McCain e Poston (2018) - The Mystery of Skepticism

The Mystery of Skepticism
Brill Studies in Skepticism
Editors
Diego Machuca (conicet)

Duncan Pritchard (University of Edinburgh)
Advisory Board
John Greco (Saint Louis University)

John Christian Laursen (University of California, Riverside)
Casey Perin (University of California, Irvine)
Dominik Perler (Humboldt-Universität zu Berlin)
Claudine Tiercelin (Collège de France)
volume 2
The titles published in this series are listed at brill.com/bss

The Mystery of Skepticism
New Explorations
Edited by
Kevin McCain
Ted Poston
leiden | boston
The Library of Congress Cataloging-in-Publication Data is available online at http://catalog.loc.gov
lc record available at http://lccn.loc.gov/2018050758
Typeface for the Latin, Greek, and Cyrillic scripts: “Brill”. See and download: brill.com/brill-typeface.
issn 2215-177x
isbn 978-90-04-39352-3 (hardback)
isbn 978-90-04-39353-0 (e-book)
Copyright 2019 by Koninklijke Brill nv, Leiden, The Netherlands.

Koninklijke Brill nv incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi,
Brill Sense, Hotei Publishing, mentis Verlag, Verlag Ferdinand Schöningh and Wilhelm Fink Verlag.
All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise,
without prior written permission from the publisher.
Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill nv provided
that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive,
Suite 910, Danvers, ma 01923, usa. Fees are subject to change.
This book is printed on acid-free paper and produced in a sustainable manner.

For the skeptic in each of us
∵
Contents
List of Contributors ix
1 The Mystery of Skepticism: an Introduction 1

Kevin McCain and Ted Poston
Part I
Understanding Traditional Skeptical Arguments
2 Procedural Reasons and the Problem of the Criterion 7

Andrew D. Cling
3 Is Cartesian Skepticism Too Cartesian? 24

Jonathan Vogel
4 Hume’s Certain Doubts: Why We Should Worry Too 46

Kevin Meeker
5 Skepticism as a Way of Life 63

Baron Reed
Part II
New Skeptical Challenges
6 Disagreement Skepticism and the Rationality of Religious Belief 83

Jonathan Matheson
7 Pragmatist Anti-Skepticism: at What Price? 105

Scott Aikin and Thomas Dabay
8 Extraordinary Skepticism 128

Earl Conee
9 Internalism, Memory, and Skepticism 161

Brett Coppenger
viii Contents
10 Inferential Internalism and the Problem of Unconscious

Inference 176
Richard Fumerton
11 The Philosopher’s Doom: Unreliable at Truth or Unreliable at

Logic 187
Bryan Frances
Part III
New Responses to Skepticism
12 The Skill Model: a Response to Skepticism about the Intellectual

Virtues 203
Sarah Wright
13 Two Strategies for Explaining Away Skepticism 227

14 Reasoning One’s Way Out of Skepticism 240

Susanna Rinard
Index 265
Contributors
Scott Aikin—Vanderbilt University

Andrew D. Cling—University of Alabama in Huntsville
Earl Conee—University of Rochester
Brett Coppenger—Tuskegee University
Thomas Dabay—Vanderbilt University
Bryan Frances—University of Tartu
Richard Fumerton—University of Iowa
Jonathan Matheson—University of North Florida
Kevin McCain—University of Alabama at Birmingham
Kevin Meeker—University of South Alabama
Ted Poston—University of Alabama
Baron Reed—Northwestern University
Susanna Rinard—Harvard University
Jonathan Vogel—Amherst College
Sarah Wright—University of Georgia
Chapter 1
The Mystery of Skepticism: an Introduction

Skepticism has a long history. So long in fact that one might think that Kant was
not only correct in his time but prophetic of the present state of epistemology
because skepticism continues to be a ‘scandal.’ Despite the lack of consensus
concerning exactly what the skeptical problem is and how best to handle it,
one might be tempted to think there is nothing new to say about skepticism.
The essays in this volume demonstrate that such a thought is misguided. They
are the cutting edge of research on underexplored skeptical challenges, dimen-
sions of the skeptical problematic, and responses to skepticism.
Part I consists of new takes on how traditional skeptical problems should be
understood. Andrew D. Cling kicks things off with a discussion of the problem
of the criterion in Chapter 2. He argues that the problem of the criterion has
even wider applicability than what might have been thought. So wide is its ap-
plicability that, according to Cling, without a solution to the problem of the cri-
terion we cannot even account for the possibility of cognitive autonomy that is
directed at truth. In Chapter 3, Jonathan Vogel examines Cartesian skepticism.
Vogel considers a complaint that some have raised that Cartesian skepticism is
too Cartesian. He argues that, contrary to those who make this complaint, one
cannot get out of the problem of Cartesian skepticism by simply denying vari-
ous idiosyncrasies of Descartes. Next, Kevin Meeker discusses Hume’s thoughts
on demonstrative reasoning (Chapter 4). Meeker argues that the full force of
Hume’s reflections has been underappreciated. He maintains that there is a
neglected “trickle down” argument that Hume employs, which threatens all
of our knowledge. Baron Reed completes Part I by arguing in Chapter 5 that in
order to be adequate any consideration of skepticism must address the ques-
tion of value in epistemology. In particular, Reed considers what can be said
for skepticism as a way of life. While he concludes that the way of life offered
by Pyrrhonian skepticism leaves much to be desired, the Academic skeptical
tradition points toward a way of life that is more admirable than either that of
the Pyrrhonists or the dogmatists.
The focus shifts from traditional skeptical problems to new skeptical
challenges in Part ii. Jonathan Matheson opens the discussion of these new
challenges with an application of the Equal Weight View of the epistemic
© koninklijke brill nv, leiden, ��9 | doi:10.1163/9789004393530_002

2 McCain and Poston
significance of peer disagreement to religious disagreements (Chapter 6). Ac-

cording to Matheson, the Equal Weight View implies that the current state of
religious disagreement gives us a defeater for any non-skeptical view concern-
ing the question of God’s existence. Nonetheless, Matheson argues that this
skeptical conclusion may not mean that religious beliefs fail to be rational.
Scott Aikin and Thomas Dabay take on the pragmatist response to skepticism
in Chapter 7. They argue that while it might seem that pragmatism offers a
good way of being both a fallibilist and an anti-skeptic, this is not so. Ultimate-
ly, Aikin and Dabay conclude that there are significant costs to the pragmatist
anti-skeptical program including giving up fallibilism. In Chapter 8 Earl Conee
explores “extraordinary skepticism”—skeptical challenges that do not appeal
to the possibility of mass deception. He considers arguments for four kinds
of extraordinary skepticism: there is no fallibly justified knowledge, we lack
knowledge because of localized skeptical possibilities, Gettier problems are so
pervasive that we lack knowledge, and our beliefs are false. Conee argues that
none of these skeptical challenges are insurmountable. After this, Brett Cop-
penger examines the skeptical threat to memorial justification (Chapter 9).
In particular, he considers the nature of this skeptical threat and how three
prominent forms of internalism (traditional, phenomenal conservatism, and
evidentialism) might try to respond. The end result, according to Coppenger,
is that while considerations of memory skepticism do not show that internal-
ism is false, there is much work for internalists to do if they hope to respond to
the skeptical threat to memorial justification. Next, Richard Fumerton consid-
ers the prospects for accepting very strong requirements for justification in
Chapter 10. Fumerton admits that the more that is required for justification
the more likely it is that skepticism is correct. Nevertheless, he argues that one
might commit to very strict standards for ideal justification while leaving room
for less ideal forms of justification, which can perhaps avoid skepticism. Final-
ly, Bryan Frances closes this part of the volume by arguing that philosophers
face a serious problem (Chapter 11). We are either highly unreliable at finding
truth or we are highly unreliable at spotting validity. Frances concludes that
there does not seem to be a good way out of this problem, so we seem to lack a
lot of knowledge that we tend to think we have.
The volume closes with Part iii, which offers some hope against the threat
of skepticism by including three new responses to this problem. In Chapter
12, Sarah Wright responds to skepticism about virtue. If virtue epistemologists
are correct that knowledge requires the exercise of intellectual virtues, then
skepticism about the existence of virtues leads to skepticism about knowl-
edge. Wright responds to this skepticism concerning virtues by defending a
skill model of the nature of virtue. This is followed by Kevin McCain’s and Ted
The Mystery of Skepticism: An Introduction 3
Poston’s exploration of explanationist responses to skepticism in Chapter 13.

According to McCain and Poston, there are at least two different explanation-
ist approaches to the problem of skepticism, one is Cartesian and the other is
naturalist. In the end, McCain and Poston argue that both varieties of expla-
nationism hold promise for responding to skepticism, though they each have
difficulties to overcome. In the final chapter of the volume (Chapter 14), Su-
sanna Rinard argues that it is possible to rationally persuade an external world
skeptic that we have knowledge of the external world. The reason for this,
Rinard maintains, is that it is possible show that premises that the external
world skeptic accepts entail a more radical skepticism about complex reason-
ing. Rinard argues that skepticism about complex reasoning is self-defeating,
and the only way that the external world skeptic can escape this self-defeat is
by abandoning external world skepticism.1
1 Thanks to Siddharth Srikakolapu for editorial assistance in formatting this volume.

Part i
Understanding Traditional Skeptical Arguments
∵
Chapter 2
Procedural Reasons and the Problem

of the Criterion
Andrew D. Cling
1 Introduction
To know a proposition, we must first know a criterion of truth but to know a crite-
rion of truth, we must first know some proposition. Therefore we cannot know any
proposition or any criterion of truth. That, roughly, is the problem of the crite-
rion. We should, however, improve on this statement of the problem. First, to
make these premises plausible we need to make it explicit that they are claims
about the kinds of reasons we need in order to have justified beliefs. The prob-
lem is also more general because it can be about epistemic values other than
knowledge. Finally, the problem depends on assumptions about circles of
reasons—“we must first know”—and about infinite regresses of reasons that
need to be made explicit.
The problem of the criterion is best understood as a family of epistemic
regress problems. Interpreted in this way, it is about the role that criteria
of truth must play in providing us with epistemic reasons for belief. All
versions of the problem fit the following schema, in which “C” is a variable
for propositions that express criteria of truth, “P” is a variable for other
propositions, and “justified” is a placeholder for terms that express positive
epistemic value:
(1a) A proposition P can be justified for a person S only if there is a criterion

of truth C such that (i) C is justified for S and (ii) C is an epistemic reason
for S for believing P.
(1b) A criterion of truth C can be justified for a person S only if there is a prop-
osition P such that (i) P is justified for S and (ii) P is an epistemic reason
for S for believing C.
(2) No proposition or criterion can be justified if that requires a circle of jus-
tified epistemic reasons for belief.
(3) No proposition or criterion can be justified if that requires an infinite
sequence of justified epistemic reasons for belief.
(4a) ∴ No proposition can be justified.

