American Philosophical Quarterly 1

Volume 50, Number 4, October 2013 2

3
4
MORAL INTUITIONISM DEFEATED? 5
6
7
8
Nathan Ballantyne and Joshua C. Thurow1 9
10
11
12
13
W alter Sinnott-Armstrong has developed
and progressively refined an argument against
offers for the premises. Then, in section 2,
we identify a critical but dubious epistemo-
14
15
16
moral intuitionism—the view on which some logical assumption concerning the nature of
17
moral beliefs enjoy non-inferential justifica- defeat that undergirds the argument. Finally,
18
tion.2 He has stated his argument in a few in section 3, we will defend our challenge to
19
different forms, but the basic idea is straight- the EDA by answering two objections.
20
forward. To start with, Sinnott-Armstrong
1. The Empirical Defeat Argument 21
highlights facts relevant to the truth of moral
22
beliefs: such beliefs are sometimes biased, in- Sinnott-Armstrong’s target is an ecumeni- 23
fluenced by various irrelevant factors, and of- cal version of moral intuitionism. Something 24
ten subject to disagreement. Given these facts, like this view has been espoused by moral 25
Sinnott-Armstrong infers that many moral theorists from W. D. Ross and H. A. Prichard 26
beliefs are false. What then shall we think to Robert Audi and Michael Huemer.3 “Moral 27
of our own moral beliefs? Either we have intuitionism,” says Sinnott-Armstrong, “is 28
reason to think some of our moral beliefs are the claim that some people are adequately 29
reliably formed or we have no such reason. If epistemically justified in holding some moral 30
the latter, our moral beliefs are unjustified. If beliefs independently of whether those people 31
we have reason to think some moral beliefs are able to infer those moral beliefs from any 32
are reliably formed, then those beliefs are not other beliefs” (2006a, p. 341; cf. 2008a, p. 33
non-inferentially justified, because then we’ll 48, and 2012, p. 13). The theorists just noted, 34
have reason to accept something—namely, among many others, are intuitionists in that 35
that they are reliably formed—that entails or broad sense. 36
supports those beliefs. But then, either way, Intuitionism has been challenged on differ- 37
our moral beliefs are not non-inferentially ent fronts by its opponents. What is innovative 38
justified, and so moral intuitionism is false. about Sinnott-Armstrong’s argument against 39
This paper takes issue with Sinnott-Arm- intuitionism is that it’s partly driven by em- 40
strong’s interesting and widely discussed pirical claims. He contends that “some recent 41
argument, which we here call the Empirical research in psychology and brain science un- 42
Defeat Argument (EDA). According to us, the dermines moral intuitionism” (2006a, p. 340). 43
EDA does not defeat moral intuitionism. In How is that supposed to work? We’ll pres- 44
section 1, we will set out the argument, briefly ent the EDA as proceeding in two steps. First 45
reviewing the rationale Sinnott-Armstrong of all, Sinnott-Armstrong directs his readers 46
©2013 by the Board of Trustees of the University of Illinois
 412 / AMERICAN PHILOSOPHICAL QUARTERLY
1 toward several types of alleged facts about
2 moral beliefs and the epistemic consequences
3 of those facts for moral beliefs:
4 Partiality: “Because people’s moral beliefs af-
5 fect their self-interest so often in so many ways
6 at least indirectly, and because people are so
7 bad at telling when their own beliefs are partial,
8 there is a presumption that moral beliefs are
9 partial” (2006b, p. 197). Self-interest, of course,
10 is not correlated with moral truth.
11 Bias: Psychologists have identified various
12 cognitive illusions and framing effects that
13 influence moral belief. “Moral beliefs that vary
14 in response to factors that do not affect truth—
15 such as wording and belief context—cannot
16 reliably track the truth” (2008a, p. 54).
17 Emotions: Research from psychology suggests
18 moral judgment is often affect-laden or driven
19 by the brain’s emotion centers. “[M]any moral
20 judgments result from emotions that cloud
21 judgment” (2006a, p. 352).
22 Disagreement: There is considerable disagree-
23 ment about morality. “[I]f we know that many
24 moral intuitions are unreliable because others
25 hold conflicting intuitions, then we are not
26 justified in trusting a particular moral intuition
without some reason to believe that it is one of
27
the reliable ones” (2006a, p. 350).
28
29 Sinnott-Armstrong takes it for granted
30 that we, his readership, are well aware of the
31 above facts and their consequences. These
32 facts, he says, imply that “a large percent-
33 age” of moral beliefs are false (2012, p. 16).
34 As a result, we have evidence that our moral
35 beliefs, taken as an entire class, are not reli-
36 ably formed (2008a, pp. 51–52). For present
37 purposes, we may assume all of this is so.
