See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257667024

Armchair methodology and epistemological naturalism

Article  in  Synthese · December 2013

DOI: 10.1007/s11229-013-0253-9

CITATIONS READS

2 503

1 author:

Janet Levin
University of Southern California
28 PUBLICATIONS   301 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Janet Levin on 15 March 2017.

The user has requested enhancement of the downloaded file.

Synthese (2013) 190:4117–4136
DOI 10.1007/s11229-013-0253-9

Armchair methodology and epistemological naturalism

Janet Levin

Received: 1 January 2013 / Accepted: 12 February 2013 / Published online: 2 March 2013
© Springer Science+Business Media Dordrecht 2013

Abstract In traditional armchair methodology, philosophers attempt to challenge a

thesis of the form ‘F iff G’ or ‘F only if G’ by describing a scenario that elicits the
intuition that what has been described is an F that isn’t G. If they succeed, then the judg-
ment that there is, or could be, an F that is not G counts as good prima facie evidence
against the target thesis. Moreover, if these intuitions remain compelling after further
(good faith) reflection, then traditional armchair methodology takes the judgment to
be serious (though not infallible) evidence against the target thesis—call it secunda
facie evidence—that should not be discounted as long as those intuitions retain their
force. Some philosophers, however, suggest that this methodology is incompatible
with epistemological naturalism, the view that philosophical inquiry should be sensi-
tive to empirical observations, and argue that traditional armchair methodology must
deemphasize the role of intuitions in philosophical inquiry. In my view, however, this
would be a mistake: as I will argue, the most effective way to promote philosophical
progress is to treat intuitions as having the (prima and secunda) evidential status I’ve
described. But I will also argue that philosophical inquiry can produce a theory that is
sensitive to empirical observations and the growth of empirical knowledge, even if it
gives intuitions the prima- and secunda-facie evidential status that traditional armchair
methodology demands—and thus that traditional armchair methodology, if properly
practiced, need not be abandoned by naturalists, or even (except for a few exceptions)
be much revised.

Keywords Intuitions · Armchair philosophy · Naturalism · Thought

experiments · Philosophical methodology

J. Levin (B)
University of Southern California, 3709 Trousdale Parkway, Los Angeles, CA 90089-0451, USA
e-mail: levin@usc.edu

123
4118 Synthese (2013) 190:4117–4136

1 Introduction

I am sympathetic to ‘epistemological’ (or ‘methodological’) naturalism: the view

that philosophy, like the sciences, should ‘investigate reality’ and that its conclusions
should be sensitive to empirical observations and the growth of empirical knowledge.1
But I’m also sympathetic to (and have defended)2 certain traditional views about the
methodology of philosophy—call it ‘armchair methodology’—that may seem incom-
patible with naturalism of this sort. My primary goal in this paper is to see how far
these views can be reconciled.
Just to be clear: what I mean by ‘armchair methodology’ is the method by which
philosophers who want to challenge a thesis of the form ‘F iff G’ or ‘F only if G’
attempt to describe a scenario that elicits the intuition (call it, following Bealer (1996),
a ‘concrete-case intuition’) that what has been described is an F that isn’t G. If the
endeavor is successful, then the judgment that there is, or could be, an F that is not G
counts as good prima facie evidence against the target thesis. Familiar examples are
Gettier cases, Frankfurt cases, trolley cases, Twin Earth cases, and brain transplant
cases, constructed to challenge philosophical theses about the nature of (respectively)
knowledge, moral responsibility, moral action, meaning, and personal identity.
To be sure, prima facie evidence against a thesis is not definitive. Sometimes, upon
further reflection (perhaps prompted by defenders of the thesis), these initial intuitions
fade, either because the scenario described no longer seems to be an instance of an F,
or, while seeming to be an F, also seems to be a G. It is only when the initial intuitions
remain undiminished after further reflection that those ‘F but not G’ judgments count
as more than prima facie evidence against the thesis. This should be ‘good faith’
reflection, of course, with the aim of finding, rather than ignoring, confounding factors
that may have illegitimately influenced one’s judgment. But if these judgments retain
their intuitive force after further (good faith) reflection on the particulars of the case,
other intuition-based judgments, established philosophical principles (and also the
empirical facts, when the subject of inquiry is an a posteriori claim about identity
or essence), then armchair methodology takes the judgment to be serious (though
not infallible) evidence against the target thesis—call it secunda facie evidence—that
should not be discounted as long as those intuitions retain their force.3

1 Michael Devitt (2010a, Ch. 12, p. 252) contrasts ‘epistemological’ with ‘metaphysical’ naturalism,
and David Papineau (2009) contrasts ‘methodological’ with ‘ontological’ naturalism. Papineau’s particular
version (2009, p. 1) holds that
(1) Philosophical claims are synthetic, not analytic.
(2) Philosophical knowledge is a posteriori, not a priori.
(3) Central questions of philosophy concern actuality, not necessity; they are aimed at understanding the
world, not the realm of metaphysical modality (though he does admit (p. 27) that modal reflection can
help us determine whether claims about such things as identity and constitution are true in the actual
world).
As I’ll argue, however, there are theories other than Papineau’s that (more or less) meet these conditions.
2 See my (2004), (2009), (2011).
3 In my view, however, there are a (very) small number of exceptions, e..g. claims involving phenomenal
concepts (and perhaps—though this is less plausible than one might imagine—other sorts of demonstra-
tives). I have argued for this elsewhere (e.g. 2004, 2011) and will not discuss this question further here.

123
Synthese (2013) 190:4117–4136 4119

This is merely a sketch of the method; there are many details that need to be filled
in. For example, although I characterize the prima and secunda facie evidence for
philosophical theses as judgments based on concrete-case intuitions, I have not spec-
ified whether the intuitions themselves should be identified with those judgments, or
instead with their non-conceptual grounds.4 However, though this distinction is cru-
cial for a full account of the nature of intuitions, my characterization of the role of
intuition-based judgments in philosophical inquiry is compatible with either view. One
may also want more details about how a principled distinction can be made between
intuitions about category membership elicited by a thought-experiment—whatever
their nature—and other spontaneous reactions to the description of the scenario in
question, such as revulsion or other emotional responses, since without such a distinc-
tion, it would be hard to give armchair inquiry much credence. However, though I will
say more later (in Sects. 2 and 4) about how to make such a principled distinction,
I will assume for now merely that this can be done.5
Finally—and perhaps most important—one may question why I focus on the role
of intuition-based judgments at all, given recently expressed skepticism (in particular,
by Cappelen 2012) about whether intuitions have any significant role in philosophical
inquiry, including inquiry prompted by Gettier examples and the other familiar thought
experiments that I discuss here.6 Now I myself do not embrace Cappelen’s view that
intuitions play no role in philosophical inquiry. But I do agree with many of his
objections to the way they are often characterized by their defenders, in particular,
that intuitions have a common, distinctive phenomenology, that they are unreflective,
or even pre-theoretic, and that they (and thus the judgments based upon them, if these
are different) have a purely rational source and cannot be overturned either by further
argument or appeal to empirical facts or theories.7 Indeed, much of this paper is devoted
to challenging these claims.
I do, however, argue in what follows that the intuitively compelling judgments,
elicited by thought experiments, about whether something depicted as an F must,
or could fail to be, a G have some sort of ‘default justificatory status’ in (at least

4 Some philosophers identify intuitions with intuitively compelling beliefs about whether item F belongs to
category G, while others identify them with the (non-conceptual) bases of such beliefs—in Bealer’s (1996)
terminology intellectual seemings, or in Sosa’s (2007) terminology as pulls to believe certain statements
on the basis of our understanding of what is required for membership in that category. (Unlike Bealer (and
perhaps Sosa), however, I would not suggest that intellectual seemings (or pulls to believe) have a purely
rational source. All that should be required, in my view, is for it to seem compelling that, given what we
know, an item with these characteristics is (or is not) a member of that category.)
5 One also needs to say something about the source of concrete-case intuitions, and what counts as enough
‘further reflection’ to give intuition-based judgments secunda facie evidential status.
6 I am grateful to an anonymous referee for pressing me to clarify my views on this question.
7 See his (2012),1.2.2, 6.2.1–3, 7.1 for a characterization of the distinctive (and problematic) features of
intuitions. For the record, however, I share Cappelen’s worry (p. 131) about whether examination of the
passages in which now-famous philosophical set pieces are originally introduced—that is, the ‘written evi-
dence of philosophical practice’—provides the best examples of how these ‘cases’ are used in philosophical
inquiry. Better, I think, to examine how they are presented to undergraduates by philosophers who, perhaps
after struggling with tentative (and sometimes turgid) initial presentations, have come to believe that the
scenario described is at least a prima facie counterexample worth discussing.