8 Cling
(4b) ∴ No criterion can be justified.1
This schema makes it explicit that the problem is about reasons. It also displays
the necessary injunctions against circles and infinite regresses of reasons.
This way of understanding the problem depends on a distinction between
propositions that express criteria of truth and other propositions. So one chal-
lenge for understanding the problem in this way is to explain that distinction.
Another challenge is to show that (1a) is plausible by explaining how criteria of
truth can be reasons for belief. A related challenge is to show why justification
might require such reasons. My goal in this paper is to answer these challenges.
I argue that criteria of truth can be procedural reasons for beliefs. Procedural
reasons for belief are factors that count in favor of believing propositions be-
cause we take them to indicate or imply that our belief-forming procedures—
our ways of believing—are reliable. I argue that if epistemic justification
requires justified procedural reasons for belief, then justification requires justi-
fied criteria of truth, as (1a) says. I close by sketching an argument for the claim
that one important kind of justification requires justified procedural reasons.
2 Criteria of Truth
Criteria of truth are principles for identifying true propositions. There are at
least three kinds of criteria of truth. Paradigmatic criteria of truth are prin-
ciples that purport to give sufficient conditions for truth. Paradigmatic criteria
can be limited to a specific domain and might give only probabilistic condi-
tions of truth. Propositions stating that belief-forming procedures are reliable
(reliability claims) and, under some conditions, principles that give sufficient
conditions for positive epistemic properties (epistemic principles) are also cri-
teria of truth because they imply paradigmatic criteria. Different versions of
the problem of the criterion result from taking premises (1a) and (1b) to be
claims about the role these kinds of criteria play in epistemic justification.
2.1 Paradigmatic Criteria of Truth

Paradigmatic criteria of truth are generalizations that give sufficient condi-
tions for truth. Detection criteria are paradigmatic criteria according to which
having a specified property X is sufficient for truth:
1 (1a) and (1b) jointly entail that justified propositions and justified criteria of truth depend
upon circles or infinite sequences of justified epistemic reasons. (2) and (3) jointly entail that
justification cannot depend upon circles or infinite sequences of justified epistemic reasons.
So if (1a), (1b), (2), and (3) were true, (4a) and (4b) would have to be true.
Procedural Reasons and the Problem of the Criterion 9
(DC) Necessarily, for all propositions P, if P has X, then P is true.
Inference criteria are paradigmatic criteria according to which standing in a

specified relation R to true propositions is sufficient for truth:
(IC) Necessarily, for all P and Q, if P is true and P stands in R to Q, then

Q is true.
Let us say that a proposition P is directly or indirectly connected by relation R

to a proposition Q just in case there is a finite sequence of propositions that
begins with P, ends with Q, and every component of which stands in relation R
to its successor, if it has one.2 Taken together, (DC) and (IC) imply this hybrid
paradigmatic criterion of truth:
(HC) Necessarily, for all P, if P has X or there is a Q such that Q has X

and Q is directly or indirectly connected by R to P, then P is true.
(HC) follows because (DC) implies that propositions that have X are true and
(IC) implies that R is truth-preserving.
The problem of the criterion is about the kinds of reasons we need in order
to have justified beliefs. It is also about the conditions that must be satisfied
in order for us to be justified in believing principles for identifying true propo-
sitions. In addition to the paradigmatic criteria I have discussed so far, the
problem also applies to principles giving sufficient conditions for truth that
are fallible, that are limited to specific domains, or that are neither logical nor
conceptual truths. To capture the general problem of the criterion, therefore,
we should expand the class of paradigmatic criteria of truth to also include
principles that give only probabilistic conditions of truth, principles that apply
only in limited domains, and principles with various kinds of alethic modality.
An example of a detection criterion is Descartes’s principle of clear and dis-
tinct perception:
2 Another way to understand the connection condition I have expressed by means of the ex-
pression “directly or indirectly connected by relation R” is by means of an ancestral relation
R* defined recursively as follows:
For all x and y, if x stands in R to y, then x stands in R* to y.
For all x and y, if x stands in R* to y and y stands in R* to z, then x stands in R* to z.
No objects are related by R* except in virtue of (i) and (ii).
Given this definition, we may say that objects a and b are directly or indirectly connected
by R if, and only if, a stands in R* to b.
10 Cling
(CD) Necessarily, for all propositions P, if I clearly and distinctly per-

ceive that P, then P is true.
An example of an inference criterion is this principle that would underwrite

some non-deductive arguments:
(SS) If x% of observed Ss have been Ps under the relevant conditions,

then (probably) about x% of Ss are Ps.3
One version of the problem of the criterion is about the role that paradigmatic
criteria play in epistemic justification. On this interpretation, (1a) says:
(1aP) proposition P can be justified for a person S only if there is a

A
paradigmatic criterion of truth CP such that (i) CP is justified for S
and (ii) CP is an epistemic reason for S for believing P.
One aspect of this version of the problem is whether justification requires that
a detection criterion be justified for us. Another aspect is whether we can be
justified in believing some propositions on the basis of others only if an infer-
ence criterion is justified for us. Every version of the problem of the criterion
raises questions about whether the corresponding kind of epistemic c ircularity
is vicious. So an aspect of this version of the problem is whether a proposition
that is supported by a paradigmatic criterion of truth can be a good reason for
believing that very criterion.
2.2 Reliability Claims

A belief-forming procedure is a set of dispositions to activate—form or s ustain—
beliefs as outputs given features of our points of view as inputs.4 The inputs to
belief-forming procedures are features of our points of view including content
provided by sensory experiences, testimony, beliefs we already have, rational
3 This is a placeholder for a fully adequate principle that would underwrite enumerative
inductions.
4 See Alston (1993: 4–5). I think that that a factor R is a reason for believing a proposition P
only for persons who are disposed to endorse the disposition to believe P given R. So belief-
forming procedures that involve reasons must be complex sets of dispositions (i) to activate
specific beliefs as outputs given features of our points of view as inputs and (ii) to endorse
those belief-forming dispositions in ourselves and others. I am not sure whether they must
also include dispositions to negatively sanction contrary belief-forming dispositions. The idea
of representing belief-forming methods as complex systems of dispositions in this way is
due, in part, to Haugeland’s account of Heidegger’s conception of norms (1982: 15–17).
intuitions, memories, and so on. We have, for example, dispositions to form

beliefs about observable things given different kinds of sensory experiences.
We have dispositions to form new beliefs by means of deductive, inductive,
and explanatory inferences from propositions we already believe. And so on.
Call any proposition to the effect that belief-forming procedure M is reliable a
reliability claim. Reliability claims are criteria of truth because they imply para-
digmatic criteria. So one version of the problem of the criterion is about the
role that reliability claims play in epistemic justification.
Let me explain. A proposition to the effect that belief-forming procedure M
is reliable implies that most of the beliefs that would be activated by means of M
would be true.5 So M is reliable implies:
(MR) ecessarily, for all P, if the belief that P is activated by procedure

N
M, then P is (probably) true.
(MR) results from (DC) by taking Χ to be the property of being the content of
a belief that is activated by procedure M and by allowing Χ to be a probabilistic
condition of truth. So reliability claims imply paradigmatic criteria. The con-
verse is also true. Detection and hybrid paradigmatic criteria of truth imply
that belief-forming procedures that conform to them are reliable.6
We can also pose the problem of the criterion, therefore, in terms of the role
that reliability claims play in epistemic justification. On this way of stating the
problem, premise (1a) says:
(1aR) A proposition P can be justified for a person S only if there is a

reliability claim CR such that (i) CR is justified for S and (ii) CR is
an epistemic reason for S for believing P.
An aspect of this version of the problem is whether having justified beliefs

requires that we be justified in believing that our belief-forming procedures
are reliable. A question about epistemic circularity posed by this version of
the problem is about bootstrapping: can reasons that are activated by a belief-
forming procedure M be good reasons for believing that M itself is reliable?7
5 Alston, for example, says that a belief-forming mechanism is reliable if it “would yield mostly
true beliefs in a sufficiently large and varied run of employments in situations of the sorts we
typically encounter” (1993: 9, his italics).
6 Inference criteria of truth imply that belief-forming procedures that conform to them are
conditionally reliable in the sense that they (usually) yield true beliefs given true inputs.
7 Weisberg (2012) gives a useful overview of the issues connected to bootstrapping.
12 Cling
2.3 Epistemic Principles

Epistemic principles state sufficient conditions for normative epistemic prop-
erties such as being a case of knowledge, being certain, being justified, and so on.
Epistemic principles have this form, where Φ is a normative epistemic property:
(E) Necessarily, for all P, if _____, then P has Φ.8
There are at least two kinds of epistemic principles. Generation principles

state non-epistemic sufficient conditions for normative epistemic properties.
Where C is a descriptive property and Φ is a normative epistemic property,
generation principles have this form:
(G) For all P, if P has X, then P has Φ.
This principle of Chisholm’s is a generation principle:
(M1) If the property of being F is self-presenting, if S is F, and if S be-

lieves himself to be F, then it is certain for S that he is F.9
Another generation principle is this proposition about the connection be-

tween coherence and justification:
(CJ) If P coheres with the system of propositions accepted by S, then P

is justified for S.10
Transmission principles state that propositions have a normative epistemic

property Φ if they stand in specified relationships to other propositions that
have Φ.11 Simple transmission principles have this form:
8 My account of epistemic principles is largely due to Van Cleve (1979: 75) although my ac-
count of generation principles is slightly different.
9 Chisholm (1989: 19).
10 (CJ) is discussed by Van Cleve (1979: 75, n. 45) who counts it as a hybrid principle that
is neither a generation principle nor a transmission principle in his sense. But if gen-
eration principles give non-epistemic sufficient conditions for epistemic properties and
coherence is non-epistemic, then (CJ) is a generation principle. Van Cleve seems to be
assuming that the non-epistemic property Χ in generation principles cannot involve in-
ferential relations between propositions.
11 Still other epistemic principles might be mixed. Mixed principles would state that propo-
sitions have a normative epistemic property Φ, if they stand in specified relationships to
propositions that have a different normative epistemic property Ψ.
(T) For all P and Q, if P has Φ and P stands in relation R to Q, then Q has
Φ.
An example of a transmission principle is this proposition to the effect that

knowledge is transmitted by known entailment:
(tke) For all P and Q, if S knows that P, S knows that P entails Q, and S
believes that Q on the basis of P, then S knows that Q.
Propositions that have epistemic value Φ are either (probably) true, or they
are not (probably) true. If propositions that have Φ are not (probably) true,
then epistemic principles are not criteria of truth and the problem of the crite-
rion is not about them.12 If propositions that have Φ are (probably) true, then
epistemic principles imply paradigmatic criteria of truth. For a g eneration
principle with form (G) and the assumption that propositions that have epis-
temic property Φ are (probably) true:
12 Consider
(ΦT) Propositions that have Φ are (probably) true.
If (ΦT) is false, then “the problem of the criterion” about epistemic principles would be
fairly easy to solve. For then the relevant version of (1a) would be:
(1aE) A proposition P can be justified for person S only if there is an epistemic principle
CEP such that (i) CEP is justified for S and (ii) CEP is an epistemic reason for S for
believing P.
If (ΦT) is false, then (1aE) is false. For if (ΦT) is false, then a criterion of truth is not an
epistemic reason for believing a proposition that satisfies its condition Χ. For an epis-
temic reason for believing P is a factor that counts in favor of believing P because we
take it to imply or to indicate that P is true. If (ΦT) is false, then epistemic principles do
not imply that their target propositions are true. They imply only that it is permissible,
obligatory, virtuous, or in some other way good to believe those propositions. If we try
to solve this problem by modifying (1aE) to eliminate reference to epistemic reasons for
belief, it becomes:
(1aE′) A proposition P can be justified for person S only if there is an epistemic principle
CEP such that (i) CEP is justified for S and (ii) CEP is a non-epistemic reason for S for
believing P.
(1aE′) says that epistemic justification requires non-epistemic reasons for belief. This is
not plausible. Furthermore, there is no evident problem about circularity here. For it is
not obvious that a proposition P cannot be a good non-epistemic reason for believing a
proposition Q when Q is a good epistemic reason for believing P. The upshot of these
considerations is that the problem of the criterion is not plausible if it is about epistemic
principles and (ΦT) is false. (I take no official view about whether (ΦT) is true.) What is
plausible, however, is a special case of the problem of the criterion according to which
epistemic principles are epistemic reasons for singular epistemic propositions and vice
versa. I discuss that argument below.
14 Cling
(ΦT) For all P, if P has Φ, then P is (probably) true
jointly entail a paradigmatic criterion:
(RT) For all P, if P has X, then P is (probably) true.
Likewise, any set of propositions that includes both a generation principle

with form (G) and a transmission principle with form (T) implies that every
proposition that has X or is directly or indirectly connected by relation R to a
proposition that has X is (probably) true. For suppose that P has X and that P
stands in relation R to Q. By (G), P has Φ. Hence, by (T), Q has Φ. Suppose that
Q stands in relation R to proposition V. Given (T), V has Φ. And so on. What
follows is this hybrid epistemic principle:
For all P, if P has X or there is a Q such that Q has X and Q is directly or

indirectly connected by relation R to P, then P has epistemic value Φ.
Given (ΦT)—propositions that have Φ are (probably) true—a hybrid paradig-

matic criterion of truth follows:
(HC) F or all P, if P has X or there is a Q such that Q has Χ and Q is di-