38 That is Step One of the EDA.4
39 Step Two attempts to show that an impor-
40 tant epistemological conclusion follows from
41 Step One. Sinnott-Armstrong’s idea is that if
42 we are aware of those facts about moral be-
43 lief and their consequences, then we should
44 recognize that none of our moral beliefs are
45 non-inferentially justified, and so that moral
46 intuitionism is false. Although Sinnott-Arm-
strong’s defense of this conditional claim has
changed as he has refined his argument, the
basic idea has remained the same.
It is useful, we think, to state Step Two of the
EDA in terms of undercutting defeaters—that
is, reasons to give up belief in some proposition
that are consistent with that proposition being
true (see Pollock 1986). Now suppose that
we accept Step One: on the basis of the facts
Sinnott-Armstrong calls attention to, we think
a large percentage of moral beliefs are false. As
soon as we recognize that any particular moral
belief of ours, B, is a moral belief, we then have
an undercutting defeater for B. That’s because
we would have a reason to think members of
the class of moral beliefs, such as B, are to a
large degree likely to be false. Thus, if there is
no defeater for the undercutting defeater here,
B is not justified.5
Sinnott-Armstrong grants that we might
gain a defeater for the undercutting defeater
for B. His proposal is that the defeater for B is
itself defeated only if we have reason to think
B falls into a subclass of sufficiently reliably
formed moral beliefs. Sinnott-Armstrong
offers the following example in support of
his proposal (see 2008a, p. 51). Imagine that
you look at a figure from afar and come to
believe that she or he is shorter than six feet
tall (call this belief “S”). Then you remember
that your beliefs concerning the height of
objects at a distance are unreliable. Indeed,
many of them are false. Your awareness that
your height beliefs at a distance are unreliable
constitutes an undercutting defeater for S. In
order to restore justification for S, Sinnott-
Armstrong suggests, you need to find some
positive and undefeated reason for believing
that S is sufficiently reliably formed; you
might gain this by walking closer to the figure
or asking a trustworthy source.
Importantly, Sinnott-Armstrong insists that
whenever we have a defeater for the defeater
for a belief B—again, a reason to think B is
a member of a subclass of reliably formed
moral beliefs—we’re thereby committed
 MORAL INTUITIONISM DEFEATED? / 413

to what he calls a “justificatory inferential necessary, in defeating a defeater for B, to be 1

structure.” He defines that term of art as committed to a justificatory inferential struc- 2
a set of propositions where some propositions ture with respect to B. An example, followed 3
provide epistemic support for others. To be com- by subsequent discussion, will bear this out: 4
mitted to such a structure is to accept or have McCoy visits the local widget factory and sees 5
a disposition to accept the supporting proposi- what seems to be a red widget being carried along 6
tions in that structure or other propositions that a conveyor belt. He believes that the widget is 7
entail or support the supporting propositions in red (call this belief “B”). Soon enough, a stranger 8
that structure. (2012, p. 13) approaches McCoy and says that the widgets are 9
So, whenever we’re committed to a justifica- actually white but are illuminated by red lights. 10
tory inferential structure with respect to B, we (Call this event “D”.) Upon seeing this conversa- 11
tion, another stranger—who seems to McCoy to 12
accept or are disposed to accept propositions
be a factory employee—tells McCoy not to listen 13
that entail or rationally support B. Thus, if we to the other stranger: he is a trickster, McCoy is
have a defeater for the defeater for B, we’re 14
told, who likes to mess around with visitors. (Call
in a position to infer B from other beliefs 15
this second event “F”.)