123
4120 Synthese (2013) 190:4117–4136

certain sorts of) philosophical inquiry8 —and it is hard to deny that this sort of inquiry
is devoted, at least in part, to resolving conflicts between established principles and
category judgments of this kind. Thus, even if Cappelen has definitively shown that
philosophical arguments do not rely on intuitions in any familiar sense, the view
I advance about how philosophers should resolve these conflicts will remain a contro-
versial thesis for many naturalists, since it diverges from the way these conflicts are
adjudicated in scientific inquiry.9
In particular, epistemological naturalists may question why such concrete-case
judgments should have any prima facie evidential weight, especially when they are
generated by descriptions of farfetched, or physically impossible, scenarios, since it
is unclear where such intuitions come from, and hard to see how they could be tested
for reliability. Second, they may question why the only legitimate way to discount
judgments based on concrete-case judgments generated by thought-experiments is to
find that they lose their intuitive force under further armchair scrutiny, rather than
be dismissed on such methodological grounds as ‘simplicity’ or ‘systematicity’ that
generally operate in scientific theory-construction. Both these features of the armchair
methodology I’m endorsing, it may seem, threaten to take philosophical inquiry out
of the realm of good scientific method.
Indeed, some influential naturalists (e.g. Devitt 2010a, 2010b, 2010c; Papineau
2009, forthcoming) 10 explicitly press this charge. In a series of essays now collected
in his recent Putting Metaphysics First (2010), Michael Devitt raises questions about
traditional armchair methods that take intuitions generated by thought experiments to
be key evidence for philosophical theses. For Devitt (2010c, p. 299), intuitions can
be legitimate (prima facie) evidence for theses of philosophical interest only if they
are understood as intuitions about category membership based on general empirical
knowledge about the workings of the world and, when relevant, expertise about the

8 And thus I believe, contrary to Cappelen, that many (if not all) philosophical thought experiments attempt
to present a scenario that elicits an ‘F but not G’ judgment that ‘seems true’ (or at least not easily dismissed—
which I think would count as success enough for those attempting to test a thesis). I address the question
of why these judgments should be taken seriously in Sect. 5.
9 Clearly, these remarks do not do justice to Cappelen’s bold and bracing challenge to the widespread
contention that an essential feature of philosophical method is the (explicit or implicit) appeal to ‘intuition’.
This would require a more extensive discussion. However, though I’ve suggested, in the text, that a more
liberal characterization of what goes on when philosophers appeal to intuition (or intuition-based categorical
judgments) can escape some of his objections, there is another objection I’d like to address, namely, his
critique of the ‘so-called method of cases’. In the body of the paper I’ve referred uncritically to intuitions
(intuitive judgments) about ‘cases’, e.g. Gettier cases, trolley cases, Frankfurt cases, and other scenarios
designed to raise the question of whether something presented as an F must be, or could fail to be, a G—and
Cappelen maintains that there is much that is problematic in philosophers’ appeal to such cases. He argues
(p. 190; see also p. 132) that ‘when we say that philosophers appeal to cases, what we mean is that they draw
our attention to philosophically significant features of the world (or the way the world could have been)’.
This seems correct, but (in my view) it also seems to capture what most philosophers think they’re doing
when they appeal to Gettier cases, trolley cases, and the like, namely, presenting evidence that there is at
least a prima facie counterexample to a philosophical thesis. Cappelen continues, however (p. 190), that
‘describing such cases is hardly a method…[Moreover] it encourages the view that when we discuss cases,
the subject matter is the description given of the world…not the described feature of the world’. I agree
with the latter contention, but insofar as philosophical method includes testing a thesis against potential
counterexample, it seems that ‘appeal to cases’ can indeed be a legitimate part of philosophical method.
10 See also Kornblith (2002); Weatherson (2003)

123
Synthese (2013) 190:4117–4136 4121

kinds in question. This, he argues, is what we’re tapping not only when we attempt to
identify members of the kind gene or echidna, but also when we respond to thought-
experiments such as Gettier cases with the judgment that Jones does not know that
someone in his office owns a Ford, or to Kripke’s case of Schmidt and Godel with the
judgment that ‘Godel’ would refer to Godel even if he did not do any of the important
things commonly attributed to him.11 Moreover, (Devitt, 2010b, p. 282) endorses the
Quine-Duhem view that theory confirmation is holistic.12 Thus, even when intuitions
have a legitimately empirical source, they have no privileged long-term evidential
status, but can be overruled if they fail to cohere with the rest of our theories in
whatever way the Quine-Duhem thesis demands.
David Papineau has a similar view of the methodology of philosophical inquiry.13
Papineau, like Devitt, holds that ‘philosophical intuitions’ derive from high-level
empirical assumptions, that proper philosophical inquiry should aim to establish claims
about the world that are subject to disconfirmation by experience,14 and that theory
confirmation in philosophy can be viewed as a holistic procedure in which conflicts
between philosophical theses and intuition-based judgments are decided on the basis of
‘overall fit with the evidence’. Indeed, Papineau maintains (2009, p. 1) that in the cases
in which philosophical inquiry can make a ‘worthwhile’ contribution, it ‘is already
akin to science, not that it needs reforming in order to become so’. He acknowledges
that philosophical thinking, whether general or more specific, is often caught in ‘the-
oretical tangles’ that may seem incapable of resolution by consulting, or reevaluating,
empirical data—and thus that the only route to progress is further reflection from the
armchair. But all that such reflection would provide, Papineau contends, is a ‘creative
restructuring’ of hidden or unacknowledged assumptions, and there is no reason to
think that those assumptions are themselves anything but empirical.
Nonetheless, despite Papineau’s claim that what’s ‘worthwhile’ about philosophical
inquiry ‘is already akin to science’, there seem to be substantial differences between
the use of intuitions in scientific and armchair philosophical inquiry, at least as gener-
ally practiced. In particular, many of the intuition-based judgments that philosophers

11 Insofar as claims based on these intuitions are regarded as knowable a priori, Devitt argues, armchair
methods are incompatible with naturalism, since ‘there is (2010a, p. 253), ‘only one way of knowing, the
empirical way that is the basis of science’.
12 He writes: ‘this [empirical] way of knowing is holistic…we have no good reason to think that our trou-
blesome knowledge of mathematics, philosophy, and logic could not be accommodated within this holistic
empirical picture’. Though Devitt here claims that logic and mathematics, in addition to ‘philosophy’, can
be known in the ‘empirical way’, I’ll confine my attention to our knowledge of philosophical theses about
such things as knowledge, moral action, and freedom—and of the philosophical intuitions on which this
knowledge is based. Compare this chapter, p. 275, and Ch. 12 (pp. 257–258).
13 Unlike Devitt, who rejects the a priori entirely as (2010b, p. 287) ‘mysterious, even mystical’, perhaps
(p. 290) ‘occult’, Papineau holds that philosophical naturalists can accept the possibility of a priori knowl-
edge. On his view, however, a priori knowledge extends only to ‘analytic’ truths such as ‘Triangles have
three angles’ or ‘Saffron is yellow’, and thus is empty, uninformative—and completely irrelevant to philos-
ophy, which ‘consists of synthetic theorizing, evaluated against experience’. Papineau, like Devitt, contends
(forthcoming, p. 5), that in their well-known thought experiments, ‘Kripke and Gettier were appealing to
familiar empirical assumptions about names and knowledge respectively, rather than to purely conceptual
intuitions’.
14 Even if (p. 4) this is merely ‘empirical information that is part of pre-existing thought, as opposed to
information prompted by novel evidence’.

