteristics of the relevant practice, one of the two other
ways of addressing the challenge may be available to the idealizer. And
if this is so, then the plausibility of ideal-response-dependence views
elsewhere should not raise suspicions about my argument against such
views of (parts of) the normative domain. Furthermore, even if there
is no rationale for the idealization, this flaw may not be equally dis-
turbing in all philosophical contexts.
Indeed, even among metanormative idealizing views there may be
relevant differences both in philosophical motivations and in our rel-
evant justificatory practices.48 To take the latter point first: perhaps our
48. It is perhaps worth mentioning that even some response-dependence theories

that explicitly reject idealization—such as Hubin’s (in “Hypothetical Motivation”) and
Noggle’s—and whose critiques of idealization are not completely unrelated to mine (at
times, Noggle [“Integrity, the Self, and Desire-Based Accounts of the Good,” 322] seems
justificatory practices of moral discourse and of reason-talk are com-

mitted to more robustly realist views of their subject matter (so that we
think of our methods as reliably tracking an independent normative
realm), but our justificatory practices of claims about values and about
what is good for one are not so committed. If so, Firth and Williams
are still in trouble, but Lewis and Railton are off the hook because the
discussion in Section III.B does not after all show that they cannot rely
on our justificatory practices in motivating their idealization. And, mov-
ing now to the philosophical concerns underlying metanormative ide-
alizing views, some idealizers are more clearly naturalist than others,49
some explicitly consider their view a reduction of the normative50 while
others explicitly deny the reducibility of the normative,51 and though
all (I think) hold some kind of internalism dear to their hearts, they
differ widely in which kind it is.52 I want to note, then, that in the case
of the normative my critique is more powerful against naturalist, re-
ductivist, and rather strongly internalist views than against other ide-
alizing views, because it is harder to see how a naturalist, reductivist,
internalist response-dependence theorist—compared to one who is will-
ing to reject reduction, qualify her internalism, and indeed be flexible
about her naturalism—can supply a rationale for idealization.
to flirt with something like my objection to idealization), are nevertheless vulnerable to

a criticism analogous to the one developed in this article. For neither theory settles for
just our actual desires as the foundations of normativity. Hubin privileges intrinsic moti-
vations and what is actually conducive to their satisfaction over all other motivations, and
Noggle privileges desires with which I more strongly identify. Now, these restrictions do
not, of course, constitute idealization, but like idealization they demand some philosoph-
ical rationale. Why, we can ask Hubin and Noggle, do some desires but not all count?
And given the fact that both theories are supposed to stem from the thought that desires
are somehow unique in that all (normative) reasons are somehow grounded in desires,
it is not clear how Hubin and Noggle can respond to the challenge. The natural answer
is, of course, not available to them, because they don’t think of the privileged desires as
tracking an independent order of normative facts. And relying merely on considerations
of extensional adequacy would be here—as in the case of idealizing views—objectionably
ad hoc. Perhaps Hubin or Noggle (or both) can after all adequately motivate their re-
striction to privileged desires. But what should be clear is that—despite their views not
being exactly idealizing views—they are prima facie subject to a challenge exactly analogous
to the one idealizers face.
49. Firth, Williams, Railton, and Lewis are all clear—and explicit—about their nat-
uralism. Johnston and Smith are harder cases.
50. For instance, Firth, “Ethical Absolutism and the Ideal Observer,” 329; Lewis, “Dis-
positional Theories of Value,” 85.
51. Johnston, “Dispositional Theories of Value,” 147.
52. For a powerful criticism of the internalism of Lewis—an idealizer—by Johnston—
another idealizer—see ibid., 160. See also ibid., 161, for a critique of internalism and a
discussion of what he takes to be the kernel of truth in it.
VI. REVISIONARY ACCOUNTS