of ours. But if we are positioned to infer B 16
from other beliefs—and if we must be so This example will support our contention 17
positioned in order to defeat the defeater to B that Step Two features a false assumption.6 18
and to remain justified in believing B—then It is but one example. There are others, with 19
B is not non-inferentially justified. slightly different details, and we think such 20
In the end, Step Two of the EDA comes to examples are commonplace for a whole host 21
this: after having taken Step One, we have a of non-inferential sources of justification.7 22
defeater for each one of our moral beliefs. Let’s consider B, D, and F, in order to show 23
That defeater is itself either undefeated or how they undermine Step Two. 24
defeated. If it is undefeated, our moral beliefs McCoy forms a belief, B, in a basic way. 25
are not justified. If it is defeated, our moral be- His belief is based on a perceptual experience 26
liefs are not non-inferentially justified. Sup- or seeming. Given that Sinnott-Armstrong’s 27
posing the argument is successful, it shows target, moral intuitionism, is a foundationalist 28
that any moral belief is either unjustified theory, let’s assume foundationalism concern- 29
or not non-inferentially justified. It follows ing perceptual justification in the example. 30
that moral intuitionism—the view on which So, B is initially non-inferentially justified 31
“some people are adequately epistemically for McCoy in some way (we remain neutral 32
justified in holding some moral beliefs inde- on the details) by the appropriate experiences 33
pendently of whether those people are able or seemings. 34
to infer those moral beliefs from any other D is a defeater for B. Hearing from another 35
beliefs” (2006a, p. 341)—is false. person that the widget is white but illuminated 36
with red lights throws enough doubt on Mc- 37
2. The Trouble with the Coy’s belief B to defeat its initial justification. 38
Empirical Defeat Argument We remain neutral on how this testimony 39
defeats. Perhaps testimony directly justifies 40
We think Step Two of Sinnott-Armstrong’s
(absent defeaters), or perhaps McCoy has 41
EDA has a fatal, but so far unrecognized,
good reasons to trust people in general and the 42
defect. Here is the trouble: it is not neces-
testifier doesn’t seem untrustworthy to him. 43
sary, in defeating a defeater for a belief B, to
F is a defeater for D. F is what is sometimes 44
have a reason to think that B was formed in
called a “defeater-defeater” (see Pollock 45
a sufficiently reliable way. Thus, it also isn’t
1987). The factory employee’s testimony 46
 414 / AMERICAN PHILOSOPHICAL QUARTERLY
1 undermines the evidential force of the other
2 person’s testimony. Since F defeats D, D no
3 longer defeats the initial justification for Mc-
4 Coy’s belief B. McCoy is, therefore, justified
5 in B upon learning F.
6 Our example has the following pattern: B is
7 initially non-inferentially justified, B is then
8 defeated by D before D is later defeated by
9 F, and, as a result, B is justified once again.
10 According to Sinnott-Armstrong, in order
11 for F to defeat D and restore justification to
12 B, F must provide reason to think that B was
13 formed reliably and, as a result, McCoy must
14 be committed to a justificatory inferential
15 structure in support of B. Note well: that is
16 not what we find in the example. The factory
17 employee’s testimony that the stranger is a
18 trickster doesn’t imply that McCoy’s belief
19 that the widget is red was reliably formed.
20 Accordingly, McCoy is not committed to a
21 justificatory inferential structure in support
22 of B by way of obtaining a defeater for D.
23 All of that said, we grant that McCoy might
24 be able to construct an argument for the con-
25 clusion that B is reliably formed—perhaps,
26 for example, he can give a track-record ar-
27 gument. But that ability is irrelevant to our
28 contention: his being able to come up with
29 such an argument is not required to defeat D.
30 Having F is by itself sufficient to defeat D,
31 independently of his ability to give arguments
32 in favor of B.
33 Therefore, contrary to Sinnott-Armstrong’s
34 argument in Step Two of the EDA, it is not
35 necessary that a thinker defeat a defeater for
36 a non-inferentially justified belief B by way
37 of a reason for believing that B was reliably
38 formed and thus committing one to a justi-
39 ficatory inferential structure supporting B.
40 A defeater-defeater can do its job merely by
41 showing that the defeater of B has no or neg-
42 ligible evidential force. That can be accom-
43 plished without giving evidence for thinking
44 B is true or that B was reliably formed, as with
45 our example. Furthermore, once the defeater
46 is defeated, B is clearly non-inferentially
justified. Whatever initially made B justified
continues to justify B once the defeater has
been defeated. McCoy’s belief that the widget
is red is justified by his perceptual experience
or seeming (perhaps together with the fact
that his perception is reliable)—even after D
has been defeated by F.
We do not deny that McCoy is commit-
ted to some inferential structure. Indeed,
he’s committed to the following inference:
F, F defeats D, hence D is defeated. And
since McCoy is aware of D, he wouldn’t
be justified in B unless he was aware of F
and recognized that F defeats D.8 Does this
mean B is inferentially justified? Not at all.
It’s important to distinguish between B’s jus-
tification depending upon an inference and
B being supported (and thus justified) by an
inference. McCoy’s belief B has the former
feature, but not the latter, though it is the
latter that Sinnott-Armstrong claims results
from defeating his empirical defeaters.
We say that B is justified only if McCoy
is positioned to make the inference we
noted: F, F defeats D, hence D is defeated.
In that sense, the justification for B depends
upon an inference. Yet, plainly B is not sup-
ported by that inference. B is supported by
experiences or seemings alone, without any
inference to B from reasons indicating that
B was reliably formed. These experiences,
all by themselves,9 without an inference,
produce justification for B, whereas F, the
defeater-defeater, preserves the justification
that the experiences produce for B. It’s like
this analogy: a spark on dry wood produces
a fire while the tarp covering the fire pit pre-
serves the fire by shielding it from the rain.
The tarp doesn’t create or sustain the fire, of
course, but it preserves the conditions for the
fire’s sustenance. The production of justifica-
tion for B is non-inferential even though the
preservation of B’s justification is inferential.