123
4122 Synthese (2013) 190:4117–4136

take seriously (e.g. about moral action and responsibility, freedom, personal identity,
and—at least arguably—knowledge and justification) do not seem to be products of
general empirical knowledge. Moreover, philosophers are not generally inclined to
dismiss intuitions that conflict with an established thesis solely on grounds of simplic-
ity or systematicity. And thus it may seem that armchair methodology of the sort I’ve
described cannot be embraced by epistemological naturalists, and should therefore be
substantially revised.
In my view, however, this would be a mistake. I believe that the most effective
way to promote philosophical progress is to treat intuitions as having the (prima and
secunda) evidential status I’ve described (except in a small and idiosyncratic class of
cases),15 and an important aim of this paper is to argue for this claim. However, as
I’ll also argue, such a methodology is compatible with a reasonable version of the
epistemological naturalism embraced by Devitt, Papineau, and many others, and thus
an armchair methodology that gives such evidential weight to judgments based on
concrete-case intuitions, if properly practiced, need not be abandoned by naturalists,
or even (again, except for a few exceptions) much revised.
I’ll begin by considering, and disputing, a number of general proposals for when it’s
legitimate to discount a judgment based on a concrete-case intuition, even when that
judgment retains its intuitive force. First (in Sect. 2), I’ll examine an account of proper
philosophical method recently proposed by Brian Weatherson (and endorsed by Pap-
ineau) that permits intuitively compelling judgments to be discounted if they conflict
with general principles that are sufficiently systematic—and I’ll argue that adopting
this methodology would impede philosophical progress by stopping inquiry too soon.
Next (in Sect. 3) I’ll consider, and dispute, the claim that there are facts about either
the persistence or provenance of our intuitions that provide grounds for discounting
the evidential force of any judgments based upon them, but argue (in Sect. 4) that a
methodology that gives such evidential weight (both prima and secunda facie) to such
judgments can meet (albeit indirectly) certain well-established criteria for a method-
ology to be open to empirical input, and thus be compatible with epistemological
naturalism.16 Finally (in Sect. 5), I’ll discuss some questions that can be raised about
the empirical plausibility of this account of the evidential status of judgments based
on concrete-case intuitions—and, though I’ll provide some reasons to think that these
questions can be answered, I’ll acknowledge that, here too, the tribunal of experience
must be the final judge.17

15 Once again, these would be cases that involve phenomenal concepts derived from introspection. See
note 3.
16 Of course, if armchair methods that give special status to concrete-case intuitions can face the ‘tribunal
of experience’ even indirectly, they cannot yield knowledge that is strictly a priori. Some traditional philoso-
phers may thereby regard any such version of armchair methodology as fatally flawed. However, even if the
distinction I sketch between scientific and armchair methodology cannot count as the distinction between
the empirical and the a priori as traditionally conceived, it seems (more or less) extensionally equivalent to
the distinction that traditional philosophers have wanted to draw and defend.
17 And thus any experimental philosophers persuaded by Cappelen (2012, Ch. 11) to give up the practice
of canvassing philosophical intuitions will find many suggestions for other things to do.

123
Synthese (2013) 190:4117–4136 4123

2 Weatherson on intuitions and philosophical methodology

In his (2003) discussion of the role of counterexamples in philosophical inquiry, Brian

Weatherson implicitly endorses the traditional armchair method in which theses of
the form ‘F iff G’ are tested by thought-experiments designed to elicit intuitions that
there could be F’s that are not G, and thus counterexamples to the thesis. He observes,
however (p. 1), that different branches of philosophy resolve conflicts between thesis
and intuitions in different ways: in epistemology, Gettier cases are considered ‘obvi-
ously and immediately fatal’ for theories that count them as instances of knowledge,
whereas in ethics, the principle of utility often is endorsed even when, in response to
thought experiments involving trolley cases, it seems intuitively that ‘the action that
maximizes utility is not the right thing to do’. And this, Weatherson continues, is a
problem: ‘Either there is some important difference…between the anti-utilitarian cases
and the Gettier cases, a difference that justifies our differing reactions, or someone is
making a mistake.’
Weatherson’s view is that there is no important difference between these cases, and
that therefore, in the interest of achieving consistency in one’s philosophical theory
construction, these practices should be brought into conformity. Moreover, he argues
(Sect. 1), it is reasonable to think that it is the epistemologists who are making the mis-
take. Thus Weatherson urges epistemologists to bring their practices into conformity
with those of moral philosophers by adopting certain general methodological princi-
ples, familiar from the natural sciences, that permit an established principle to override
an intuition-based judgment to the contrary if doing so greatly increases the overall
coherence of one’s theory. In particular, he argues (p. 5) that if the principle under
threat can ‘account for as many intuitions as possible, while remaining systematic’,
then it can be retained and the counterexample dismissed.18
However, it seems that, contrary to Weatherson, traditional armchair methodology
can identify a difference between trolley cases and Gettier cases that justifies this
differential treatment of them. Suppose that the reason we accept the principle of utility,
even when it recommends, as in a classic trolley case, an action that seems morally
wrong, is that when thinking more carefully about the action, it begins to seem (for
example) that shoving a single heavy person onto a trolley track to prevent the trolley
from hitting five others is more similar to the uncontroversial cases than it initially
appears. And suppose that this, in turn, makes it intuitively compelling to believe
that shoving the person onto the trolley track would be morally acceptable, perhaps

18 Weatherson’s four criteria (2003, pp. 8–9) for judging a theory are:

(1) Counterexamples count against it—but (a) ones that generate stronger intuitions that they are correct
count more, and (b) obscure or fantastic cases count against a theory less, and (c) if we can explain
why we have a mistaken intuition, this permits us (in many cases) to reject it.
(2) Analyses can’t have too many unacceptable theoretical consequences.
(3) Concept should be analyzed in theoretically significant terms.
(4) Analysis must be simple.