One kind of idealizing view—one that is not exactly orthodox, but that
is nevertheless not uncommon in the contemporary literature, and that
has already been alluded to above—does escape my criticism, and I want
to conclude by briefly showing how. Some idealizers explicitly concede
that their view is best seen not as an adequate descriptive account of,
say, reasons, values, or one’s good, but rather as a revisionary account.53
They typically despair on the unreconstructed discourse and practice,
thinking of it as badly mistaken, confused, or even incoherent, and they
proceed to suggest their idealizing view as the best reconstruction of
the relevant discourse. By lowering expectations regarding descriptive
adequacy, they can avoid some of the obvious difficulties similar views
face when presented as nonrevisionary, descriptive accounts.
But revisionary accounts have their own standards of acceptability.
First, of course, revisionary theorists have to convince us that the un-
reconstructed discourse is indeed badly confused or incoherent. And
then, they have to argue that theirs is the best revisionary suggestion.
To show that, they have to show that their revisionary theory is good as
a theory (coherent,54 simple, etc.) and also that it is close enough to
the original unreconstructed discourse to count as a revisionary account
of it (this is not, of course, as close as it would need to be in order to
qualify as a descriptive account of it). Furthermore, it must be shown
that the suggested revisionary account has pragmatic advantages, ad-
vantages that make it a good idea to employ the reconstructed discourse
suggested rather than to discard it altogether or to employ some other
possible reconstruction. One way—almost the canonical way—of estab-
lishing the advantages of a suggested revisionary account is showing that
it can give us all we ever wanted from the original, unreconstructed
discourse.55 Of course, it cannot fit the unreconstructed discourse com-
pletely (for the latter is confused or incoherent, and the revisionary
account is supposed to be neither), but it can fit it well enough to supply
(perhaps almost) all we wanted from the original discourse.
Now, even revisionary accounts cannot be too ad hoc—they are,
after all, supposed to score reasonably high on the list of theoretical
virtues. So an idealizing revisionary account still faces the challenge of
53. See Brandt, A Theory of the Good and the Right; Railton, “Moral Realism,” 157; and
Michael Johnston, “Objectivity Refigured: Pragmatism without Verificationism,” in Reality,
Representation and Projection, ed. John Haldane and Crispin Wright, 85–130 (Oxford: Oxford
University Press, 1993), 108. Lewis (“Dispositional Theories of Value,” 93) expresses non-
chalance regarding the classification of his view as a reduction or a revisionary account.
54. Lillehammer can be understood as arguing that no revisionary account of (nor-
mative) reasons along naturalist response-dependence lines can be coherent.
55. See Johnston, “Objectivity Refigured: Pragmatism without Verificationism,” 108–9.
coming up with a rationale for the idealization. What I want to note,

however, is that the revisionist idealizer is going to have an easier time
facing this challenge. Of course, the revisionist idealizer cannot rely on
the natural answer any more than the nonrevisionist idealizer can. And
it is not clear to me whether the revisionist idealizer is in better shape
than the nonrevisionist regarding the ad-hoc-ness of relying merely on
considerations of extensional adequacy in motivating the idealization.56
But it is clear that she is in better shape regarding the possibility of
motivating the idealization by referring to characteristics of our justifi-
catory practices.57 Remember, the problem for the nonrevisionist was
that she was trying to extract a rationale for idealization that is consistent
with her response-dependence view from a practice that does not need
this rationale, because from the point of view of participants the ide-
alization is motivated in the natural, more robustly realist way according
to which ideal conditions are those conducive to the tracking of re-
sponse-independent normative facts. This gap between the idealizer’s
theory and the practice caused troubles for the nonrevisionist idealizer,
but it need not cause troubles for the revisionist. She already concedes
such a gap. And she can (and typically does) argue that the more ro-
bustly realist commitments of the unreconstructed discourse are exactly
what makes it badly confused and that the best revisionary account—
the one that is best able to give us all we wanted from the original
discourse—will get rid of such commitments but keep our practice as
similar to the original one as is otherwise possible. The way to do that,
the revisionist can plausibly argue, is by coming up with an alternative
explanation—and indeed justification—of our justificatory practices,
one that does not rely on (purportedly) confused beliefs in an inde-
pendent order of normative facts. Assuming the details of such reason-
ing can be filled in in a plausible way, the revisionist idealizer is off the
hook.
56. As Bruce Brower noted, there may be some reason to think that the revisionist
idealizer is here in a tougher spot than the descriptive one, because the former doesn’t
even have to achieve extensional adequacy (she is a revisionist, after all) and so is not
entitled to rely on the need to achieve extensional adequacy in motivating her idealization.
But seeing that revisionists still want their account to be close enough to the relevant
discourse and practice, they cannot afford too much extensional inadequacy, and it is
therefore not surprising to find them too idealizing in order to achieve (close enough)
extensional adequacy.
57. Lillehammer (“Revisionary Dispositionalism and Practical Reason,” 176) seems
to suggest that the revisionist cannot appeal to standards implicit in our practice, because—
being a revisionist—she is willing to discard our practice. But Lillehammer ignores the
fact that to be acceptable a revisionist account too must be at least close enough to a
description of the relevant discourse and practice.
Revisionary accounts face their own difficulties, of course.58 But the