As it happens, a central point of dispute
between moral intuitionists of various stripes
and non-intuitionists is whether justification
 MORAL INTUITIONISM DEFEATED? / 415

for moral beliefs is (or can be, or often is) pro- not show that production of justification must 1
duced non-inferentially. Almost any theory of be inferential. 2
justification for moral beliefs, and non-moral Before concluding, let us consider two 3
beliefs alike, will be committed to saying that objections to our argument. 4
the preservation of justification is inferential: 5
for facing and evading defeaters is an everyday 3. Two Objections 6
occurrence for all sorts of beliefs. To see why, Consider this reply to our argument: “An 7
consider the following defeater for any and all undercutting defeater for a belief B, such as 8
of our beliefs: we often encounter defeaters for the empirical defeaters presented by Sinnott- 9
our beliefs. We know that we often encounter Armstrong, raises a question about B at the 10
defeaters for our beliefs because we often higher epistemic level: Is B justified? If you 11
form beliefs in poor conditions, such as when have no reason to believe that your belief is 12
we are exhausted, or looking too hastily, or justified, and you recognize as much, then 13
multi-tasking, or relying on biased informa- you won’t in fact be justified in believing B. 14
tion. That defeater is defeated for most of us The higher-level question must be resolved, 15
and for most of our current beliefs by simply since the defeater pushes the issue to that 16
avoiding known sources of defeat. We don’t level. But if you have reason to believe that 17
normally do any of this consciously, but we B is justified, despite the defeater, then you’ll 18
do it nonetheless. Here is an example. You be in a position to infer B from the belief that 19
automatically look closely at a book’s spine to you are justified in believing B. Thus, when 20
make out the title, rather than quickly glanc- you are faced with an undercutting defeater, 21
ing. If you had instead glanced hurriedly and in order to defeat the defeater, you will be 22
later remembered this, you might have gotten committed to an inferential structure sup- 23
a defeater for your belief about the title. But porting B.” 24
since you recall that you took a careful look in This won’t do. The objection mistakenly 25
this instance, the “I looked too hastily” defeater assumes that the higher-level question of 26
gets defeated. Your justification for your belief whether B is justified must be resolved in 27
is preserved by inference, even though that order for B to be justified. Let’s say that the 28
justification was produced non-inferentially. true epistemic principle has the following 29
Saying that the preservation of justification form: a belief B is justified if and only if con- 30
for our moral beliefs is inferential is not say- dition C is met and there are no defeaters for 31
ing much—the same goes for nearly any kind B. Then suppose that C is met. To illustrate, 32
of belief, no matter our favored theory of how we can suppose that C is that it seems to the 33
justification is produced. thinker that B is true. Then along comes a 34
Remember that the critical dispute, and defeater. As a result, the principle implies 35
the one under investigation here, is whether that B is no longer justified. Now a defeater 36
the production of justification for our moral for that defeater arrives. By the principle, B 37
beliefs is at least sometimes non-inferential. is justified for the thinker once again. Over 38
That’s the dispute that the EDA is designed and above that, the thinker needn’t be justified 39
to settle. Sinnott-Armstrong aims to show in believing that B is justified. Arguably, if 40
that moral beliefs need confirmation in light the thinker had serious doubts about whether 41
of the defeaters he presents, and he thinks B was justified, and could not resolve those 42
that such confirmation requires the ability doubts, then B would lose its justification (see 43
to infer that some subset of moral beliefs is Bergmann 2005). But suppose she lacks any 44
true (see 2006b, pp. 212–213). But, as we’ve such doubts. She has what it takes to be justi- 45
argued, Sinnott-Armstrong’s argument does fied and defeats the defeaters she faces. The 46
 416 / AMERICAN PHILOSOPHICAL QUARTERLY

1 epistemic principle implies that B is justified the truth or justification of the empirical
2 for her. defeaters, then the only way to defeat them
3 Turn to a second objection. Our complaint is to obtain evidence that moral beliefs are
4 with the EDA implies that defeaters can be reliably produced.” This reply runs into two
5 defeated without furnishing evidence to think problems. First, it dramatically restricts the
6 that the original belief was reliably formed. significance of the EDA’s conclusion. After
7 To be sure, sometimes particular defeaters are all, we’ve noted there is another way to defeat
8 defeated by evidence that the original belief the defeaters—namely, by questioning the
9 was reliably produced. With that in mind, truth or justification of the defeaters. The re-
10 one might attempt to rehabilitate the EDA ply just sweeps that possibility aside, despite
11 by contending that the fact that further reflection and evidence
12 RP A thinker can defeat the empirical defeaters could cast doubt on the truth or justification
13 (namely, Partiality, Bias, Emotions, Dis- of the defeaters. Again, questioning the truth
14 agreement; see section 1) only if the thinker or justification of the defeaters certainly
15 has reason to think that moral beliefs are seems to allow moral beliefs to remain non-
16 reliably produced. inferentially justified.