123
4124 Synthese (2013) 190:4117–4136

obligatory—even if, at the same time, we feel revulsion at the prospect.19 In this case,
further reflection on the principle of utility (and the many cases for which it provides
intuitively acceptable verdicts) would not merely reinforce the belief in the principle’s
scope and systematicity, but also make it less intuitively compelling to think that the
trolley case is a counterexample to it (although it may highlight certain unpleasant
features of ‘doing the right thing’). So far, so traditional: no further methodological
principle such as ‘Maximize systematicity’ would need to be invoked.20
In contrast, it is hard to see how thinking about other situations in which a person’s
justified true belief does seem to be knowledge would make it intuitively compelling
to believe (as in some classic Gettier cases) that Jones knows that someone in his
office owns a Ford, or that the person who got the job has ten coins in his pocket. But
if this is so, then there is an explanation of the difference in the ways that conflicts
between intuition and theory are treated in ethics and epistemology: both use the same
(traditional) armchair methodology, but in the trolley cases, but not the Gettier cases,
further reflection produces a change in our intuitions, in what intuitively seems to
be the case. And this is why we are justified in continuing to regard those intuitions
as ‘fatal’ to the prospects of principles such as ‘knowledge is justified true belief’,
even though, despite initial intuitions (and ongoing ‘feelings’) to the contrary, we may
accept the principle of utility.
But even if traditional armchair methodology can make a principled distinction
between Gettier cases and trolley cases along these lines, one may ask whether it is
good practice, in general, to take established principles to trump ‘recalcitrant’ judg-
ments that retain their intuitive force after further reflection. Weatherson suspects that
philosophers would not be likely to embrace such a methodology, and I agree. In
practice, it seems that while conflict between intuition and principle may prompt us
to reexamine our conviction that some imagined scenario is a counterexample to that
principle, if our intuitions say ‘yes’ after due scrutiny, then we generally take yes for
an answer—once again, for better or worse—and continue to look for a principle that
accommodates those intuitions.
One can say much the same about Papineau’s (forthcoming, p. 14) proposal for
adjudicating disputes about the nature of free will, personal identity, and other philo-
sophically interesting phenomena. In arguing that the seeming irrelevance of empirical
information to these disputes does not mean that they can be resolved only by a pri-
ori reflection, Papineau suggests that these disputes ‘can’t be resolved at all because
some crucial concept is indeterminate’. That is, he continues, it’s reasonable to take
those who assert, and those who deny, that free will is compatible with determinism,
or that personal identity depends on psychological continuity, to be using free will
or personal identity in different ways—as defined by one, rather than another, frag-
ment of an imprecise common sense theory—and thus that the conflict is merely a

19 Utilitarians, after all, do often argue in this way. For example, consider the well-known ‘dirty hands’
response to anti-utilitarian arguments, in which it is argued that one’s discomfort at the prospect that one
would be held responsible gets in the way of one’s judgment about what it is morally right to do.
20 This is a good example of an argument that assumes a distinction between the concrete-case intuitions
about category membership elicited by a thought-experiment, and other reactions (such as revulsion) to the
description of the scenario in question. But although it is not easy to make this distinction, it seems that
further ‘good faith’ reflective scrutiny can do a reasonable job. I discuss this question further in Sect. 4.

123
Synthese (2013) 190:4117–4136 4125

‘terminological dispute’. Presumably he would make the same claim about pairs of
conflicting intuitions that retain their pull on further reflection.
For some disputes this may be an acceptable suggestion.21 But it often seems
unsatisfactory to regard philosophical conflicts as terminological, even if it seems that
they cannot be resolved by empirical evidence, since this fails to explain (for example)
why the free will debate remains fresh, and indeed sometimes unsettling, and why the
question of what constitutes personal identity remains compelling. Yes; we can offer
an explication of free action, such as the ones mentioned above, that permits an action
to be free even if the agent couldn’t have done otherwise. Or we can accept the claim
that determinism precludes an agent’s ability to do otherwise, and thus her freedom
of action—but argue (as Frankfurt famously does) that actions that spring from the
agent’s motives can be praiseworthy or blameworthy even if they are not free. But this
does not dull the suspicion that we’ve changed the subject, rather than merely focused
on one among a number of our concepts of what it is for an action to be free. Nor does
it explain why it seems so hard to agree on a characterization of the class of actions
for which we can be held morally responsible if determinism is true.
Similarly, we can distinguish psychological from bodily continuity, claim that they
provide characterizations of different types of identity through time—and leave it at
that. But this too seems to be a matter of changing, not clarifying, the subject, and it
too remains unsatisfying, since (at least arguably) an important goal of this inquiry is
to determine which future or past person (if any) is unequivocally me. Once again, we
are inclined to resist the suggestion that (for one reason or another) there is no need
for further inquiry.
Our question, of course, is whether this is good practice, and no doubt Papineau and
Weatherson, along with Devitt, would urge philosophers to join scientists in discount-
ing ‘recalcitrant’ intuition-based judgments in favor of simplicity and systematicity,
or regarding philosophers who cannot resolve a dispute to be talking about different
things. As I’ve noted earlier, I dispute this view, and hold that taking intuition-based
judgments to have special evidential weight not only reflects standard philosophical
practice, but that discounting them solely on grounds of simplicity or systematicity
would inhibit philosophical progress. But I haven’t yet said much about why this is
the case.
The reason, in my view, is that if we dismiss philosophical intuitions that retain their
intuitive pull, we may be giving up inquiry too soon. For example, it is tempting to urge
philosophers to ignore any intuitive pull they may feel to believe that specific human
actions, if fully determined by prior events and the laws of nature, cannot be free—and
thus accept some sort of compatibilist view that characterizes free actions as actions
produced by a certain sort of cause (or as actions compatible with the agent’s values,
or second-order volitions). However, there is recent work by compatibilists on the free
will problem that endorses a characterization of free action that incompatibilists pur-
port to accept, namely, actions for which the agent could have done otherwise—and
argues that there are actions that meet these conditions even if determinism is true

21 For example, consider Sosa’s (2007) suggestion that the clash in intuitions supporting evidential and
reliabilist theories of knowledge is best resolved by taking each set of intuitions to reflect a different concept
of knowledge.

123
4126 Synthese (2013) 190:4117–4136

(see Vihvelin, forthcoming, and Perry, forthcoming). In this case, continued philo-
sophical reflection provides (at least arguably) substantive results.
Will these arguments convince incompatibilists? One cannot say for sure, of
course—but one also cannot rule out the possibility that these considerations could
eventually make compatibilism seem intuitively compelling, even when one is paying
attention to the possibility that determinism is true, and if so there will be progress in
resolving a long-standing philosophical question. But a methodology that dismisses
intuition-backed judgments that stand as potential counterexamples to established
philosophical theses in favor of ‘simplicity’ or ‘systematicity’ would inhibit any such
progress.22
One might worry, however, that I’m settling into the armchair too soon. Surely there
are some considerations that permit us to dismiss certain concrete-case intuitions as
illusory, or otherwise problematic, even if they retain their force on further reflection.
After all, there are perceptual illusions, such as the Muller-Lyer lines and the Ames
Room, that continue to look a certain way even after we recognize that they repre-
sent the world incorrectly—and we don’t take them seriously as data for a theory of
perception. And so, one might suggest, there may be certain sorts of intuition-based
judgments that remain compelling upon further reflective scrutiny, yet have no legit-
imate evidential status. This is an attractive view, endorsed by many, but it seems
unlikely to explain the continuing sway of our intuitions about knowledge, justifica-
tion, free agency, personal identity, and moral action. I will argue for this claim in the
next section.

3 Is there a privileged class of philosophical intuitions?

In an ongoing body of work, Ernest Sosa addresses just this question, and suggests
that the difference between legitimate and spurious intuitions (that retain their intuitive
force) can be viewed as a difference between intuitions that do, and intuitions that
don’t, derive from a (rational) competence.23 He explicitly compares intuiting with
perceiving, and argues that just as we take only those perceptual deliverances that our
sensory systems are competent to discriminate as evidence for perceptual beliefs, so we
should regard concrete-case intuitions as evidence only if they arise from a competence
to ‘discriminate…the true from the false, in some subfield of the modally strong…with
no reliance on introspection, perception, memory, testimony, or inference…’ (2007,
p. 61).

22 There is also recent work on the Gettier problem (Schroeder, forthcoming) that presents new char-
acterizations of what it is to have knowledge that encompasses features that most theorists agree to be
necessary for knowledge, and argues that it rules out Gettier cases as well, and also some recent work
(Wasserman, forthcoming) on what it is to be an intentional action that helps to explain and perhaps resolve
some ‘theoretical tangles’ involving the relation between doing A intentionally and having an intention to
do A.
23 See his 2005, 2007, 2008. Sometimes Sosa suggests that ‘competent’ responses to thought-experiments
are best construed as pulls to believe some proposition, and other times as concrete-case beliefs; as already
noted in Sect. 1, this does not matter here.