rectly or indirectly connected by relation R to P, then P is (prob-
ably) true.
If (ΦT) is true, then one version of the problem of the criterion is about the
role of epistemic principles in epistemic justification. On this interpretation,
(1a) says:
(1aE) A proposition P can be justified for person S only if there is an

epistemic principle CEP such that (i) CEP is justified for S and (ii)
CEP is an epistemic reason for S for believing P.
(1aE) says—implausibly—that a proposition P is justified only if a justified

epistemic principle is a reason for believing P. So (1aE) requires that we be jus-
tified in believing a principle about justification in order for any proposition to
be justified for us. This is implausible because epistemic principles are stron-
ger than paradigmatic criteria and reliability claims. For if (ΦT) is true, then
epistemic principles imply the corresponding paradigmatic criteria and reli-
ability claims, but not vice versa. A more plausible principle in the neighbor-
hood says only that any epistemic proposition—a singular proposition about
the epistemic value that a proposition has for a person—can be justified only
if it is supported by a justified epistemic principle. That principle is an impor-
tant assumption in a special case of the problem of the criterion that is posed
by this argument for skepticism about epistemic propositions and epistemic
principles:
(1aE*) A
n epistemic proposition PE can be justified for person S only if
there is an epistemic principle CEP such that (i) CEP is justified for
S and (ii) CEP is an epistemic reason for S for believing PE.
(1bE*) An epistemic principle CEP can be justified for a person S only if
there is an epistemic proposition PE such that (i) PE is justified for
S and (ii) PE is an epistemic reason for S for believing CEP.
.
.
.
(4aE*) ∴ No epistemic proposition can be justified.
(4bE*) ∴ No epistemic principle can be justified.
(1aE*) is plausible since it is plausible that we can be justified in believing an

epistemic proposition such as Frank is justified in believing that the cat is on
the mat only if we are justified in believing an epistemic principle stating a suf-
ficient condition for justification that the cat is on the mat satisfies for Frank.
Likewise (1bE*) is plausible since it seems that epistemic principles can be sup-
ported only by the epistemic propositions they entail.13 This argument is not,
however, an argument for general skepticism and should, I think, worry only
epistemologists.
3 Procedural Reasons
A reason for a person S for believing a proposition P is a factor that counts in

favor of S’s believing P.14 An epistemic reason R for S for believing P is a factor
that counts in favor of S’s believing P because S takes R, perhaps together with
13 Many contemporary epistemologists seem to identify the problem of the criterion with
this metaepistemological variant. See, for example, Chisholm (1973), Van Cleve (1979),
and Steup (1992, 1995). For discussion of metaepistemological variants of the problem of
the criterion see Cling (1997).
14 We may, in turn, explain counting in favor of in terms of our dispositions to make and to
endorse belief-forming procedures. So R counts in favor of believing P for S if, and only if,
S is disposed to believe that P given R and S is disposed to endorse forming the belief that
16 Cling
other factors, to imply or to indicate that P is true.15 A procedural reason for S

for believing P is a type of epistemic reason. It is a factor R that, perhaps to-
gether with other factors, S takes to imply or to indicate that P is true because
S takes R to imply or to indicate that the way in which S believes P is reliable.
To have a procedural reason for believing a proposition P is to have a reason to
trust one’s own reliability with respect to P.
Procedural reasons are factors that are relevant to the reliability of our ways
of believing. What is a way of believing? What is it for a way of believing to
be reliable? What must we believe in order to take a way of believing to be
reliable?
In one sense, a way of believing is a general process, method, rule, practice, or
procedure by means of which a person activates—forms or sustains—beliefs,
perhaps about a more-or-less delimited subject. So, for example, we may speak
of the ways in which persons activate beliefs about macroscopic objects, about
politics, about mathematics, about the past, about God, and so on. A belief-
activating procedure is reliable just in case it is a sufficiently effective means
to the goal of believing what is true and avoiding belief in what is false. Hence,
one thing it can mean to say that S’s way of believing that P is reliable is that the
general procedure by means of which S activates the belief that P is reliable.
Can we say that the way in which a person believes a particular proposition
is reliable without making reference to a general belief-forming procedure?
Can we take our ways of believing to be reliable without having beliefs about
general belief-forming procedures or without having the concept of reliability
itself? Yes, we can. If a general belief-forming procedure is reliable, then the
particular beliefs that are activated by means of that procedure are reliably
P given R. Perhaps S must also be disposed to negatively sanction contrary belief-forming

dispositions.
15 I include “indicate” in addition to “imply” in my account of reasons to leave it open that
non-propositional contents can be reasons for belief. Thus, that there is smoke (induc-
tively) implies that there is fire but smoke indicates that there is fire.
It follows from my account that a non-epistemic reason R for S for believing P is a factor
that counts in favor of believing P but that S does not take to imply or to indicate that P is
true. My account of epistemic reasons for belief is intended to solve the problem posed
by this argument: (1) An epistemic reason to believe that P is (=) a reason to believe that P
is true, (2) any reason to believe that P is a reason to believe that P is true, (3) therefore, any
reason to believe that P is an epistemic reason to believe that P. If this argument is sound,
there is no distinction between epistemic and non-epistemic reasons for belief. I reject
(1) because although all epistemic reasons to believe that P are reasons to believe that P is
true, not all reasons to believe that P is true are epistemic reasons to believe that P. On my
view, an epistemic reason to believe that P is a special kind of factor that counts in favor
of believing P: a factor we take to indicate or to imply that P.
connected—attuned—to the corresponding states of affairs. For from the

proposition that a belief-forming procedure M is reliable and the proposition
that S activates a belief in a given proposition A by means of M, it follows that:
(PR1) If A were false, then S would (probably) not believe A
and that:
(PR2) If S were to believe A, then A would (probably) be true.16
These implications show the relationship between general reliability claims,

on the one hand, and propositions such as (PR1) and (PR2) that are about the
attunement of particular beliefs to the corresponding states of affairs, on the
other. Because of this relationship, (PR1) and (PR2) show how we can mean-
ingfully say that—and have a relevant belief that—the way in which a person
believes a particular proposition is reliable without making reference to any
general belief-forming procedure and without using a general concept of reli-
ability. For in the limiting case of a specific proposition A, we can describe S’s
way of believing A as reliable without committing ourselves to any proposi-
tions about a general method by means of which S accepts, rejects, or with-
holds A and without using a general concept of reliability. For, in this case, we
can describe S’s way of believing A as reliable by saying that propositions such
as (PR1) and (PR2) are true.17
To believe that one’s way of believing P is reliable is to believe propositions
such as (PR1) and (PR2). Because they are partly constitutive of the r eliability
16 (PR1) is about what has been called the safety of belief and (PR2) is about what has been
called the sensitivity of belief. Some philosophers believe that one or the other of these
conditions must be satisfied if our beliefs are to be cases of knowledge (see Black (2011)
for a good survey). I am interested, instead, in whether (PR1) and (PR2) might be kinds of
reasons for belief. In particular, I am interested in the role they might play in our beliefs
about the reliability of our ways of believing. So I am raiding externalist scripture to do
internalist devil’s work.
17 (PR1) and (PR2) are, however, relevant only to the aspect of reliability that concerns the
avoidance of false beliefs. Avoiding false beliefs is only one aspect of reliability, however.
Another aspect is succeeding in believing the truth. It would be a mistake, I think, to say
that persons are reliable detectors of enemy bombers provided only that they minimize
the number of their false reports of enemy bombers. To be reliable, they must also be
able to detect enemy bombers in a significantly high number of cases in which enemy
bombers are present. In light of my discussion of procedural reasons for belief, this opens
up the possibility that subjunctives that are relevant to the identification of the truth—if
P were true, I would believe P and if I were to fail to believe P, P would not be true—are also
procedural reasons for belief. I shall not investigate this complication here.
18 Cling
of S’s way of believing P, I shall call propositions such as (PR1) and (PR2) spe-
cific procedural reasons for S for believing P. Unlike evidential reasons, specific
procedural reasons do not normally indicate or imply their target proposi-
tions by themselves. Instead, they can be epistemic reasons for believing target
propositions because we take them to indicate or to imply that we are properly
attuned to the corresponding states of affairs. The procedural reasons for S for
believing that P include S’s specific procedural reasons for believing P together
with any reasons S has for believing those specific procedural reasons.
When we believe a proposition P we commit ourselves to the specific pro-
cedural reasons for believing P whether or not we believe those procedural
reasons. This commitment is shown by the fact that a reason for doubting a
specific procedural reason for believing P is a reason for withholding P.18 Sup-
pose, for example, I believe that:
(AA) I am an above-average employee.
Suppose that I have no evidence for (AA). Also suppose that I study some
psychological research and come to believe, on that basis, that persons tend
to overestimate the relative quality of their performance of relevant tasks
because they tend to ignore evidence indicating that the quality of their per-
formance is not above average.19 This research can give me a reason for with-
holding (AA) despite the fact that I do not take it to indicate or to imply that
(AA) is false. Nor is my reason for withholding (AA) the fact that the research
gives me a reason for doubting my evidence for (AA). By hypothesis, I have no
such evidence. Instead, the research gives me a reason for withholding (AA)
because it gives me a reason for doubting that if I were not an above-average em-
ployee, I would not believe that I am an above-average employee or for d oubting
that if I were to believe that I am an above-average employee, then I would be an
above-average employee. Thus it is a reason for withholding (AA) because it is
a reason for doubting a specific procedural reason for this proposition to which
I am committed whether I believe that reason or not.
By the same token, having good reasons for a procedural reason for believ-
ing P might improve the epistemic standing that P has for us. This is so because
persons are in a better epistemic position with respect a proposition if they
have reasons for believing that their way of believing it is reliable. Knowing
18 I conjecture that this principle is correct: R is a reason for S for doubting P just in case
R is a reason for S for believing not P or R is a reason for S for rejecting or withholding a
reason for believing P. If it is correct, this principle would underwrite my argument for
procedural reasons.
19 Taylor and Brown (1988).
that the hiring committee has followed a rigorous and reliable screening pro-
cedure, for example, can give each member of the committee a good reason
for believing that the employee who was hired on the basis of that procedure
is the best-qualified applicant. This is so despite the fact that propositions
about the reliability of the selection procedure will not normally be directly
relevant to the qualifications of any particular applicant or to the ranking of
the applicants. Likewise, justifiably believing that one has gathered evidence
for an empirical hypothesis H in a way that minimizes bias can give a person
a good reason to accept the results of an investigation about whether H is true
whether or not that evidence implies that H is true, implies that H is false, or
is inconclusive.20
Because paradigmatic criteria of truth imply that there is a reliable connec-
tion between propositions that have (or are suitably related to propositions
that have) the relevant feature X and the corresponding states of affairs, crite-
ria of truth are ideally suited to support specific procedural reasons for belief.
This shown by this argument pattern:
(i) Necessarily, propositions that have X are true. [Premise]21

(ii) I would believe that A only if A had X and if A did not have Χ I
would not believe A. [Premise]
(PR1) ∴ If A were false, then I would not believe that A. [From (1) & (2)]
and
(PR2) ∴ If I were to believe that A, then A would be true. [From (1) &
(2)]
This argument pattern shows that paradigmatic criteria of truth are well suited
to provide us with reasons for believing specific procedural reasons for belief.
Premises corresponding to (i) are paradigmatic criteria of truth. Premises cor-
responding to (ii) are claims about ways of believing A: I believe A because it
has Χ. If (i) and (ii) were true, then the specific procedural reasons expressed
by (PR1) and (PR2) would have to be true.
20 For example, with respect to hypotheses with the form All Ss are Ps there are selection
procedures for picking out the Ss to be examined and for determining whether or not an
S is a P. Propositions about selection procedures are one kind of procedural reason for
belief. The concept of a selection procedure is due to Achinstein (2001: 40, 200–201).
21 To avoid what I hope are irrelevant complexities, I consider only the case of an exception-
less paradigmatic criterion of truth. For criteria that give only probabilistic conditions of
truth, the logic will be more complex. Also for simplicity I omit reference to the clause
about being directly or indirectly connected to a proposition that has X.
20 Cling
The converse also holds: unless criteria of truth are among our reasons, we
do not have good reasons for believing our specific procedural reasons. My spe-
cific procedural reasons for believing (AA) include:
(PR1AA) I f it were false that (AA) I am an above-average employee, then