difficulty highlighted in this article is not one of them. Indeed, my
argument can be read as supporting the following conclusion: the only
kind of ideal-response-dependence theory that is likely to be able to
support the idealization it incorporates in a way that is not objectionably
ad hoc is the revisionary kind.
VII. CONCLUSION
I conclude, then, that (nonrevisionist) idealizers are not likely to be
able to present a rationale for their idealization that is consistent with
the philosophical concerns underlying their views.59 Of the three ways
of supplying such a rationale I can think of, none is available to the
(typical) idealizer. If this is indeed so, we have, I think, a strong reason
to reject such idealization. Until the idealizer comes up with such a
rationale for her idealization, then, the response-dependence theorist
is faced with a choice: settle for actual, nonidealized responses or else
abandon response-dependence altogether. If the former is as unattrac-
tive as it seems,60 the prospects for an adequate response-dependence
theory of normativity do not seem promising.
58. For one of these, see Enoch, “An Argument for Robust Metanormative Realism,”
sec. 4.2.
59. Sobel (“Explanation, Internalism, and Reasons for Action,” 226) seems to ac-
knowledge a challenge close to (an instance of) the one I put forward to the idealizer
and expresses confidence that it can be successfully met by his favorite idealizer (which
he calls “subjectivist”). But he does not show how this can be done.
60. As Zimmerman, an idealizer of sorts, puts it: “No sensible neo-Humean wishes
to ground practical reasons for a person exclusively in the person’s actual motives” (“Why
Richard Brandt Does Not Need Cognitive Psychotherapy,” 390).

Why Idealize PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Why Idealize PDF

Uploaded by

Copyright:

Available Formats

Why Idealize?

Ethics 115 (July 2005): 759–787

claims that this is what it is to have a reason to F (an internal reason,

idealize, but rather as a reason to reject response-dependence theories

II. WHY IDEALIZE—THE NATURAL ANSWER

If someone has little notion what it would be like to live as a free

order of normative facts. As emphasized above, they think of the relevant

III. WHY IDEALIZE—POSSIBLE ANSWERS

will be objectionably ad hoc, a case of cheating, really. What would save

normative facts to, say, motivations of whatever sort, be they actual or

to show that typical response-dependence theorists are motivated by

B. Our Justificatory Practices

and justifying reason or value claims, or other relevant normative claims.

way of coming to know which actions are religiously required? The

discourse about religious obligation must, if he is honest, concede that

practices is not that an idealized response is what the relevant normative

32. See, e.g., Lewis, “Dispositional Theories of Value,” 92.

response-dependence views to our relevant justificatory practices. Sec-

IV. OTHER OPTIONS?

Notice that if the argument of the previous sections goes through,

41. An anonymous referee suggested a related motivation for idealization: achieving

hoc—than it otherwise would have been. And this—to paraphrase David

That an idealized response-dependence view cannot address this chal-

the normative to the nonnormative. Relatedly, Lillehammer (“Revisionary Dispositionalism

48. It is perhaps worth mentioning that even some response-dependence theories

justificatory practices of moral discourse and of reason-talk are com-

to flirt with something like my objection to idealization), are nevertheless vulnerable to

VI. REVISIONARY ACCOUNTS

coming up with a rationale for the idealization. What I want to note,

Revisionary accounts face their own difficulties, of course.58 But the

You might also like