123
Synthese (2013) 190:4117–4136 4127

This may well be a promising approach—but only if there is a good enough analogy
between our perceptual and rational ‘competences’; only if there is an explanation of
what it is to be a legitimate intuitive seeming that is analogous to the explanation of
what it is to be a legitimate perceptual seeming. Now, there is no doubt more than
an analogy here for some of our intuitions, for example, its seeming that something
could be a spatial configuration, or a color incompatibility, of a certain sort. These
sorts of intuitions, it seems, have their source in what we can (visually) imagine, which
ultimately has its source in what we can (visually) perceive—and thus any facts about
our perceptual competences could also yield a verdict about whether or not seemings
of these other sorts can be trusted. But this is not true for intuitions about whether
something is knowledge, justified belief, free action, or moral action, and so we need
to ask whether there is a more general characterization of the conditions under which
we can legitimately discount the evidential standing of perceptual seemings that can
apply to this broader class of concrete-case intuitions as well.
One example that Sosa often works with is the difference between a (quick) glance
at a row of eight evenly spaced dots on a blackboard versus a row of eight dots displayed
as four pairs. Sosa argues, plausibly, that the latter, but not the former, can produce
a perceptual seeming that justifies the belief that there are eight dots on the board,
given certain well-established facts about our inability to discriminate differences in
cardinality above a certain threshold. So here is a case in which the verdict about
whether a perceptual belief arises from a competence depends on our having an inde-
pendent theory about (e.g.) the discriminative capacities of our perceptual systems.
Another means for determining whether a perceptual seeming derives from a compe-
tence is whether it conforms to the deliverances of different perceptual systems—e.g.
touch and vision. It is because vision and touch produce different perceptual seem-
ings, along with other information about the workings of these faculties, that we can
discount visual illusions such as the Muller-Lyer lines and the Ames Room (as well
as Descartes’s towers in the distance).
To be sure, both sorts of ‘corroboration’ of perceptual seemings have their ultimate
source in perception itself, and thus, as Sosa points out (2007, p. 64), if we must
question everything that has a perceptual source, we could have no justified perceptual
beliefs. But he contends that if a ‘divide and conquer strategy’ that permits these sorts
of agreement to give credence to our perceptual seemings, then ‘it could be adapted
for the support of intuition’ as well.
However, it’s not clear that this is so. Though we do have theories that would permit
us to invoke our rational limitations to disqualify intuitions about certain complex or
ambiguous phenomena, it’s hard to see how these considerations would apply to the
concrete-case ituitions, generated by thought-experiments, that we’ve been consider-
ing here.24 In addition, there seems to be no ‘intellectual’ analogue of the disparate

24 Now I suppose that some (future) theory of cognition could be relevant to some of these cases. For
example, some philosophers (McGinn 1989; Nagel 2000) suggest that the reason thought-experiments
involving zombies and disembodied minds seem so compelling as counterexamples to materialism is that
we are incapable (at least for now) of formulating concepts that display a link between the mental and the
physical. Perhaps empirical investigation can confirm this hypothesis, and if so, give us reason to dismiss
those intellectual seemings as due to some (temporary or permanent) cognitive deficit. My own view is that
this is implausible; that if there is an explanation for why intuitions about zombies and disembodied minds

123
4128 Synthese (2013) 190:4117–4136

sources of perceptual seemings (for example, vision and touch) that may, or may not,
yield intuitions that agree; intuitive seemings, though perhaps influenced by a variety
of factors, instead seem to arise from a single source.
Sosa sometimes suggests (e.g. 2007, p. 64) that the sort of agreement in intuitions
that could mark them as deriving from a rational competence is agreement among
different individuals that a candidate counterexample should be described a certain
way; in this case, he writes, ‘one could lean one’s intuitions evidentially on those of
others’.25 But agreement itself can’t be the mark of rational competence, since it may
derive from non-rational sources as well, e.g inertia, the emotional pull generated by
the thought-experiment, or one of the many heuristics that psychologists have found
people to use (and often overuse) in their reasoning. We want to eliminate agreement
deriving from those sources from adding weight to the authority of our own intuitions,
and our only way of doing so, it seems, is to use armchair techniques, to the best of our
ability, to determine whether our own intuitions remain robust. But if we can do that,
then why bother to look for agreement? In short, it seems that the only plausible account
of what it is for an intuition (at least of the sort we’ve been discussing) to arise from
a rational competence is for it to be able to survive the continued reflective scrutiny
characteristic of armchair dialectic—and thus it cannot provide an independent way
of discounting intuitions that such reflective scrutiny yields.26
There is yet another source of worry, however, about a method that gives intu-
itions such evidential status, namely, that philosophers can become so invested in their
theories—or so used to seeing the world through their lenses—that they are inclined to
discount conflicting intuitions without giving them a fair appraisal. Granted, philoso-
phers are not alone in having such proclivities; this is an occupational hazard for
any theorist, and is not exclusive to this domain. Nevertheless—and here’s the rub for
philosophers sympathetic to naturalism—scientific theories that are products of biases

Footnote 24 continued
are to be discounted, it will proceed from further reflection about the nature of our phenomenal concepts
and the mechanisms by which they refer—which, of course, can be accomplished from the armchair. But
even if this is a plausible explanation of our anti-materialist intuitions, it’s hard to see how the concrete-case
intuitions about whether something is knowledge, justification, freedom, or moral action that retain their
pull under reflective scrutiny could be shown to be products of some conceptual lacuna or cognitive deficit.
25 See also Goldman (2010). Sosa also invokes stability, over time, in the judgment of a single individual,
but this criterion shares many of the problems of the first.
26 Sosa also suggests that if there are stable and enduring disagreements about thought-experiments among
different populations, even after all precautions have been taken, we can conclude that that the populations are
using different concepts, and thus that ‘supposed intuitive disagreements…are to a large extent spurious, that
different epistemic values are in play, and that much of the disagreement will yield to a linguistic recognition
of that fact…’ (2007, p. 69). But here too it’s plausible to regard such disagreement as due to a divergence
in our concepts of knowledge only if we are able to recognize that there are distinct concepts, deployed
by ourselves and ‘the other’, that could both have some claim to the name of knowledge. And this too is
a matter of what we, individually, find intuitively compelling after due reflection. (One can say the same
about divergences in one’s own intuitions over time.) It seems, therefore, that even the plausibility of an
explanation of intuitive disagreement depends on how it fares in the standard back-and-forth of philosophical
dialectic. There are others (e.g. Cummins 1998) who have raised similar worries about whether philosophical
intuitions can be ‘calibrated’, and take the negative answer to show that intuitions should be discounted
altogether. My view is that this shows that (careful) armchair methodology is all that we’ve got, and thus
the question is whether it can be vindicated in some other way.

123
Synthese (2013) 190:4117–4136 4129

or overlooked data will, at least eventually, face the tribunal of experience, and one
might doubt that beliefs based on concrete-case intuitions (about phenomena such as
free will, personal identity, knowledge, moral action, etc.) can do so as well. And if
they cannot do so, then it seems that philosophers sympathetic with naturalism must
make a potentially costly choice.
It seems to me, however, that although we can’t expect beliefs based on concrete-
case intuitions to confront the ‘tribunal of experience’ in quite the way that scientific
theories do, they nonetheless can be subject to pressure from, and thus in an important
sense answerable to, empirical (and methodological) facts. I’ll argue for this view in
the next section.