I would not believe that (AA) I am an above-average employee.
and
(PR2AA) I f I were to believe that (AA) I am an above-average employ-

ee, then it would be true that (AA) I am an above-average
employee.
(PR1AA) and (PR2AA) are propositions about the patterns exhibited, across a
range of possible situations, by the state of affairs represented by (AA) and
my belief that (AA). (PR1AA) is true just in case the relevant possible situations
in which (AA) is false are situations in which I do not believe (AA). (PR2AA)
is true just in case the relevant possible situations in which I believe (AA)
are situations in which (AA) is true. The mere fact that I believe (AA) does
not give me a good reason for believing (PR1AA) or (PR2AA). If it did, then
everyone would have a good reason for believing that every way of believing
any proposition is reliable. Unless the patterns that would make (PR1AA) and
(PR2AA) true are mysterious, there must be some feature X of my epistemic
predicament with respect to (AA) that explains why (PR1AA) and (PR2AA) are
true. To explain (PR1AA) and (PR2AA), feature X must be a characteristic of my
predicament with respect to (AA) that is sufficient for the truth of (AA) and
necessary for my believing that (AA). For unless X is sufficient for the truth
of (AA) and also necessary for my believing (AA), X would not explain why
I fail to believe (AA) when it is false, as (PR1AA) requires. Likewise, unless X
is sufficient for the truth of (AA) and necessary for my believing that (AA), X
would not explain why the relevant situations in which I believe (AA) are also
situations in which (AA) is true, as (PR2AA) requires. If neither of these things
were true—if (AA) does not satisfy a sufficient condition of truth that is also a
necessary condition for my believing (AA)—then (PR1AA) and (PR2AA) would
be utterly mysterious. For in this case there would be no factor that accounts
for my failure to believe (AA) when it is false or for (AA)’s truth in cases in
which I believe it.
This argument does not yet show that paradigmatic criteria of truth must be
among our reasons for believing specific procedural reasons. It shows that we
can have a good reason for believing (PR1AA) and (PR2AA) only if our reasons
include a proposition at least as strong as this:
(I*)Necessarily, if (AA) has X, then (AA) is true.
But (I*) is not general and, therefore, does not express a paradigmatic criterion
of truth.
The problem with (I*) is that because it is a singular proposition, it does not
give me a good reason for believing (PR1AA) or (PR2AA). Suppose that, for all I
know, (AA) has X but that most other similar propositions that have X under
similar conditions are not true. Under these conditions, (I*) does not give me a
good reason for believing (PR1AA) or for believing (PR2AA).
For unless it is generally the case that propositions that have X are true, the
fact that (AA) has X does not explain why (PR1) or (PR2) are true. For X to
explain the reliability of my way of believing (AA) it must be a general truth
that propositions that have X are true. This, however, means that in order to
have a good reason for believing specific procedural reasons, I must believe a
paradigmatic criterion of truth.
These considerations support two conclusions. First, paradigmatic criteria
of truth are the sorts of propositions that might provide us with good reasons
for believing specific procedural reasons. Second, paradigmatic criteria must be
among our reasons for believing specific procedural reasons. Because reliability
claims and, under some conditions, epistemic principles imply paradigmatic cri-
teria of truth, all three kinds of criteria of truth are the sorts of propositions that
might give us good reasons for believing specific procedural reasons.
4 Conclusion
It follows from my account that if epistemic justification requires justified

procedural reasons for belief, then it requires justified criteria of truth, as (1a)
says. I favor pluralism about epistemic justification because I find it plausible
that there are different worthwhile cognitive goals and that there are differ-
ent properties that make it epistemically permissible, obligatory, virtuous, or
in some other way epistemically good for us to believe propositions in light
of these goals. Because of this, I think an important question to ask about
(1a) is whether there is a type of epistemic justification about which it is true
and thus, whether there is a type of epistemic justification that is threatened
by the problem of the criterion. I also think that we should understand the
22 Cling
significance of different types of epistemic justification—the value of epis-

temic v alues—in light of broadly ethical conceptions of the sorts of things that
would make life meaningful. So we should ask about the kinds of justified be-
liefs we need in order to have different kinds of meaningful lives. I conclude by
sketching an argument for the view that one kind of meaningful life requires
that we have procedural reasons for our beliefs.
I find it plausible that one kind of meaningful life requires that we exercise
autonomy in pursuit of our goals. Autonomy, however, requires not only that
we understand and endorse our goals but also that we understand and endorse
our means to those goals. One valuable goal is to believe what is true while
avoiding belief in what is false. So we can exercise autonomy in pursuit of this
goal only if we believe it to be worthwhile and we believe that our means for
achieving it—our ways of believing—are reliable. This, as I have argued, re-
quires that we have specific procedural reasons for our beliefs. Because these
reasons must be justified in the same way, this kind of justification requires
justified criteria of truth that are reasons for our beliefs. If sound, this argu-
ment would show that without a solution to the problem of the criterion, we
cannot account for the possibility of truth-directed cognitive autonomy.22
Bibliography
Achinstein, P. 2001. The Book of Evidence. New York: Oxford University Press.
Alston, W. 1993. The Reliability of Sense Perception. Ithaca, NY: Cornell University Press.
Black, T. 2011. “Modal and Anti-Luck Epistemology.” In Sven Bernecker and Duncan
Pritchard (eds.), The Routledge Companion to Epistemology, 187–198. New York:
Routledge.
Chisholm, R. 1972. The Problem of the Criterion. Milwaukee: Marquette University Press.
Chisholm, R. 1989. Theory of Knowledge, third edition. Englewood Cliffs, N.J.: Prentice-
Hall.
Cling, A. 1997. “Epistemic Levels and the Problem of the Criterion,” Philosophical Stud-
ies 88: 109–140.
Haugeland, J. 1982. “Heidegger on Being a Person,” Noûs 16: 15–26.
Steup, M. 1992. “Problem of the Criterion.” In J. Dancy and E. Sosa (eds.), A Companion
to Epistemology, 378–381. Cambridge, MA: Blackwell.
22 Thanks to Ted Poston and Kevin McCain for very helpful comments on an earlier version
of this essay. Thanks also to audiences at the 2016 meetings of the Alabama Philosophical
Association and the Tennessee Philosophical Association. Special thanks to Scott Aikin
for his hospitality and helpful comments at the latter.
Steup, M. 1995. “Problem of the Criterion.” In R. Audi (ed.), The Cambridge Dictionary of
Philosophy, first edition, 653. New York: Cambridge University Press.
Taylor, S. & J. Brown. 1988. “Illusion and Well-Being: A Social Psychological Perspective
on Mental Health,” Psychological Bulletin 103: 193–210.
Van Cleve, J. 1979. “Foundationalism, Epistemic Principles, and the Cartesian Circle,”
The Philosophical Review 88: 55–91.
Weisberg, J. 2012. “The Bootstrapping Problem,” Philosophy Compass 9: 597–610.
Chapter 3
Is Cartesian Skepticism Too Cartesian?

Jonathan Vogel
1 Introduction
I understand Cartesian skepticism to be the doctrine that we lack all or nearly

all knowledge of the external world. A prominent response is that Cartesian
skepticism is too Cartesian. It arises from outmoded views in epistemology and
the philosophy of mind that we now properly reject. We can and should move
on to other things.1
§2 takes up three broadly Cartesian themes: the epistemic priority of ex-
perience, underdetermination, and the representative theory of perception. I
challenge some common assumptions about these, and their connection to
skepticism.2 §3 shows how skeptical arguments that emphasize causal con-
siderations can avoid some suspect Cartesian commitments. §4 explores the
related idea that perceptual knowledge is what I call instrumental knowledge.
Drawing on that connection, §5 examines a skeptical argument based on mini-
mal assumptions, and explains why a Moorean response to it is unsatisfactory.
The problem of skepticism is still with us, and it requires a substantive solution.
2 Three Cartesian Principles
Cartesian skepticism is of interest to the extent that is supported by a classic

line of argument. Here is one version:
Deceiver Argument
1. One’s beliefs about the external world are supported, if at all, by one’s
“sensory evidence”.
2. One’s “sensory evidence” consists in experiences or beliefs about the
character of one’s experiences.
1 “The real discovery is the one which enables me to stop doing philosophy when I want to”,
Wittgenstein (1969, § 133).
2 From now on, ‘skepticism’ means Cartesian skepticism, unless otherwise indicated.

Is Cartesian Skepticism Too Cartesian? 25
Combined, (1) and (2) amount to an epistemic principle with a Cartesian

pedigree:
(epe) The Epistemic Priority of Experience: One’s beliefs about the ex-
ternal world are supported, if at all, by one’s experiences or beliefs about
the character of one’s sensory experiences.
epe will be disputed by philosophers of various stripes, including apriorists,

foundationalists who think that beliefs about the external world are basic, and
contextualists who reject claims of (absolute) epistemic priority.3
Let me introduce some notation. M stands for some state of affairs in the
external world. biv(M) designates a state of affairs in which (i) M doesn’t ob-
tain and (ii) the subject is a brain in a vat such that the effects on her sensory
system are what they would be if M obtained. Continuing the argument:
3. One’s experiences and/or beliefs about experiences provide no more evi-

dential support to M than to biv(M).
From epe and (3):
4. M and biv(M) are equally well supported by one’s evidence.
At this point, the skeptic may invoke another principle that is Cartesian in spirit:
(UD) Underdetermination: If two contrary propositions are equally well

supported by S’s evidence, then S doesn’t know either one.4
Given (4) and UD, we have:
5. One doesn’t know M or biv(M).

6. Therefore, one doesn’t know M.
7. Therefore, one has no knowledge of the external world.
Another traditional thesis comes into play in connection with (2). The de-
tails may vary, but the gist is that the “immediate” objects of perception are
always mental. These intermediaries might be “sense-data, ideas, impressions,
phantasms or other queer entities.”5 I’ll choose sensations as a blanket term for
such items. We then have:
3 For relevant discussion see Vogel (1997) and Vogel (2008).

4 For the role of underdetermination in skeptical arguments see Vogel (2004).
5 The list is due to Byrne (2004: 307).
26 Vogel
(rtp) The Representative Theory of Perception: We perceive objects in

the external world by being aware of sensations, so that our awareness of
external objects is mediated or indirect.6
rtp is widely regarded as an unwholesome relic of Cartesianism. A stock re-

sponse to the Deceiver Argument is that it can be rejected because it presup-
poses or relies upon rtp.7 There are several versions of this idea.
An anti-skeptic might claim that epe entails rtp. Since rtp is false, so is
epe, and the Deceiver Argument is unsound. But consider a type of dogmatism
about perceptual justification. On this view, perceptual beliefs are justified by
sensory experiences with matching contents, and epe holds. An alternative to
rtp is intentionalism, which treats sensory experiences as intentional states
and denies the existence of sensations. Dogmatism as specified is consistent
with intentionalism. Therefore, epe doesn’t entail rtp, and the Deceiver Argu-
ment is unscathed.
Another thought is that rtp undergirds the Deceiver Argument by providing
a rationale for epe. According to rtp, we are immediately aware of sensations,
not worldly objects. Since we lack direct awareness of worldly objects, any
justification we might have for beliefs about them is bound to be inferential.
Inferentially justified beliefs about the world have to rest on non-inferentially
justified beliefs about something else. rtp offers a candidate, namely sensa-
tions. Since beliefs about the world are justified by beliefs about sensations,
epe is true. However, if we reject rtp, this motivation for epe is removed, and
the Deceiver Argument collapses.
The current objection is no stronger than the first, and for a similar rea-
son. Suppose that perception is metaphysically complex, as rtp says. Besides
the mind and the world, there are sensations that mediate between them. It
doesn’t follow that perception is epistemologically complex in the same way.
We might have immediate, non-inferential justification for what we believe
by perception, even if that simple cognitive relation is implemented or sub-
served by a complex metaphysical one. To assume otherwise—to assume that,
structurally, the epistemology of perception mirrors the metaphysics of per-
ception—is to make a mistake Sellars warned against when he criticized the
Myth of the Given.8 rtp and epe are independent, so denying rtp doesn’t
dislodge epe.9
6 See Jackson (1974: 1); Maund (2014: 5–6).