17 We doubt that RP can be sustained, and we’ll There’s a second problem for the reply. The
18 now establish the point. empirical defeaters work by first identifying
19 Sinnott-Armstrong’s empirical defeaters features of moral beliefs that apparently sug-
20 may be defeated by simply gaining evidence gest the beliefs aren’t reliably formed, and
21 that some of the defeaters are false. Thus, we then arguing that most moral beliefs have
22 don’t need reason to think our moral beliefs these features. Upon becoming aware of the
23 are reliably produced in order to defeat the defeaters, we thereby have reason to believe,
24 empirical defeaters. To see why, notice that of any moral belief, that it likely has these
25 each of those defeaters has the following form: features, and so is likely to be unreliably
26 empirical information E shows that moral be- formed. But if we learned that some subset
27 liefs in general are not reliably produced. We of our moral beliefs lacks those features,
28 can have evidence that such a defeater is false then we wouldn’t have the defeaters for those
29 if we have reason to think either that E is false beliefs. Suppose the beliefs in that subset
30 or that E does not show that moral beliefs in were non-inferentially justified before we
31 general are not reliably produced. For instance, learned about the empirical defeaters. Since,
32 someone could gain reason to think that, for by hypothesis, those beliefs don’t have the
33 some classes of beliefs, Partiality and Bias features that generate defeat (and we realize
34 aren’t significant enough to show that those this), they are not defeated and, thus, they
35 beliefs are not reliably produced. Someone would remain non-inferentially justified. In
36 might also argue that the information described this kind of situation, the empirical defeaters
37 earlier in Emotions does not show that moral are defeated by being avoided, and they’re
38 beliefs are not reliably produced.10 Since the avoided by obtaining evidence that the defeat-
39 other empirical defeaters can also be defeated ing features aren’t possessed by a subset of
40 in this way, it’s highly doubtful that RP is true: our moral beliefs. Furthermore, notice that
41 Sinnott-Armstrong’s empirical defeaters can the defeater is defeated without invoking
42 be defeated without reason to think that moral evidence that the beliefs in the subset are
43 beliefs are reliably produced. reliably formed. So, again, RP is false.
44 One might reply as follows: “If we don’t This is worth briefly illustrating. The “Dis-
45 question (or if we lack reason to question) agreement” defeater can be avoided if we
46
 MORAL INTUITIONISM DEFEATED? / 417

check whether there is considerable disagree- good reason to believe lack the defeating fea- 1
ment over a given moral belief, B, and then tures, the beliefs in M would still be defeated. 2
learn there isn’t.11 The idea here isn’t that we The idea might be that Sinnott-Armstrong’s 3
use widespread agreement as evidence that B empirical defeaters make salient the follow- 4
is true or that B was reliably produced, even ing possibility: that M’s members (or many 5
if we could run an argument to that effect. In- of them) do possess some of the defeating 6
stead, the idea is that we position ourselves so features, or perhaps some other defeating 7
that the defeater doesn’t apply. Similarly, the features, but we haven’t so far recognized 8
“Bias” defeater is harmless when we’ve got this. The idea is that, given we have possibly 9
reason to think we are not in a situation that failed to recognize M’s members have defeat- 10
typically leads to biases. Likewise for defeat- ing features, that’s enough for the members 11
ers posed by Partiality and Emotions—if we of M to be defeated. 12
have reason to think that a subset of our moral Does this suggestion save the EDA? Ac- 13
beliefs isn’t subject to any defeating kinds of cording to us, no. That mere possibility— 14
partiality and emotional influence, then the again, that M’s members possess defeating 15
defeaters are avoided for that subset.12 features we haven’t yet recognized—is not 16
It could turn out that only a few of our significant enough to fully defeat the beliefs 17
moral beliefs avoid defeaters by way of in M. At most, we’d have a partial defeater. 18
the strategy described above. If that’s how The mere possibility is too distant to do 19
things go, one could argue that the empirical enough damage. By way of explanation here, 20
defeaters still defeat those few beliefs, albeit allow us to compare the present suggestion 21
in a less direct way. Suppose the subset of for salvaging the empirical defeaters to an 22
beliefs that we have reason to think lack the argument from Sextus Empiricus’s Outlines 23
defeating features is small in number. Perhaps of Scepticism (Bk. I.xiii). Our reply to the 24
the empirical defeaters could still challenge suggestion parallels what we regard as the 25
members of the subset because the defeaters sensible response to Sextus’s argument. 26
would provide reason to think that the sub- Writes Sextus: 27
set’s members have some of the defeating [W]hen someone propounds to us an argument 28
features, but that we haven’t yet been able to we cannot refute, we say to him: “Before the 29
recognize or detect those features. founder of the school to which you adhere was 30