4 Intuitions and empirical evidence

Michael Devitt, in his (Ch. 12, p. 256) spells out two ways in which the belief that p
can be sensitive to empirical evidence, namely,
(i) The ‘evidence might bear for or against p itself’; that is, the content of that
(statement of) evidence can conflict with the content of p.
(ii) The ‘evidence might throw light on the goodness of [an individual’s or a commu-
nity’s] thinking about p’.
These conditions seem to be reasonable expressions of what it takes for a belief to be
the product of ‘the empirical way of knowing’. Now Devitt holds that a strict naturalist
must affirm (i) as a way in which all beliefs can be answerable to the evidence; affirming
(ii) alone is too easy, he contends, and does not yield a view of theory confirmation
that is ‘captured by the metaphor of the seamless web’. This may be debated,27 but
I’ll accept this constraint in what follows.
Is it even possible, however, for empirical evidence to ‘bear against’ concrete-
case judgments prompted by Gettier cases, trolley cases, thought-experiments about
brain switches, and the like in ways that can diminish their intuitive force? It is easy
to see how further reflection could do so. For example, on first considering thought
experiments involving the transplant of A’s brain into B’s body, and vice versa, it may
initially seem intuitive to judge that the resulting person with A’s brain is the same
person as A before the operation—but then we encounter Bernard Williams’s famous
thought experiment in his (1970), and may subsequently be inclined to judge that the
person with B’s brain is the same person as pre-operative A. Or we find neither option
intuitive—and go back to the drawing board.28
It is also easy to see how, in various other cases, consideration of empirical evidence
could prompt such a change in our judgments. For example, after hearing a purely
qualitative description of the contents of the lakes and streams on Twin Earth, we
may find it intuitively compelling that S and her Twin Earth counterpart mean the

27 However, the considerations marshaled in the previous section suggest that no such considerations
would be relevant to philosophical intuitions.
28 These reconsiderations and revisions can occur even after considerable thought has been given to the
topic. For example, Judith Thomson, in her (2008) has revised her characterization of what is going on in
some of the trolley cases that have been important in her own inquiry.

123
4130 Synthese (2013) 190:4117–4136

same thing when they call that stuff ‘water’—but after learning some empirical facts
that cannot be deduced from that qualitative information, namely, that the molecular
composition of that stuff is XYZ and not H2 O, we may no longer feel the intuitive
pull of that judgment. And presumably, this could also happen when a category belief
prompted by a philosophical thought-experiment contains a term that purports to pick
out a natural kind (or other kind with a ‘hidden essence’), either demonstratively or
by means of a salient but contingent ‘mode of presentation’.29
However, the cases I’m concerned with seem best treated not as natural kinds, but
(in Locke’s locution) as nominal kinds; that is, phenomena to be characterized not
by demonstrations or reference-fixing descriptions, but by descriptions that purport to
specify conditions that are (logically) necessary and sufficient for being an instance of
that kind. Some philosophers, contrary to at least some naturalists,30 regard knowledge
as a nominal, rather than a natural kind. This is also the standard view of moral action,
responsible action, free agency, and personal identity as well, since it is hard to see
these as kinds that can, even in principle, be isolated and studied by scientists. But
if this is so, then it is hard to see how, as Devitt requires, empirical evidence could
be relevant to intuition-based judgments generated by trolley cases, Frankfurt cases,
brain and body switches, or the like.
Now Devitt (and also Papineau, another confirmation holist) can argue that even
if empirical evidence seems irrelevant to some particular intuition-based belief, that
belief can count as sensitive to empirical evidence if it coheres sufficiently well with
a theory that as a whole is confirmed (or at least is not disconfirmed) by the evidence
in Devitt’s sense (i). However, it’s not clear why replacing generalizations such as
‘knowledge is justified true belief’ or ‘the morally obligatory action is the one that
produces the greatest happiness’ or ‘personal identity requires bodily continuity’ with
the judgments typically generated by Gettier cases, trolley cases, and Locke’s prince
and cobbler case should make any difference to a theory’s degree of empirical confir-
mation as a whole.31 And if it doesn’t, then it’s hard to see how the verdicts reached
by reflecting on philosophical thought experiments could in any sense, no matter how
attenuated and indirect, be answerable to experience, and thereby be taken seriously
by philosophers who would be naturalists.

29 For example, if we understand knowledge (as some do) as true belief produced by a reliable mechanism,
it is easy to see how empirical evidence could play a role in overturning an initial judgment that S does or
does not know that p. If we learn that S has had a run of true beliefs about the emotional states of others, or
about which ticket will win the lottery, then—not knowing much about the source of these beliefs—it may
seem intuitive that these are merely lucky guesses. But if scientists determine that S’s beliefs are generated
by the right sort of mechanism, then consulting this empirical information would presumably diminish the
intuitive force of this initial judgment, just as learning that the stuff in the lakes and streams on Twin Earth
is composed of XYZ molecules diminishes the intuition that S and her counterpart mean the same thing
when calling that stuff ‘water’.
30 Papineau has suggested, if not fully endorsed, the claim that knowledge should be treated as a natural
kind. And see Kornblith (2002), for a firm endorsement of the view.
31 To be sure, as Weatherson and others argue, the theories that emerge when ‘recalcitrant’ intuition-backed
judgments are rejected may be simpler, but, as I’ve argued, simpler theories are not always satisfying—and
if the intuitive judgments that conflict with them are compatible with the empirical evidence, there seems
to be no particular reason (other than simplicity) for a naturalist to reject them.

123
Synthese (2013) 190:4117–4136 4131

It seems, however, that there is another way in which empirical evidence can count
as bearing against intuitively compelling judgments, namely, if their intuitive force
could, in principle, be diminished (or augmented) by reflection on empirical facts
and methodological principles. If it’s possible that further reflection, over time, on
these facts and principles could shift our general understanding about what’s required
to be a member of a particular kind or instance of a particular property—so that an
initially compelling description of a situation later seems incorrect, or at least not
obvious—then this, I contend, is enough to permit a naturalist to invoke intuition-
backed concrete-case judgments as (secunda facie) evidence, as long as they retain
their intuitive force.
Now it may seem implausible to think that any empirical discoveries (or any increase
in methodological sophistication) would change the intuitive force of judgments about
such things as Gettier cases or trolley scenarios.32 But this doesn’t matter; what’s
important for a judgment to meet Devitt’s condition (i), it seems, is merely that it be
the sort of belief that could lose its intuitive force in the face of genuinely conflicting
empirical evidence; for example, if increased familiarity with empirical discoveries
and scientific methodology could change our inclinations to categorize objects in
certain ways, or to take certain properties to be essential features of items of some
kind. If this is possible, I suggest, then philosophical theses based on concrete-case
intuitions that seem isolated from empirical evidence can nonetheless be viewed as
sensitive to the way the world is, and thus the range of philosophical inquiry compatible
with naturalism may be greater than many naturalists believe.
This will be so, of course, only if there is no reason to think that (considered)
judgments based on concrete-case intuitions are impermeable by further empirical
evidence. However, Papineau (2009; in defense of his claim that intuition-based judg-
ments about knowledge, free will, personal identity, and the like are products of
synthetic general assumptions, even though they seem impervious to empirical evi-
dence) suggests that these intuitions may be encapsulated in some cognitive system.
In particular, he suggests that philosophical intuitions may arise from certain hard-
to-dislodge ‘assumptions’ produced by ill-understood ‘subpersonal mechanisms’ that
operate much like the unconscious ‘edge-detectors’ that determine that we see changes
in intensity in the visual field as edges of objects.33 And if this is so, he argues, there
is no case in which those intuitions could diminish in force, even if we recognize that
they could be false if applied outside certain specific domains.
If Papineau’s hypothesis about the encapsulation of concrete-case intuitions (or the
assumptions that prompt them) is correct, then these intuitions cannot meet any version
of Devitt’s Condition (i). But there are a number of problems with this hypothesis.