7 For references, see Macarthur (2003).
8 Sellars (1956).
9 There are other ways to argue for epe. One important line of thought invokes accessibilism,
the doctrine that what justification one has for a proposition is always knowable by reflec-
tion. See Williamson (2000, Chapter 8).
Some might believe that rejecting rtp undercuts skepticism in any case.
Suppose rtp is false, and genuine perception involves an unmediated grasp of
a mind-independent state of affairs. These wide perceptual states are, or yield,
knowledge of the world. Therefore, skepticism can’t gain any traction unless
rtp is true. Since rtp is false, skepticism subsides.10
There is a lacuna in this train of thought, at the very least. That perception
is wide doesn’t ensure that perception affords knowledge. A metaphysically
rich state that incorporates an object might be indistinguishable to you from a
metaphysically leaner state that doesn’t. To that extent, you would be unable
to tell whether you are the victim of massive sensory deception, and skepti-
cism remains a threat. The lesson once again is that metaphysical immediacy,
in and of itself, needn’t be epistemically significant. If further considerations
aren’t brought forward to establish that significance, denying rtp cuts no ice
against skepticism.11
3 Anti-Cartesian Skepticism
An essential component of skeptical arguments has been left out of the discus-
sion to this point. For the Cartesian skeptic, the causal character of perception
creates a special epistemic obstacle, liability, or risk. Tyler Burge (1986: 196)
writes:
I will construe Descartes as capitalizing on the causal gap that we tend to

assume there is between the world and its effects on us: different causes
could have produced “the same” effects, certainly the same physical ef-
fects on our sense organs. I will interpret him as conceiving a person as
radically mistaken about the nature of the empirical world. I shall see
him as…imagining that the entities that lie at the ends of relevant causal
chains (and perhaps the causal laws) are very different from what the
person thinks.
Why is the “causal gap…between the world and its effects” inimical to
knowledge? The long-standing answer is that the effects constitute a “veil of
10 Mc Dowell (1982) is a classic source for this kind of approach, though there is much more
to his view than what appears in the text. See Vogel (forthcoming) for extended discus-
sion of this kind of position.
11 See, inter alia, Vogel (1997), Martin (2004), and Vogel (forthcoming). To be clear, endors-
ing disjunctivism in particular, rather than rtp, doesn’t help us to repel the skeptic. In-
tentionalism is another alternative to rtp, as noted above. Byrne (2004) can be taken to
suggest that intentionalism has anti-skeptical consequences.
28 Vogel
perception” which blocks cognitive access to the objects in our environment.12

More explicitly, let’s grant for now that rtp is correct. We are directly aware of
sensations rather than material objects. The only way to learn about material
objects is by a causal inference from the sensations they produce. That is:
(inf) Causal Inference: S’s perceptual belief that P is justified, if at all, by

an unaided causal inference from the occurrence of (some of) S’s sensa-
tions to P.13
inf as stated implies rtp and epe. The skeptic will offer various reasons why
the inference from the occurrence of sensations to conclusions about the
world can’t be accomplished. The result is that there is a “veil of perception”
that makes justified belief about the external world impossible.14
The possibility of framing Cartesian skepticism in this way may seem un-
impressive. As noted in §2, many philosophers reject rtp out of hand. Others
may accept rtp, but deny inf. They think that causal inference doesn’t play an
essential role in justifying our beliefs about ordinary objects, so the skeptic’s
challenge to such inferences is beside the point.15
In fact, the Deceiver Argument can be recast in a form that is independent
of epe, rtp and inf. Consider a version of eliminativism with respect to sen-
sory experience:
Eliminativism. Perception gives rise immediately to beliefs about one’s

environment, and perceptual processes don’t produce sensory experi-
ences as such.16
The criticism that the Deceiver Argument presupposes rtp will be moot if
the skeptic can proceed under the assumption that eliminativism is correct.
What’s more, eliminativism implies that our perceptual beliefs aren’t justi-
fied by sensory experiences, since these don’t exist.17 It’s very plausible that
12 “This can seem to leave us in the position of finding a barrier between ourselves and the
world around us. There would then be a veil of sensory experiences or sensory objects
which we could not penetrate but which would be no reliable guide to the world beyond
the veil.” Stroud (1984: 32–3). Bennett (1971) introduced the metaphor.
13 Whether the inference proceeds from beliefs about sensations or sensations themselves
is a sore point, but nothing here turns on it.
14 Berkeley and Kant, if not Descartes himself, are canonical sources for the view that skepti-
cism is underwritten by inf.
15 See Jackson (1974, Chapter 6), Maund (2014: 5–6).
16 Armstrong (1961) defends a position somewhat like eliminativism, but see Note 19.
17 What could perceptual beliefs be, if eliminativism is true? We can’t say that they are be-
liefs prompted by sensory experience, since according to eliminativism there is no such
if we don’t have any sensory experiences, we don’t have beliefs about sensory
experiences, either. In that case, our perceptual beliefs aren’t justified by be-
liefs about sensory experiences.18 Therefore, eliminativism excludes epe as
well as rtp and inf.19
The Deceiver Argument can be restated in a way that respects eliminativism:
Belief Argument
1. If two mutually exclusive states of affairs X and Y would produce the same
perceptual beliefs in you, you can’t know by perception that X obtains or
that Y obtains.
2. M and biv(M) are mutually exclusive states of affairs that would produce
the same perceptual beliefs in you.
3. Therefore, you don’t know by perception that M obtains or that biv(M)
obtains.
4. Therefore, you don’t know by perception that M obtains.
5. Therefore, as a general matter, you have no perceptual knowledge of the
external world.
6. Therefore, you have no knowledge whatsoever of the external world.
This train of thought is consistent with eliminativism. To that extent, the ob-
jection that the skeptic is wedded to rtp, epe, or inf is ill-taken.
The Belief Argument may save the skeptic from some difficulties, but others
crowd in. For one thing, (2) is open to dispute. A widely held view is that the
causal history of one’s beliefs affects the content of what one believes. Let’s say,
for example, that circumstances are normal and you believe (L) that there is a
lion before you. Your beliefs with the content L are generally caused by lions,
and so are about lions. If you were a brain in a vat, however, none of your be-
liefs would be caused by lions. Therefore, none of your beliefs would be about
lions. Therefore, you wouldn’t believe L. Therefore, the states of affairs corre-
sponding to L and biv(L) wouldn’t produce the same perceptual beliefs in you.
Therefore, (2) is false.20
thing. There are other answers, however. Armstrong holds that perceptual beliefs are ones
that are caused in a distinctive way by a person’s sensory apparatus (1961: 105).
18 Byrne (2004) and others deny that we have such beliefs in any case. For the opposite view,
see Kelly (2008).
19 According to Armstrong (1961), perceptual beliefs are sensory experiences. A proponent of
this view would accept the first part of eliminativism as I presented it, while rejecting the sec-
ond. If perceptual beliefs can justify themelves, this truncated form of eliminativism would
be consistent with the letter of epe. But the issue concerning us is whether the skeptic can do
without epe, not whether a particular kind of anti-skeptic could accept eliminativism.
20 Other versions of content externalism could be deployed for the same purpose. An addi-
tional escape route may open up if perception puts one into a mental state that isn’t a belief.
30 Vogel
This critique of the Belief Argument might be resisted in a number of ways,

but pursuing these issues would take us too far afield.21 A different r eworking
of the Deceiver Argument keeps clear of such intricacies. It’s natural to sup-
pose that the causal dimension of perception gives it the epistemic signifi-
cance it has. Perception generates knowledge insofar as the world impinges on
one’s sensory system in a distinctive way. This thought can be turned around.
Differences in the world that don’t lead to differences in sensory experience
can’t be known to obtain by perception. So, let’s suppose that eliminativism is
true, and rtp, epe and inf are false. It is consistent with these stipulations that
you should find yourself in the following situation.
Twins. The twins Rex and Homer are molecule-for-molecule identical,

and you see Homer. Rex and Homer produce the same effects on your
retinas, and on your visual system more generally. In this situation, you
can’t know by perception that Homer in particular is nearby.
Similar cases are easy to come by. You can’t know by perception that what you
see is fool’s gold or real gold, because both substances affect you in the same
way. You can’t know by perception whether a painting is the original or an ex-
act copy for the same reason. And so on.
The skeptic will try to assimilate his far-reaching denial of knowledge to
our failure to know in cases like Twins. He will argue that being in Homer’s
presence and being in the presence of a suitably rigged computer affect your
visual system in identical ways, just as the presence of Homer and the pres-
ence of his twin would do. Hence, you don’t know by perception you are in the
presence of Homer rather than a nefarious computer, just as you don’t know
by perception that you are in the presence of Homer rather than Rex. If no
other information shores up your belief, you don’t know you are in Homer’s
presence, full stop. This kind of argument can be generalized, leading to full-
scale skepticism.22
The lesson the skeptic wants to draw from Twins and examples like it can be
formulated as:
Equipollence. If two mutually exclusive states of affairs X and Y would

produce the same effects on your sense organs, then you don’t know by
perception that X obtains or that Y obtains.23
21 One thought in the vicinity is that content externalism doesn’t license epistemological
externalism; see among others Audi (2001), Madison (2008), Tillman (2012), and Wikforss
(2007).
22 I am following the development in Vogel (1997).
23 Equipollence is a restricted and non-evidentialist analogue of the principle that a
body of evidence can’t justify rejecting a hypothesis that entails the evidence. See
The Perception Argument builds on Equipollence:
Perception Argument
1. Equipollence
2. M and biv(M) are mutually exclusive states of affairs that would produce
exactly the same effects on your sense organs.
3. Therefore, you don’t know by perception that M obtains or that biv(M)
obtains.
4. Therefore, you don’t know by perception that M obtains.
5. Therefore, as a general matter, you have no perceptual knowledge of the
external world.
6. Therefore, you have no knowledge whatsoever of the external world.
This line of thought doesn’t so much as mention experience. Thus, it avoids

commitment to rtp, epe and inf.24
The Perception Argument has another important feature. A proponent of
the argument denies that we have perceptual knowledge of the external world,
but she may raise no scruples about inductive knowledge as such. For exam-
ple, she may allow that extrapolation from past instances provides knowledge
of future ones. Unlike knowledge by induction, knowledge by perception re-
quires a successful causal transaction between the subject and the world. The
skeptic denies that such transactions can bear the epistemic weight assigned
to them; she needn’t commit herself to anything more.25
When confronted with the Perception Argument, many anti-skeptics will
reject Equipollence as too strong. In the formulation presented above, the
variables X and Y range over all metaphysically possible states of affairs. The
anti-skeptics maintain that you can know by perception that X obtains even
though states of affairs other than X would have the same effect on your sense
organs. Suppose so. The skeptic’s error lies in his assumption that a subject’s
knowing M by perception requires that the response of her sensory system
to M differ from the response it would have to all other states of affairs. This
isn’t to say that the Perception Argument goes astray by assuming rtp, epe,
Vogel (2014). This connection is important, and I intend to return to it in future

work.
24 According to Williamson (2000: 164), a cardinal sin of skeptical arguments is their com-
mitment to a particular symmetry claim: “Non-sceptics postulate a special asymmetry
between the good case and the bad cases in a sceptical argument…Sceptics try to under-
mine the asymmetry by claiming that the subject has exactly the same evidence in the
two cases”. However, the Perception Argument as stated doesn’t involve any assumptions
about evidence at all; see Note 23. For this reason, the Perception Argument also does
without UD, in letter if not in spirit.
25 See Vogel (2004), Vogel (2014), Vogel (forthcoming).
32 Vogel
or inf. An anti-skeptic can’t simply declare that Cartesian skepticism is too

Cartesian, and be on her way. Rather, she has to develop and defend a substantive
epistemological position that provides for the failure of Equipollence, or blocks
the transition from (5) to (6).26
4 Perceptual Knowledge and Instrumental Knowledge
We can gain some perspective on the Perception Argument and prepare for
the next section by examining what I’ll call instrumental knowledge. Some ex-
amples are:
Knowing that a printer is on by seeing its power light.