We agree that if the subset of beliefs that born, the argument of the school, which is no 31
we have reason to think lack the defeating doubt sound, was not yet apparent, although it 32
features were small in number, then its mem- was really there in nature. In the same way, it 33
bers might be defeated in this way. However, is possible that the argument opposing the one 34
if the subset were large enough and unified in you have just propounded is really there in na- 35
various ways (e.g., the beliefs concern certain ture but is not yet apparent to us; so we should 36
not yet assent to what is now thought to be a
types of actions or motives), then it seems 37
powerful argument.” (2000, p. 12)13
that the defeaters would be avoided. For all 38
we know, there is a large enough unified set Let’s say that someone offers us an apparently 39
of beliefs that lack the defeating features. strong argument. We know that there are often 40
If so—and we had reason to think it were replies to strong arguments. We think, and 41
so—then beliefs in that set would remain then think some more, but don’t come up with 42
non-inferentially justified. a good reply. To be sure, there may be a good 43
One might retort that even if there is a large, reply that we haven’t noticed. Recognizing 44
unified set of moral beliefs, M, that we have this possibility is enough, Sextus claims, to 45
46
 418 / AMERICAN PHILOSOPHICAL QUARTERLY
1 defeat our belief that the original argument
2 is strong. Sextus is correct: this possibility
3 is salient, just as the possibility noted above
4 concerning the empirical defeaters is salient.
5 But, normally, these possibilities seem too
6 distant to fully defeat our beliefs. In both
7 cases, we may find ourselves with a partial
8 defeater, but this merely lowers the degree
9 of justification for our belief. So, even if the
10 justification for our M beliefs is diminished,
11 our justification remains non-inferential (on
12 the assumption that our beliefs in M were
13 originally non-inferentially justified, which
14 Step One of the EDA grants for argument’s
15 sake). It is just that whatever justifies our
16 belief doesn’t justify as much as it would
17 have if the partial defeater were not present.
18 Let’s take account of our argument against
19 RP. We’ve said that RP is false for a pair of
20 reasons: (i) we could gain evidence that the
21 defeaters are false without having reason to
22 think that some subset of our moral beliefs
23 is reliably formed, and (ii) we could discover
24 that a large, unified subset of moral beliefs
25 M doesn’t possess the defeating features, and
26 thus is not subject to the defeater. Then we
27 noted that, given (ii), the beliefs in M may be
28 partially defeated, but they would still remain
29 non-inferentially justified.
30 But even if we’re wrong and RP is true,
31 the EDA still fails. According to RP, we need
32 reason to think that our moral beliefs are reli-
33 ably produced in order to defeat the defeaters.
34 Suppose that we come by such a reason, R.
35 This reason defeats the defeaters and commits
36 us to a justificatory inferential structure. R
37 provides evidence for our moral beliefs, in
38 virtue of the fact that the moral beliefs can be
39 inferred from R (together with other informa-
40 tion). So, our moral beliefs are justified, in
41 part, inferentially. From this, it doesn’t follow
42 that they are not justified non-inferentially.
43 That’s because the epistemic status of beliefs
44 can be overdetermined, and our moral be-
45 liefs are often epistemically overdetermined
46 in this kind of situation. We’re supposing,
again, that some of our moral beliefs were
originally justified non-inferentially by some
condition C (their seeming true, say). Sinnott-
Armstrong’s empirical defeaters defeated the
justification C provides, but once we get R,
the defeaters are defeated, and C can once
again justify our moral beliefs directly. And
so our moral beliefs are justified inferentially
by R and non-inferentially by C. Since those
moral beliefs are still justified—in part, non-
inferentially—moral intuitionism is not de-
feated. True, our beliefs would not be justified
without R in this situation, but this doesn’t
mean that R is the sole source of justifica-
tion production for our moral beliefs. R does
double duty, both producing and preserving
justification: it produces justification for our
moral beliefs inferentially and preserves the
justification that C provides.
4. Conclusion
Sinnott-Armstrong has developed a novel
argument against moral intuitionism. We’ve
argued that his Empirical Defeat Argument
features a mistaken assumption about what
is required to defeat a defeater. We’ve also
argued that various fixes will not save the
argument. Nevertheless, we believe that the
general project of examining the possible
epistemic implications of empirical research
on judgment and reasoning is important. The
project has the potential to yield insights,
and we commend Sinnott-Armstrong for
moving it forward. Sinnott-Armstrong writes
that “philosophers cannot continue to ignore
developments in psychology, brain science,
and biology.” He adds a word of caution: “Of
course, philosophers need to be careful when
they draw lessons from empirical research”
(2006a, p. 339). We couldn’t agree more.