32 This, presumably, is because it’s hard to see how further empirical discoveries or methodological shifts
could conflict with any views we now hold about the nature of these phenomena or the norms that govern
them. But perhaps I’m being just as shortsighted as those who thought that planetary orbits had to be circular,
or that vacuums could not exist.
33 These mechanisms would work, Papineau suggests (ms p. 11), by ‘tak[ing] in information which do[es]
not presuppose the relevant categories, and us[ing] it to arrive at judgments about who knows what, and
which words name which things, and when someone is the same person as someone else, and so on’. This
suggestion may be another way of fleshing out Sosa’s 2005, 2007, 2008 suggestion that some intuition-based
judgments should be treated as analogous to perceptual illusions.

123
4132 Synthese (2013) 190:4117–4136

First, it is a serious empirical claim that needs supporting evidence—which Papineau

indeed attempts to give. He argues (2009, p. 18) that there are independent reasons for
thinking that ‘synthetic assumptions [about phenomena such as knowledge, names,
persons, free will] are built into the automatic mechanisms that allow us to make
particular judgments about philosophically salient categories…[since] judgments like
these are important to us in our daily life’.
But this argument is not convincing. First, it’s not clear that philosophical disputes
about the nature of knowledge, names, free will, etc. do have much significance in
our daily lives. Little in our ordinary doings is affected by whether or not Kripke’s
theory of names, or the tripartite analysis of knowledge, or the psychological continuity
account of personal identity is true—which, of course, explains why so many thought-
experiments that attempt to probe intuitions about what’s possible seem so outlandish
(and why any disquiet about philosophical dilemmas tends to vanish when we return to
the backgammon table). Second, if our philosophical intuitions derive from automated
sub-personal mechanisms like edge-detectors, it’s hard to explain why anyone fails to
share our intuitions about what a name refers to, or whether something is a case of
knowledge, moral action, personal identity or free will.
Moreover, if our intuitions about knowledge, freedom, personal identity, moral
actions, etc. derive from encapsulated assumptions of any specificity, it’s hard to
explain why they are cognitively penetrable by the deliverances of further reflection.
In the back-and-forth of standard armchair inquiry, the intuitive force of some of our
initial judgments often does decrease or increase when we think further about the
details of the case and more deeply about what we understand the phenomenon in
question to be. In sum, these intuitions do not seem to be cognitively impenetrable
in general, and there’s no particular reason to think that there are any subpersonal
cognitive mechanisms that constrain them,34 given that they don’t seem to make any
crucial contributions to our everyday navigation of the world.35
In the end, of course, this is an empirical question: maybe we do make some
judgments based on intuitions that are virtually impossible to dispel, no matter what.

34 Except, perhaps, for those attempts to imagine certain scenarios that inherit built-in constraints on
perception. In these cases, Sosa’s analogy between recalcitrant intuitions and perceptual illusions may
be apt.
35 In addition, in examining the mechanics of armchair reflection both now and throughout the history
of philosophy, there seem to be many cases in which empirical findings have influenced concrete-case
intuitions, for example, the dispute between Locke and Descartes about whether ‘our concept of body’
requires only that bodies be spatially extended, or also ‘solid’, and Locke’s primary-secondary quality
distinction. (See my 2004 for further discussion of cases such as these.) And if empirical considerations can
influence one’s initial, seemingly rational, judgments, why shouldn’t they also be able to effect change (for
an individual or community) in what intuitively seems required to be an item of some particular kind? For
example, consider a description of a biological creature that is functionally equivalent to ourselves when
in pain, but is not undergoing C-fiber stimulation (or whatever it is that goes on in human brains when we
are in pain). We would presumably judge that the creature is not in the same brain state as we are when in
pain—and if so this scenario may seem to support functionalism, rather than the identity theory. However,
some philosophers (e.g. Bechtel and Mundale (2009) note that neuroscientists, in practice, individuate types
of neural states more liberally—requiring them, for example, to share only a common structure and pattern
of firing—and thus the scenario need no longer be considered to be one in which the creatures do not
share our brain states. Here, it seems, any change of mind would be due to the willingness to accept this
redescription, which clearly would be influenced by empirical information.

123
Synthese (2013) 190:4117–4136 4133

And maybe, in addition, there is no good explanation of why these should be exceptions
to the rule. But at least in cases such as knowledge, free will, personal identity and
moral action, there is not much evidence that this is the case, and at least some evidence
against—and this would thereby bring a good deal of classic armchair inquiry into the
naturalists’ domain.
Still, even if there is reason to think that our category judgments could shift under
pressure from empirical evidence, two questions remain: why, if they can do so, would
this count as a shift in our understanding, rather than an ordinary change in our beliefs
about some phenomenon, or, as Papineau suggests, a sharpening of our ‘indeterminate’
concepts; and third, why this would count as a shift in our understanding of some
phenomenon rather than a departure that counts as changing the subject and talking
about different things.
The answer to the first question is the one that philosophers have invoked throughout
the history of the discipline, namely, that in the back-and-forth of armchair reflection,
we often reach a stopping point at which we think: this is what it is to be an F or a
G, and we think that the conditions specifying ‘what it is’ give the right verdict for
possible cases, even when those possibilities obtain in worlds quite distant from our
own. In short, there seems to be a qualitative difference between the ‘stopping point’
category judgments one reaches in philosophical thought-experiments and those that
derive from theoretical specification of the entities in other scientific theories36 —and,
as I’ve just argued, it is implausible to appeal to hypotheses such as encapsulation to
provide an alternative explanation of why this occurs. And thus it seems reasonable to
regard concrete-case intuitions about Gettier cases, trolley cases, prince and cobbler
scenarios, etc. as products of the understanding.
However, to say that these intuitions are products of the understanding in this sense
does not imply that our concepts of knowledge or moral action or personal identity
have an occult source; there may be a naturalistically acceptable account of how they
are acquired.37 In addition, we may regard these concepts as applying (or purporting
to apply) to ontologically naturalistically acceptable items in the world. Finally—
and this is the claim relevant to epistemological or methodological naturalism—the
category beliefs that are products of our understanding, in this sense, need not be
considered as unrevisable come what (empirically) may; I’ve just offered an account
of the conditions under which they can count as being answerable to the empirical
facts.
There is, however, a further question, namely, why we should think that in these
cases we’ve revised our understanding of some phenomenon, in light of empiri-
cal evidence, rather than changed the subject. My answer, once again, is simply
that this is the way we do think of what has occurred—and, as Papineau (and
many others) point out, there is no need to think that we’ve changed the subject

36 This need not mean, however, that these judgments have any particular distinctive phenomenology.
37 For example, they may derive from extrapolating from clear cases of the phenomenon in various ways,
or they may be products of natural selection. The important question, however, is not how they arise, but
whether they can be revised in the light of empirical information.

123
4134 Synthese (2013) 190:4117–4136

unless one cannot count as understanding a term except by accepting the theory that
defines it.38
For example, it seems that just this sort of thing goes on both when further reflection
makes it intuitive to drop a component of what we once regarded as ‘criterial’ for the
application of a term, and also when we question whether our ‘criteria’ (e.g. ‘Knowl-
edge is JTB’ or ‘Actions can be free only if the agent could have done otherwise’)
capture the distinctions we were trying to make on the basis of our understanding.
Sometimes further reflection suggests that an addition is required (the knowledge
case?); other times it prompts us to recognize that the criteria do indeed capture the
phenomenon we had in mind (the free action case?), and most important—as I’ve been
arguing here—sometimes exposure to empirical information about the nature of the
world, and to the methodological principles that guide scientific inquiry, can help to
effect these revisions.
Cummins (1998) once disparaged intuitions as being the products of exclusive
contact with ‘friends and family’, rather than of a priori insight into the nature of things.
One way to put his point is that philosophical intuitions need not be ‘encapsulated’
by sub-personal mechanisms to seem irrefutable by empirical evidence; all it takes is
for one to get used to thinking of things in a certain way—and to have this way of
thinking reinforced by others in one’s cohort. But the ‘friends and family’ model could
also explain how the intuitions that prompt one’s category-judgments can change,
imperceptibly, as long as one’s friends and family include intellectual sophisticates,
and—even better—empirical scientists.39 Friends and family, that is, can sometimes
be a force for good.