Knowing how much fuel is in the tank by consulting the gas gauge.
Knowing that a particular direction is North by reading a compass.
Knowing that a sample is acidic by seeing that it turns litmus paper red.
These examples have certain features in common.
(1) The object or source of instrumental knowledge (the printer’s being on,
the tank’s being full, …) causes a distinctive effect in a detector of some
kind (the power light, the gas gauge…).
(2) The knower perceives the effect on the detector.
(3) The effect represents the state of the source.
(4) The knower infers the state of the source from the state of the detector.
(5) The detector is an artifact made for the purpose of obtaining the kind of
knowledge in question.
(5) isn’t essential to instrumental knowledge, as I understand it. You could

know that it’s raining by hearing the sound the rain makes on the leaves of a
tree. The leaves serve as a detector of rain, but they aren’t an artifact made for
that purpose. Some questions might be raised about (3). I’ll let it stand, noting
that it might need to be revised or dropped.27
26 It might seem that Equipollence is tainted with Cartesianism insofar as it requires knowl-
edge by perception to be certain or infallible. One reply is that even if X and Y produce
different effects on your sensory system, respecting Equipollence, you might still confuse
the presence of the one with the presence of the other. Satisfying Equipollence doesn’t
exclude the possibility of error, or guarantee knowledge for that matter.
27 If instrumental knowledge requires representation, do worries about content externalism
apply here? According to (3) as stated, instrumental knowledge that X requires that X be
It can seem natural to suppose that perceptual knowledge is a species of in-

strumental knowledge. Perception, and perceptual knowledge, arises when an
object in the world causes a subject to enter a state that represents the object
in a distinctive “sensory” way. Thus, conditions (1) and (3) are satisfied. Further,
according to rtp, the subject directly perceives the representation (the sensa-
tion, in this instance). So, a version of condition (2) is met. If inf holds as well,
then your perceptual beliefs about external objects are justified by inference
from their effects on you, along the lines of (4). In short, given rtp and inf,
perceptual knowledge is a species of instrumental knowledge.
This result won’t encourage a skeptic who wants to keep clear of Cartesian-
ism. But consider what is unquestionably instrumental knowledge of objects,
namely knowing that the opera is beginning by watching a simulcast on TV. (1)
The goings on in the opera house cause an appropriate pattern of pixels on the
TV screen. (2) The viewer sees the state of the screen. (3) The state of the screen
represents (pictures, in fact) the performance. (4) Mobilizing background in-
formation about the TV, the viewer infers that the opera is beginning (rather
than, e.g., that an unusual video game is beginning).28 Now, suppose that the
TV has an output cable that can feed another video device. And suppose fur-
ther that Charlie has a special port installed in his head that connects to the
visual system of his brain. Charlie plugs the cable into the port, bypassing the
TV set, and he has visual experience as of the opera beginning. What happens
in the opera house affects Charlie in such a way that the state of Charlie’s visual
system represents it faithfully. Therefore, conditions (1) and (3) are met. How-
ever, the others aren’t. (2) Charlie doesn’t perceive anything other than the per-
formance. As to (4), let’s grant, in opposition to inf, that i nference plays no role
in the acquisition of perceptual knowledge. For Charlie, everything is as though
his visual system were getting its input from his eyes. If he doesn’t employ infer-
ence under normal circumstances, he won’t employ it when the signal to his
visual system comes through the port in his head. Hence, (4) doesn’t apply.
Still—and this is the key point—there doesn’t seem to be an epistemically
significant difference between Charlie’s watching the opera on the TV and his
“watching” it via the feed to his brain.29 Plugging the cable into Charlie’s brain
doesn’t increase his powers of discrimination or enable him, in any other way,
to know things about the opera that he wouldn’t know by looking at the TV.
represented in the “good case” when X obtains. Whether the instrument can represent X
when X doesn’t obtain is left open. There is more to say, but not on this occasion.
28 See Wright (2002).
29 As a matter of fact, Charlie satisfies the criteria for “recognitional perception” identi-
fied by Prinz (2006), and he seems to have what Lewis (1980) would consider “prosthetic
vision.”
34 Vogel
That is, Charlie’s capacity for perceptual knowledge is what it would be if per-
ceptual knowledge were instrumental knowledge as set out above. We arrive at
this conclusion even if we endorse eliminativism and reject epe, rtp, and inf.
Charlie can be described as having instrumental knowledge about the opera in
an (innocuously) extended sense.30
Now, instrumental knowledge seems subject to the following constraint,
which is a generalization of Equipollence:
Causal Constraint. If X and Y would produce the same state in a detector

(i.e., generate the same signal), then you can’t know by using the detector
that X obtains and Y doesn’t.
Reflection on various instances of instrumental knowledge makes the Causal

Constraint plausible. Both a loose gas cap and a bad spark plug will cause a
car’s check-engine light to go on. That’s why noticing the light doesn’t provide
knowledge that the gas cap is loose or that a spark plug is bad. Both hydrochlo-
ric acid and nitric acid turn litmus paper red. That’s why you can’t know by
using litmus paper whether the sample you have is hydrochloric acid or nitric
acid. Finally, think of a thermometer that is accurate between 0 degrees and
100 degrees. Above 100, it malfunctions and reads somewhere between 80 and
90. The thermometer says 87. You can’t know by consulting it whether the tem-
perature is 87 or over 100. It is easy to multiply examples like these, supporting
the Causal Constraint.
A Cartesian skeptic maintains that the causal character of perception clash-
es with its supposed status as a source of knowledge about the external world.
Conceiving of perceptual knowledge as instrumental knowledge bolsters this
point of view. If perceptual knowledge is instrumental knowledge, then per-
ceptual knowledge falls under the Causal Constraint. Equipollence holds as a
special case, allowing the Perception Argument to get underway. A skeptic can
proceed on these terms without relying on suspect Cartesian assumptions like
epe, rtp, and inf.
Still, the Causal Constraint, like Equipollence, may strike us as too demand-
ing or fastidious. By way of illustration, imagine that your home is equipped
with a burglar alarm and only you have the code to turn it off. Entry by an
intruder will send out a warning, but so will a short circuit within the device. If
30 Stroud likens the epistemic situation of a perceiver to that of someone watching a televi-
sion, in the course of explaining how the veil-of-perception doctrine leads to skepticism
(1984: 33, 210). I’m saying very nearly the opposite: whether perception is metaphysically
direct or indirect, inferential or non-inferential, is irrelevant so far as the skeptic is con-
cerned. Rather, the issue is (what the skeptic takes to be) the epistemically tenuous nature
of the causal relation between the world and the perceiving subject.
the Causal Constraint applies, the alarm doesn’t provide you with knowledge
that someone has broken into your home—there may have been an electrical
fault instead. But, the thought goes, when the alarm is activated, you can and
should disregard the possibility that it has been set off by a short circuit. You
are in a position to know that a break-in has occurred all the same.31
5 The Bottleneck Argument and Moorean Deduction
The Perception Argument rests on Equipollence. This principle is supported

by a range of examples, but, as noted earlier, philosophers from various camps
will regard it as too strong. However, a close cousin of the Perception Argu-
ment emphasizes the causal roots of skepticism and appears less vulnerable.
Let’s imagine a militantly anti-Cartesian skeptic named Nelson. Nelson is an
eliminativist, as described above. He rejects rtp, epe, and inf, and is agnostic
about UD. However, he does accept the Closure Principle for Knowledge. Nel-
son asserts:
Bottleneck Argument
1. Let biv stand for S is a thoroughly deceived brain in a vat. S doesn’t know
by perception that ¬biv.
2. S doesn’t know by deduction from something she does know by percep-
tion that ¬biv.
3. S has no other source of knowledge that ¬biv.
4. Therefore, S doesn’t know that ¬biv.
The Closure Principle gives us:
5. If S has any knowledge about the external world, S knows ¬biv.

6. Therefore, S doesn’t have any knowledge about the external world.
Nelson might treat (1) as a datum. Ordinarily we would say there are some
things you can see to be the case, and some not. You can that there’s blood in
the vial, but you can’t see that it’s type A blood or type O blood. You can see
that there’s a zebra in the pen at the zoo. You can also see that there isn’t an
elephant in the pen. But you can’t see that there isn’t a cleverly disguised mule
in the pen. Plausibly, knowing ¬biv is like that. You can’t see—and in general
you can’t know by perception—that ¬biv.
31 See Dretske (2008: 20–21).

36 Vogel
There is more to say in favor of (1). We might hold that (1) is true because
thought outruns perception. You can believe ¬biv, but you can’t see that
¬biv, because there is no way that you’re not being a brain in a vat looks to
you. In other words, what you know by seeing is limited by what you can
represent v isually, and you can’t visually represent ¬biv. Nelson will have some
reservations about this suggestion. As an eliminativist, he regards limitations
on perceptual representation as epistemologically insignificant. Perception
gives rise directly to beliefs, so in principle you can have a perceptual represen-
tation of anything that you can believe. You can believe ¬biv, of course, so to
that extent you should be able to know ¬biv by perception.32
Nelson would prefer to motivate (1) by emphasizing the causal character of
perception. Imagine you know by perception that there is a rabbit in the yard.
You know this, at least in part, because you perceive the rabbit, and perception
is a causal process. More generally:
If S knows by perception that a state of affairs X obtains, then X causes S

to believe that X.
Now, in simple situations, counterfactual dependence is the hallmark of causa-

tion. So perceptual knowledge obeys:
Counterfactuality. If S knows by perception that a state of affairs X ob-

tains, then if X weren’t the case, S wouldn’t believe that X.
The reason why you don’t know ¬biv by perception is that the necessary con-
dition laid down by Counterfactuality isn’t met. That condition is:
If you were a brain in a vat, you wouldn’t believe that you aren’t a brain
in a vat.
If you were a brain in a vat, you would believe that you aren’t one. So, given
Counterfactuality, you don’t know by perception that ¬biv.33
The claim that a brain in a vat would believe that she isn’t one may raise
worries about content externalism (see above). Those can be avoided by shift-
ing the focus of Counterfactuality from S’s beliefs to the state of S’s sensory
system. If you perceive that ¬biv, the truth of ¬biv must affect your sensory
32 For related discussion, see below.

33 Williamson (2000: 150–151) raises some similar points. However, he doesn’t note that that
the causal character of perception provides a reason why a sensitivity requirement ought
to apply to perceptual knowledge, but not to knowledge in general.
system in some way—say, by putting it into some state Q.34 Then, you know
¬biv by perception only if:
If you were a brain in a vat, your sensory system wouldn’t be in state Q.
If this change is made, an important difference between the Bottleneck Argu-

ment and the Perception Argument emerges. According to Equipollence, you
know X by perception only if, in all possible worlds where X is false, the ef-
fect on your sensory system differs from the one that X produces. According
to Counterfactuality, you know X by perception only if, in the nearest possible
world(s) where X is false, the state of your sensory system differs from the one
X produces. The modified Bottleneck Argument, like the Perception Argu-
ment, demands that you have some way to know that all alternatives to X are
false, but the modified Bottleneck Argument places less of this responsibility
on your sensory apparatus than the Perception Argument does.
As we have seen, the Bottleneck Argument is underpinned by the Closure
Principle and Counterfactuality. This combination prompts an objection.
Counterfactuality is a special case of the Sensitivity Condition for knowledge.
A sensitivity theorist holds that S knows X only if:
Sensitivity Condition. If X were false, S wouldn’t believe X (in the same