Nathan Ballantyne,
Fordham University
Joshua C. Thurow,
University of Texas at San Antonio
 MORAL INTUITIONISM DEFEATED? / 419

NOTES 1
2
For helpful comments on earlier versions of this essay, we are grateful to Michael Bergmann, Thomas 3
Crisp, William Dyer, John Greco, Russ Shafer-Landau, Mark Timmons, and several anonymous referees.
4
We owe gratitude to Walter Sinnott-Armstrong for a lively presentation of his argument and subsequent
discussion at the Purdue Summer Seminar on Perceptual, Moral, and Religious Skepticism in June 5
2011. We also thank an audience at the 2012 American Philosophical Association Pacific meeting, and 6
Brian Talbot, our commentator. 7
8
1. This paper is the product of full and equal collaboration between its authors.
9
2. See Sinnott-Armstrong (2006a), (2006b), pp. 184–220, (2008a), (2012). For discussion of the 10
argument, see Shafer-Landau (2008); Tolhurst (2008); Joyce (2009), pp. 217–219; Bedke (2010), pp. 11
1077–1080; Smith (2010); Tropman (2011); and Littlejohn (2012), pp. 108–110. Nadelhoffer and Feltz 12
(2008) present additional empirical evidence that supports one of the argument’s premises. Sinnott-
13
Armstrong (2008b) replies to Shafer-Landau (2008); and Tolhurst (2008). Sinnott-Armstrong’s argument
has enjoyed a wide influence within recent debates in moral psychology and meta-ethics. (According to
14
Google Scholar’s count, as of March 2013, two of the papers in which he develops the argument, 2006a 15
and 2008a, have been cited a total of nearly one hundred times.) For recent discussion and defense of 16
moral intuitionism, see Audi (1996, 2004); and Huemer (2005, 2008); and papers in Hernandez (2012). 17
18
3. Ross (1930); Prichard (1957); Audi (1996, 2004); and Huemer (2005, 2008).
19
4. Step One has been critically discussed by Bedke (2010); and Smith (2010). 20
5. As we’ve noted, the details of Step Two have changed since Sinnott-Armstrong’s first statement 21
of his argument. In Moral Skepticisms (2006b), what we’ve called Step Two is captured in principles 22
1–5 of section 9.3 and his discussion in section 9.5.2; the argument in “Moral Intuitionism Meets 23
Moral Psychology” (2006a) is basically identical. We find Step Two in premise (1) in “Framing Moral 24
Intuitions” (2008a), and premise (3) in “An Empirical Challenge to Moral Intuitionism” (2012). Our 25
statement of Step Two expresses the core features of Sinnott-Armstrong’s statements of the argument.
26
6. The “widget factory” example—not counting what is introduced by F—is standard fare in episte- 27
mology these days, but was originally devised by the late John Pollock. 28
7. Consider another example that features memory as the source of non-inferential justification. Imagine 29
that today is Madeleine’s wedding anniversary: she’s with her husband and mother, fondly reminisc- 30
ing about the special day eleven years earlier. Madeleine believes that she had a wild dance with her 31
unpredictable and somewhat creepy uncle Rex (call this belief “B”). Laughing, she asks her husband 32
if he remembers the dance with Rex. “No, honey,” says her husband, “Rex wasn’t there. Remember? 33
He didn’t make it—he crashed his pickup on the road from Vegas, after yet another gambling binge. 34
‘What happens in Vegas stays in Vegas’ . . . not quite.” (Call Madeleine’s husband’s statement “D”.)
35
Then Madeleine’s mother pipes in: “Oh, you’re thinking of another wedding—Rex didn’t make it to
36
Madeleine’s sister’s wedding!” (Call the mother’s statement “F”.) Each of the claims we make about
the “widget factory” example hold for this example too. 37
38
8. This point doesn’t pre-suppose McCoy would need to use the terminology of defeat in order for F 39
to help him evade D for B.
40
9. Externalist theories might require that the process of believing based on those experiences be reli- 41
able or in accord with a thinker’s proper function. 42
10. See, for example, Smith (2010) for such arguments. 43
11. What sort of belief do we have in mind? Shafer-Landau proposes this (“R-rated”) belief: that 44
“the deliberate humiliation, rape, and torture of a child, for no purpose other than the pleasure of the 45
46
 420 / AMERICAN PHILOSOPHICAL QUARTERLY

1 one inflicting such treatment, is immoral” (2008), p. 83. We say that’s a plausible example of a non-
2 inferentially justified moral belief that would survive a check for disagreement.