5 Conclusion

Still, a naturalist might ask, why should judgments based on concrete-case intuitions
be taken seriously at all—and not just brushed away as detritus extruding from the web
of belief? The best answer, I think, appeals to an indispensability argument, namely,
that they are the only source of evidence for philosophical theses about the nature of
things that take account of possibilities that go well beyond what we have observed and

38 As Papineau puts it (forthcoming, p. 15), in endorsing a Ramsey-Carnap-style approach to philosophical

theory construction, ‘if meaning doesn’t depend on what theory is accepted, rejecting a meaning-constituting
theory applies no pressure whatsoever to change meanings. Correspondingly, if you would continue to
maintain that ‘there are Fs’ even though you came to reject T, the obvious inference to draw is that T is not
criterial for Fs…[but]…[e]ven on the Ramsey-Lewis view, speakers could in principle change the meaning
of ‘F’ if they came to reject some criterial T, and so end up saying ‘there are Fs’ even if F has previously
by definition required T’.
39 Externalists claim that the usage of a term by one’s linguistic community can determine what one
means by the term, whether or not one has cognitive access to that usage; meanings, they say, ain’t (or ain’t
exclusively) in the head. But it seems equally trsue that one’s actual exposure to differences in the way
one’s community uses a term can imperceptibly change the way one applies the term, and even explicitly
attempts to explain its meaning. What ain’t in the head, in this case, is any recognition that one is applying
the term any differently than before.

123
Synthese (2013) 190:4117–4136 4135

what is predicted by our empirical theories.40 Granted, one can reject philosophical
inquiry altogether, on the grounds that it’s irrelevant to scientific theory construction.
Or accept it, as Quine does for epistemology, only insofar as it can be regarded as
a ‘branch’ of science. Perhaps that’s what a strict Quinean naturalist must do. But
for weaker souls who think that phenomena such as moral action, freedom, personal
identity—and, for many of us, knowledge—are not in fact natural kinds, but which
nonetheless have a (nominal) nature worth investigating, then the fact—if it is one—
that the intuitions are not completely impermeable by (our recognition of) empirical
information should be reassuring.
It’s a further question, of course, whether this in indeed a fact, since otherwise it
is hard to see how naturalists could even consider taking intuition-backed judgments
to have the evidential weight that I give them. I’ve argued that, though embracing
intuitions later found to conflict with theory has led to some embarrassing philosophical
moments, it seems that over the course of the history of philosophy, the intuitive ‘bad
apples’ lose their force—for example, because we no longer impute certain properties
to the phenomenon we’re trying to analyze (e.g. ‘perfection’ to planetary orbits), or
because (as noted as well by Weatherson) we find a different, intuitively preferable,
way of describing a scenario that prompts us to see it as an instance of, rather than a
counterexample to, some general claim.
To be sure, this is no guarantee that further reflection always leads to progress;
that is an empirical question. And there is no guarantee that we can’t, or someday
won’t, give up the conviction that there is a distinction between doing and allowing, or
knowledge and justified (or reliably produced) true belief, or that free will is compatible
with determinism (if in fact we’re inclined to believe this now). And finally—but,
once again, in my view very rarely—further reflection may provide the seeds of a
more satisfying explanation of why certain sorts of intuition-based judgments can be
discounted, even if their intuitive pull does not fade.
It may be that we’ll eventually have to choose between naturalism and our intuitions
that there are certain things, or certain distinctions in the world. But, to paraphrase
Augustine, ‘give me naturalism, but not yet…do not heal me too soon from an intuitive
conviction that might be able to be satisfied rather than extinguished’—since when it’s
satisfied, it’s a very good thing. So even philosophers sympathetic to naturalism who
want to defend their theories of knowledge, meaning, moral action, free will, personal
identity, and the like would do best to return to their armchairs and get back to work.

Acknowledgments I am grateful to Kadri Vivhelin, Michael Devitt, and an anonymous reviewer for
Synthese for comments and criticisms that have improved both the substance and the presentation of this
paper.

References

Bealer, G. (1996). A priori knowledge and the scope of philosophy. Philosophical Studies, 81(2–3),
121–142.

40 Note, by the way, that this argument is different from the argument made by BonJour, etc. that intuitions
can be evidence for philosophical theses because we have the intuition that this is the case.

123
4136 Synthese (2013) 190:4117–4136

Bechtel, W., & Mundale, J. (2009). Multiple realizability revisited: Linking cognitive and neural states.
Philosophy of Science, 66, 175–207.
Cappelen, H. (2012). Philosophy without intuitions. Oxford: Oxford University Press.
Cummins, R. (1998). Reflections on reflective equilibrium. In W. Ramsey & M. DePaul (Eds.), Rethinking
intuition. Lanham: Rowman and Littlefield.
Devitt, M. (2010a). Naturalism and the a priori, Chapter 12. In Devitt (Ed.), Putting metaphysics first.
Oxford: Oxford University Press.
Devitt, M. (2010b). No place for the a priori, Chapter 13. In Devitt (Ed.), Putting metaphysics first. Oxford:
Oxford University Press.
Devitt, M. (2010c). Intuitions, Chapter 14. In Devitt (Ed.), Putting metaphysics first. Oxford: Oxford
University Press.
Goldman, A. (2010). Philosophical Naturalism and Intuitional Methodology. In Proceedings and Addresses
of the American Philosophical Association (Forthcoming)
Kornblith, H. (2002). Knowledge and its place in nature. Oxford: Oxford University Press.
Levin, J. (2004). The evidential status of philosophical intuition. Philosophical Studies, 121, 193–224.
Levin, J. (2009). Experimental philosophy. In J. Knobe & S. Nichols (Eds.), Critical notice of anthology.
Analysis reviews.
Levin, J. (2011). Imaginability, possibility, and imaginative resistance. Canadian Journal of Philosophy,
41(3).
McGinn, C. (1989). Can we solve the mind–body problem? Mind, 98, 349–366.
Nagel, T. (2000). The psychophysical nexus. In P. A. Boghossian & C. Peacocke (Eds.), New essays on the
a priori (pp. 433–471). Oxford: Oxford University Press.
Papineau, D. (2009). The poverty of analysis. Proceedings of the Aristotelian Society, 83(1), 1–30.
Papineau, D. (forthcoming). The Philosophical Insignificance of A Priori Knowledge. In: Shaffer M., &
Veber M. (eds) New essays on the a priori.
Perry, J. (forthcoming). Wretched Subterfuge.
Schroeder, M. (forthcoming). Reasons First.
Sosa, E. (2005). A defense of the use of intuitions in philosophy. In M. Bishop & D. Murphy (Eds.), Stitch
and his critics. Oxford: Blackwell.
Sosa, E. (2007). A virtue epistemology. Oxford: Oxford University Press.
Sosa, E. (2008). Experimental philosophy and philosophical intuition. In J. Knobe & S. Nichols (Eds.),
Experimental Philosophy. Oxford: Oxford University Press.
Thomson, J. J. (2008). Turning the trolley. Philosophy and Public Affairs, 36(4), 359–374.
Vihvelin, K. (forthcoming). Causes, laws, and free will: Why determinism doesn’t matter. New York: Oxford
University Press.
Wasserman, R. (2011). Intentional action and the unintentional fallacy. Pacific Philosophical Quarterly, 92,
524–534.
Weatherson, B. (2003). What good are counterexamples? Philosophical Studies, 115, 1–31.
Williams, B. (1970). Philosophical review (Vol. 79, No. 2). Durham: Duke University Press.

123
View publication stats