way).
Dretske and Nozick have taught us that mixing the Sensitivity Condition with
the Closure Principle leads to skepticism. If we reject the Sensitivity Condition,
as we should, the Bottleneck Argument gets stopped in its tracks. Nelson will
be unmoved by this response. He doesn’t claim that all knowledge is subject
to the Sensitivity Condition. He recognizes that at least some non-perceptual
knowledge needn’t satisfy that requirement, and he has a cogent reason for
maintaining that perceptual knowledge, in particular, does obey it.35
If we retain Counterfactuality, we can still escape the Bottleneck Argument
by rejecting the Closure Principle. The drawback is that the principle is true.
And dialectics aside, the Closure Principle matters in this context because it
connects knowledge of the world with knowledge that we aren’t victims of
massive sensory deception. I don’t think we should doubt for a moment that
34 Some disjunctivists deny that a worldly state of affairs and one’s perceiving that state of
affairs are related by causation. But even these philosophers ought to acknowledge that
the world’s affecting one’s sense organs is a genetic condition for perceptual knowledge
(see below). Snowdon (1980) might be a holdout.
35 See Vogel (2012).
38 Vogel
we know a lot, including that we aren’t brains in vats. But how do we know
that we aren’t? And how does our knowledge of ordinary things accommodate,
without capitulating to, the possibility of massive deception? Abandoning the
Closure Principle may obviate such questions, which go to the heart of why
skepticism is philosophically interesting in the first place.
Assume, then, that both Counterfactuality and the Closure Principle are
correct. One might reply to the Bottleneck Argument by invoking what I’ll
call Moorean deduction: you know by perception that, e.g., you have a hand.
You can infer from what you know that you aren’t a deceived brain in a vat. In
this case, your knowledge that you aren’t a deceived brain in a vat is deductive
rather than perceptual. You know (by perception) that you have a hand, and
you know (deductively) that you aren’t a deceived brain in a vat. The Closure
Principle holds. Nevertheless, we escape the Bottleneck Argument because
Premise (2) is false.
Nelson will resist this response, and rightly so. He begins by pointing out that
Moorean deduction, if legitimate, ought to apply to instrumental knowledge
across the board, rather than perceptual knowledge more narrowly. S uppose
you have instrumental knowledge of a proposition P but don’t have instru-
mental knowledge of the clearly entailed proposition Q. You can know Q by a
Moorean deduction from P, maintaining the Closure Principle.36 Returning to
the example in §3, when the alarm goes off you know that:
(B) A break-in has occurred.
B obviously entails
(¬S) It isn’t the case that the alarm has had a short circuit while no
break-in has occurred.
The security system can’t detect that there hasn’t been a short circuit instead of
a break-in, so you don’t have instrumental knowledge that ¬S. The suggestion
on offer is that you know ¬S by a Moorean deduction from B.
To evaluate this proposal, we need to observe that instrumental knowledge
has a genetic aspect and a normative aspect. Genetically, instrumental knowl-
edge is the upshot of a certain kind of interaction with a detector. Normatively,
a belief acquired as the result of that sort of interaction counts as knowledge
(other things being equal). Suppose you deduce ¬S from B. Your belief that
¬S isn’t instrumental knowledge on genetic grounds, because you arrive at it
36 Note that Moorean deduction has to proceed from instrumental knowledge, so it

wouldn’t be available to a brain in a vat.
by inference, rather than by attending to the alarm as such. But, from a nor-
mative standpoint, this genetic difference seems inconsequential. Deduction
can’t yield knowledge that the detector itself is unable to provide. Hence, the
Moorean response to the Bottleneck Argument is no good.37
This judgment needs elaboration and defense. Nelson might argue as follows.
By assumption, you have instrumental knowledge that B. You are aware that:
(A) The alarm has gone off.
and you take its doing so to indicate a break in. That is, you accept:
(A ⊃ B).
You infer B from A via (A ⊃ B) and thereby come to know B. Now, B entails ¬S,
so we may suppose that you accept:
(B ⊃ ¬S).
If you have your wits about you, you will put these two material conditionals
together and recognize that:
(A ⊃ ¬S).
It seems that when you are alerted by the alarm and accept A, you could be-
lieve ¬S via the conditional (A ⊃ ¬S), just as you can believe B via the condi-
tional (A ⊃ B).
Suppose you do move directly from A to ¬S. Nelson’s thought is that if you
satisfy the genetic criteria for instrumental knowledge with respect to B, you
ought to satisfy them with respect to ¬S as well. Then, if you don’t acquire in-
strumental knowledge that ¬S, your belief that ¬S must not meet the n ormative
requirements for knowledge. According to the Moorean, that is indeed how
things stand. He claims that, to attain knowledge that ¬S, you must deduce
¬S from B rather than A. But how does taking a detour through B permit your
belief that ¬S to satisfy the normative requirements for knowledge? Nelson
thinks that the Moorean has nothing satisfactory to say at this point, making
his response to the Bottleneck Argument untenable.
The Moorean might protest: According to Nelson, you learn that an intruder
has broken in by inferring as much from the alarm’s going off. Moorean deduc-
tion does no real work, because the two transitions you make (from A to B and
37 For the record, my own view is that Premise (3) of the Bottleneck Argument is incorrect.
40 Vogel
from B to ¬S) could be combined into one. However, what Nelson says bears
on the Bottleneck Argument only if perceptual knowledge is inferential in the
first place—a Cartesian assumption we are rejecting.38
This rejoinder misses the point. What undoes Moorean deduction is that
if the power of a detector is limited, its power—and hence its capacity to
generate knowledge—can’t be enhanced by someone’s making an inference
downstream.39 A successful Moorean deduction would, in effect, accomplish
exactly that. The total number of inferences involved isn’t the issue per se.
To get a better grip on things, we can imagine that the detector does the
Moorean deduction for you, so to speak. Let’s say you have an alarm system as
described above, but it works by sending a text message to your phone saying
“Alarm activated by an intruder!” We’re supposing that if you receive this mes-
sage you will have instrumental knowledge that a break-in is occurring. Let’s
add to the story that your phone has been re-programmed by a logically astute
and well-meaning engineer. When your phone receives the text message from
the security system, your phone does some computing and displays “Alarm not
activated by a short circuit rather than a break in!” Do you have instrumental
knowledge that a short circuit isn’t responsible for the alert? We assumed ini-
tially that your security system is unable to detect the absence of a short circuit
as such. Adding an electrical fault sensor to the system would change that; re-
programming your phone wouldn’t. So, when your phone carries out the infer-
ence for you, as it were, you can’t know instrumentally that there hasn’t been a
short circuit. But how would things be any different if the inference happens to
be located inside your head? In short: if Moorean deduction is successful, then
¬S can’t be known instrumentally, but can be known by deduction. Yet if ¬S
can be known by deduction, then ¬S can be known instrumentally (via the
reprogrammed phone). Therefore, Moorean deduction is unsuccessful.
Let’s step back and survey the ground covered over the last few pages. The
Bottleneck Argument rests on the Closure Principle and Counterfactuality.
Allowing Moorean deduction offers a way to accept those assumptions while
resisting the argument’s skeptical conclusion. But if deduction can add to per-
ceptual knowledge as the Moorean envisions, deduction should be able to ac-
complish the same thing with respect to instrumental knowledge in general.
There are two interlocking reasons to deny that instrumental knowledge can
38 The possibility that perception is “cognitively penetrated” may undermine the distinc-
tion between perceptual and inferential knowledge on which this objection depends. See
Siegel (2012).
39 I’m ignoring various fine points. The term “power” is non-specific and non-rigorous. En-
gineers discuss the sensitivity, precision, accuracy and resolution of detectors, none of
which can be enhanced by inference. See Northrop (2005).
be supplemented in this fashion: (1) The distinction between instrumental

knowledge and deductive knowledge stemming from it is superficial. (2) The
power of a detector can’t be enhanced by any inferences one might make. Since
Moorean deduction is futile, the Bottleneck Argument remains unanswered.
The Moorean has more to say, though. He will insist, correctly, that there are
situations in which a proposition can be known by perception, and a logical
consequence of that proposition can’t be known in the same way. This happens
when what we can think outruns what we can perceive. As an illustration, we
can represent visually (Z) there is a zebra in the pen, but we can’t represent visu-
ally (Z & C) there is a zebra in the pen and the Compactness Theorem is true.
That is why we can know Z by perception, but not Z & C. We can, however, know
Z & C by deduction from Z. Deduction is efficacious in this situation because it
allows us to bring supplementary representational resources to bear on what
we perceive to be so. Otherwise, we have everything we need to know Z & C.40
The question is whether Moorean deduction fits this paradigm. For compar-
ison, let’s return to the example in which your burglar alarm sends a message
to your phone. In its original configuration, the phone can’t represent or pro-
vide knowledge that the security system hasn’t been set off by a short circuit.
The reprogrammed phone can represent that a short circuit isn’t triggering the
message you receive, but it can’t give you knowledge to that effect. That is be-
cause, in and of itself, changing the representational capacity of an instrument
can’t increase its power and, thereby, the knowledge that the instrument can
provide. This observation applies to Moorean deduction. If you can’t see that
you aren’t a brain in a vat, mobilizing additional representational resources via
deduction is no way to make up for that lack. Moorean deduction is thus unlike
inferring Z & C from Z, where bringing further concepts to bear can increase
what we know.41
40 Inference comes into play in this instance because visual representation is hyperinten-
sional. But the hyperintensionality of perception doesn’t create the need for Moorean
deduction; see below.
41 The considerations I’ve raised are related to certain themes in the work of Fred Dretske.
He maintained that an information channel (source of knowledge) can’t provide knowl-
edge that conditions for the proper functioning of that channel are met. Dretske (2008)
provides an overview of his position; for discussion of Dretske’s views see Luper (2006).
The results argued for in the text can be obtained by insisting that the Sensitivity Con-
dition extends beyond instrumental knowledge to beliefs derived from instrumental
knowledge by inference. Dretske would be sympathetic, insofar as he was inclined to view
all knowledge as instrumental knowledge. But giving full sway to the Sensitivity Condi-
tion is undesirable (see immediately below). A better alternative might be to say that the
difference between instrumental knowledge and deductive knowledge proceeding from
it lacks weight (see above). Therefore, the Sensitivity Condition applies to both, though it
need not apply in other more problematic cases.
42 Vogel
The Moorean might try another tack. He will grant that instrumental knowl-
edge obeys a Sensitivity Condition, as perceptual knowledge more narrowly
does. That is, S knows P via a detector only if:
(sik) Sensitivity of Instrumental Knowledge. If P were false, S wouldn’t

believe via the detector that P.
However, knowledge gained by Moorean deduction from instrumental knowl-

edge that Z isn’t subject to a comparable sensitivity condition. Knowing P by
deduction from Z doesn’t obey:
(sdk) Sensitivity of Deductive Knowledge. If P were false, one wouldn’t

believe by deduction from Z that P.
The Moorean will analyze our working example as follows. Imagine that a bur-
glar sets off the security system, and your phone reads B. Presumably, if the
alarm hadn’t been activated by an intruder, it wouldn’t have been activated at
all. Your phone wouldn’t have received the message, and you wouldn’t have be-
lieved B by reading it. sik is satisfied, so you can have instrumental knowledge
that B. By contrast, the alarm system can’t provide instrumental knowledge that
¬S. If a short circuit were to occur, you would receive a text message and come
to believe ¬S. That is, if ¬S were false, you would believe ¬S by consulting your
phone, violating sik. You can, however, know ¬S by a Moorean deduction from B.
Admittedly, if a short circuit had occurred, you would have received warning of
a break-in, and inferred that the alert wasn’t due to a short circuit. That is to say,
if ¬S were false, you would believe ¬S by deduction from B. sdk isn’t satisfied.
But sdk is wrong, so there is no bar to your knowing ¬S. All in all, restricting
the sensitivity condition to instrumental knowledge gives Moorean deduction
something to do, and allows Moorean deduction to do it.
But why think that sik is true, while sdk is false? sik stipulates that S has
instrumental knowledge only if S’s believing and the state of the world are cor-
related in a certain way. For a reliabilist of a certain stripe, this kind of align-
ment between mind and world (via a detector) is constitutive of instrumental
knowledge.42 If the reliabilist countenances Moorean deduction, she has to
allow that the same sort of alignment isn’t necessary for deductive knowl-
edge. It is then incumbent on the reliabilist to (i) identify what instrumental
42 The alternative, which strikes me as correct, is that sik reflects how we can come to know;
it doesn’t specify what knowledge is, or in virtue of what someone knows something. In
the language used above, sik is a genetic requirement for instrumental knowledge, rather
than a normative one.
k nowledge and deductive knowledge have in common that makes them both
knowledge simpliciter; and (ii) explain how sensitivity can sometimes be con-
stitutive of knowledge, but not necessary for knowing by Moorean deduc-
tion.43 I doubt that an account along these lines can be given. But without it,
invoking and dispensing with sensitivity requirements to secure the legitimacy
of Moorean deduction seems like a lost cause.
It’s worth noting in this connection that treatments of Moorean deduction
in the literature have centered on the status of warrant transmission over en-
tailment and the failure of transitivity for evidential support.44 However, as
we have seen, the legitimacy of Moorean deduction becomes an issue within
a framework that is concerned with instrumental knowledge, and assigns no
role to evidence as such. It seems possible to account for the failure of Moorean
deduction in these terms, for at least some cases. To this extent, the focus on
transmission and the reach of evidence misses something important.45
There is another lesson to be drawn. All epistemologists have to allow for
the difference between the perceptual and non-perceptual dimensions of
knowledge, without losing sight of the unity of knowledge itself. How can
this be done? Cartesianism, broadly speaking, affords an answer: perception
causes sensory experiences, which provide evidence for everything we know
about the world. This claim incorporates and expands upon epe. Rival views
reject epe, with the hope that doing so will avoid skepticism. These efforts
will fail if they wind up distorting the nature of knowledge in an attempt to
save it. The attempt to marry instrumental knowledge with Moorean deduc-
tion risks doing just that; it risks breaking knowledge up into two fundamen-
tally different things, that are also supposed to be the same thing. We would
do better to accept epe, and argue that experience itself provides reason to
deny that we are victims of massive sensory deception.46 On this approach,
Cartesianism plays a role in defeating skepticism rather than promoting it,
and the thought that Cartesian skepticism is too Cartesian gets things the
wrong way around.47
43 Roush (2005) holds a position somewhat like this one.