3 12. Huemer’s (2008) “revisionary intuitionism” recommends, among other things, avoiding disagree-
4 ment and checking for potential sources of bias as a method to screen out problematic intuitions. For
5 all that, Huemer seems to—and should, if our argument is correct—regard moral beliefs based on
6 intuitions that survive such a screening as non-inferentially justified.
7 13. See Machuca (2011) for more on Sextus’s argument from possible disagreement.
8
9
10
11 REFERENCES
12 Audi, Robert. 1996. “Intuitionism, Pluralism, and the Foundations of Ethics,” in Moral Knowledge, ed.
13 Walter Sinnott-Armstrong and Mark Timmons (New York: Oxford University Press), pp. 101–136.
14 ———. 2004. The Good in the Right (Princeton, NJ: Princeton University Press).
15 Bedke, Matt. 2010. “Intuitional Epistemology in Ethics,” Philosophy Compass, vol. 5, no. 12, pp.
16 1069–1083.
17 Bergmann, Michael. 2005. “Defeaters and Higher-Level Requirements,” Philosophical Quarterly, vol.
18 55, no. 220, pp. 419–436.
19 Hernandez, Jill Graper, ed. 2012. The New Intuitionism (New York: Continuum Press).
Huemer, Michael. 2005. Ethical Intuitionism (New York: Palgrave Macmillan).
20
———. 2008. “Revisionary Intuitionism,” Social Philosophy and Policy, vol. 25, no. 1, pp. 368–392.
21
Joyce, Richard. 2009. “The Skeptick’s Tale,” Philosophy and Phenomenological Research, vol. 78,
22 no. 1, pp. 213–221.
23 Littlejohn, Clayton. 2012. “Ethical Intuitionism and Moral Skepticism,” in The New Intuitionism, ed.
24 Jill Graper Hernandez (New York: Continuum Press), pp. 106–127.
25 Machuca, Diego. 2011. “The Pyrrhonian Argument from Possible Disagreement,” Archiv für Geschichte
26 der Philosophie, vol. 93, no. 2, pp. 148–161.
27 Nadelhoffer, Thomas, and Adam Feltz. 2008. “The Actor-Observer Bias and Moral Intuitions: Adding
28 Fuel to Sinnott-Armstrong’s Fire,” Neuroethics, vol. 1, no. 2, pp. 133–144.
29 Pollock, John. 1986. Contemporary Theories of Knowledge (Lanham, MD: Rowman & Littlefield).
30 ———. 1987. “Defeasible Reasoning,” Cognitive Science, vol. 11, no. 4, pp. 481–518.
31 Prichard, H. A. 1957. Moral Obligation (Oxford: Clarendon Press).
Ross, W. D. 1930. The Right and the Good (Oxford: Clarendon Press).
32
Sextus Empiricus. 2000. Outlines of Scepticism, ed. Julia Annas and Jonathan Barnes (Cambridge:
33
Cambridge University Press).
34 Shafer-Landau, Russ. 2008. “Defending Ethical Intuitionism,” in Moral Psychology, Vol. 2, ed. Walter
35 Sinnott-Armstrong (Cambridge, MA: MIT Press), pp. 83–95.
36 Sinnott-Armstrong, Walter. 2006a. “Moral Intuitionism Meets Moral Psychology,” in Metaethics After
37 Moore, ed. Terry Horgan and Mark Timmons (New York: Oxford University Press), pp. 339–365.
38 ———. 2006b. Moral Skepticisms (New York: Oxford University Press).
39 ———. 2008a. “Framing Moral Intuitions,” in Moral Psychology, Vol. 2, ed. Walter Sinnott-Armstrong
40 (Cambridge, MA: MIT Press), pp. 47–76.
41 ———. 2008b. “How to Apply Generalities: Reply to Tolhurst and Shafer-Landau,” in Moral Psychol-
42 ogy, Vol. 2, ed. Walter Sinnott-Armstrong (Cambridge, MA: MIT Press), pp. 97–105.
43 ———. 2012. “An Empirical Challenge to Moral Intuitionism,” in The New Intuitionism, ed. Jill Graper
Hernandez (London: Continuum Press), pp. 11–28.
44
Smith, Jonathan. 2010. “On Sinnott-Armstrong’s Case against Moral Intuitionism,” Ethical Theory and
45
Moral Practice, vol. 13, no. 1, pp. 75–88.
46
 MORAL INTUITIONISM DEFEATED? / 421

Tolhurst, William. 2008. “Moral Intuitions Framed,” in Moral Psychology, Vol. 2, ed. Walter Sinnott- 1
Armstrong (Cambridge, MA: MIT Press), pp. 77–82. 2
Tropman, Elizabeth. 2011. “Non-Inferential Moral Knowledge,” Acta Analytica, vol. 26, no. 4, pp. 3
355–366. 4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46