44 See Moretti and Piazza (2013).
45 In my judgment, Dretske (2005) conflates these considerations with the applicability of
the sensitivity requirement, which is a mistake.
46 Stroud (1996) suggests that epe emerges as a response to skepticism, rather than a pre-
supposition of it. Still, for Stroud, epe stems from something like rtp.
47 My thanks go to Yuval Avnur, Selim Berker, Dan Greco, Ari Koslow, Kevin McCain, Ted
Poston, Bernhard Salow, Susanna Siegel and Declan Smithies for their help in thinking
through these issues. I’m also grateful to the audience at the Orange Beach Epistemol-
ogy Workshop for stimulating discussion. The National Endowment for the Humanities
granted me a Fellowship in support of this research, which I’m pleased to acknowledge.
44 Vogel
Bibliography
Armstrong, D. 1961. Perception and the Physical World. New York: Routledge.
Audi, R. 2001. “An Internalist Theory of Normative Grounds,” Philosophical Topics 29:
19–46.
Bennett, J. 1971. Locke, Berkeley, Hume: Central Themes. Oxford: Oxford University Press.
Burge, T. 1986. “Cartesian Error and the Objectivity of Perception.” In P. Pettit & J.
McDowell (eds.), Subject, Thought, And Context, 117–136. Oxford: Clarendon Press.
Byrne, A. 2004. “How Hard are the Skeptical Paradoxes?” Noûs 38: 299–325.
Dretske, F. 2005. “Is Knowledge Closed under Known Entailment? The Case Against
Closure.” In M. Steup & E. Sosa (eds.), Contemporary Debates in Epistemology, 13–
26. Blackwell.
Dretske, F. 2008. “Epistemology and Information,” In P. Adriaans & J. van Benthem
(eds.), Philosophy of Information: Handbook of the Philosophy of Science, 8, 29–47.
Amsterdam: Elsevier-North Holland.
Jackson, F. 1977. Perception. Cambridge: Cambridge University Press.
Kelly, T. 2008. “Evidence: Fundamental Concepts and the Phenomenal Conception,”
Philosophy Compass 3: 933–955.
Lewis, D. 1980. “Veridical Hallucination and Prosthetic Vision,” Australasian Journal of
Philosophy 58: 239–249.
Luper, S. 2006. “Dretske on Knowledge Closure,” Australasian Journal of Philosophy 84:
379– 394.
Macarthur, D. 2003. “McDowell, Scepticism, and the ‘Veil of Perception,’” Australasian
Journal of Philosophy 81: 175–190.
Madison, B.J.C. 2009. “On the Compatibility of Epistemic Internalism and Content
Externalism,” Acta Analytica 24: 173–183.
Maund, B. 2014. Perception. New York: Routledge.
McDowell, J. 1995. “Knowledge and the Internal,” Philosophy and Phenomenological
Research 55: 877–893.
Moretti, L. & T. Piazza. 2013. “Transmission of Justification and Warrant,” The Stanford
Encyclopedia of Philosophy (Fall 2018 Edition), Edward N. Zalta (ed.), URL = <https://
plato.stanford.edu/archives/fall2018/entries/transmission-justification-warrant/>.
Northrop, R. 2005. Introduction to Instrumentation and Measurements, 2nd edition.
Boca Raton, FL: CRC-Taylor & Francis.
Prinz, J. 2006. “Beyond Appearances: the Content of Sensation and Perception.” In T.
Gendler & J. Hawthorne (eds.), Perceptual Experience, 434–460. Oxford: Oxford Uni-
versity Press.
Roush, S. 2005. Tracking Truth: Knowledge, Evidence, and Science. Oxford: Oxford Uni-
versity Press.
Sellars, W.S. 1956. “Empiricism and the Philosophy of Mind,” Minnesota Studies in the
Philosophy of Science 1: 253–329.
Siegel, S. 2012. “Cognitive Penetrability and Perceptual Justification,” Noûs 46: 201–222.
Snowdon, P. 1980. “Perception, Vision and Causation,” Proceedings of the Aristotelian
Society 81: 175–192.
Stroud, B. 1984. The Significance of Philosophical Scepticism. Oxford: Oxford University
Press.
Stroud, B. 1996. “Epistemological Reflection on Knowledge of the External World,” Phi-
losophy and Phenomenological Research 56: 345–358.
Tillman, C. 2012. “Reconciling Justificatory Internalism and Content Externalism,” Syn-
these 187: 419–440.
Vogel, J. 1997. “Skepticism and Foundationalism,” Journal of Philosophical Research 22:
11–28.
Vogel, J. 2004. “Skeptical Arguments,” Philosophical Issues 14: 426–455.
Vogel, J. 2008. “Internalist Responses to Skepticism.” In John Greco (ed.), The Oxford
Handbook of Scepticism, 553–556. Oxford: Oxford University Press.
Vogel, J. 2012. “The Enduring Trouble with Tracking.” In Kelly Becker & Tim Black (eds.),
The Sensitivity Principle in Epistemology, 122–151. Cambridge: Cambridge University
Press.
Vogel, J. 2014. “E & -H.” In Dylan Dodd & Elia Zardini (eds.), Scepticism and Perceptual
Justification, 87–107. Oxford University Press.
Vogel, J. (forthcoming). “Wide Epistemic States.” In J. Dutant & F. Dorsch (eds.), The
New Evil Demon Problem. Oxford: Oxford University Press.
Wikforss, Å. 2007. “Semantic Externalism and Psychological Externalism,” Philosophy
Compass 3: 158–181.
Williamson, T. 2000. Knowledge and its Limits. Oxford: Oxford University Press.
Wittgenstein, L. 1969. On Certainty. (G.E.M. Anscombe & D. Paul, eds. and trans.). Ox-
ford: Basil Blackwell.
Wright, C. 2002. “(Anti-)sceptics Simple and Subtle: G.E. Moore and John McDowell,”
Philosophy and Phenomenological Research 65: 330–348.
Chapter 4
Hume’s Certain Doubts: Why We Should Worry Too

Kevin Meeker
1 Introduction
Searching for a philosopher unaware of Hume’s thoughts on induction would

be a waste of time – or maybe the basis for a clever Monty Python skit.1 Look-
ing for a philosopher familiar with Hume’s thoughts on deduction, however,
would be almost as futile. Well, almost. Many Hume scholars are of course
aware of his discussion of deductive (or demonstrative) reasoning in Book i,
Part iv, Section 1 of A Treatise of Human Nature, entitled “Of scepticism with
regard to reason.”2 Although Hume specialists often ignored this passage in the
not too distant past, they have recently paid more attention to it (see, e.g., De
Pierris 2015; Durland 2011; Garrett 2015; Meeker 2013). Moreover, some contem-
porary philosophers now recognize that these Humean reflections can help
us understand fundamental epistemological issues: Michael Huemer’s 2002
anthology reprints the argument for instance (270–274).3 Even medical epide-
miologists refer to this argument in the course of discussing causal reasoning
(see Morabia 2013, 1527). Despite this sporadic newfound interest, I shall argue
that in certain respects Hume scholars and epistemologists have yet to come
to grips fully with the powerful challenge of Hume’s reflections on demonstra-
tive reasoning.
1 I’d like to thank Susanna Rinard, Kevin McCain and Ted Poston for providing helpful com-
ments on an earlier version of this paper. I am also grateful to Jeanne Peijnenburg for inviting
me to present this paper at a Hume Workshop on the regress argument of T1.4.1 at the Rijk-
suniversiteit Groningen. The audience at this wonderful gathering provided me with tremen-
dous feedback as well, as did the other presenters at the workshop: Don Garrett, David Owen
and Jeanne Peijnenburg. To thank them for their help over the years in thinking about these
topics, this paper is dedicated to Don Garrett and David Owen.
2 All subsequent references to the Treatise come in two parts. The first is a reference to a para-
graph in the 2000 edition; the second is a page reference to the 1990 edition. So this section
would be (T1.4.1–12, 180–187). A reference to a more specific passage, such as Hume’s famous
is/ought paragraph, would be (T3.1.1.27, 469–470).
3 For another interesting discussion of the relationship between Hume’s thoughts on reason
and contemporary epistemology and philosophy of mind, see Couvalis (2011).

Hume’s Certain Doubts 47
Hume’s thoughts contain at least two important claims: first, “all knowledge
degenerates into probability” (T1.4.1.1, 180) and, second, “all the rules of logic
require… a total extinction of belief and evidence” (T1.4.1.6, 183). The two are,
of course, related. Nevertheless, more controversy engulfs the second claim be-
cause some interpret it as a radically sceptical conclusion that human beliefs
lack any epistemic merit. Most Hume commentators dispute this interpreta-
tion and a host of contemporary epistemologists deny the plausibility of such
radically sceptical stances. I have argued against both groups in the past (see,
e.g., Meeker 2013). Here I partly build on previous work to support this claim
indirectly. That is, I won’t concentrate on Hume’s notorious main argument for
the second claim. Instead, I focus on the surprising ramifications of a relatively
neglected argument in support of the first claim, which I call the “trickle down
argument.” In the first major section of this paper, I provide some brief back-
ground on the overall argument of i.iv.1. The second main section shows how
the underappreciated trickle down argument helps us to understand the moti-
vation and structure of the reasoning of this section. Finally, in the last major
Section i contend that this argument provides a worrisome sceptical threat to
our knowledge, even if knowledge is different from Hume’s conception of it.
2 Interpretive Controversies and the Structure of Hume’s Regress
Hume distinguishes between knowledge and probability as follows: “By knowl-

edge, I mean the assurance arising from the comparison of ideas… By prob-
ability, that evidence, which is still attended with uncertainty” (T1.3.11.2, 124).
Of course, not any assurance or certainty arising from a comparison of ideas
will count as knowledge. According to Hume, we must strive to “… attain a
proper stability and certainty in our determinations.”4 To be agreeably brief,
then, Hume seems to provide the following theory of knowledge: S knows p
if and only if (i) S’s assent to p arises from a comparison of ideas and (ii) S is
justifiably certain that p.5
Given this conception of knowledge, let us consider Hume’s own words
about demonstrative reasoning:
4 (ehu 12.24). References to Hume’s Enquiry concerning Human Understanding are to David
Hume, An Enquiry concerning Human Understanding, ed. Tom L. Beauchamp (Oxford: Ox-
ford University Press, 1999), and are abbreviated as “ehu,” followed by the section and para-
graph numbers of the relevant passage.
5 See Meeker (2013: ch. 2) for a fuller defense of this interpretation. I presume that, for Hume,
justifiable or proper certainty entails truth.
48 Meeker
In all demonstrative sciences the rules are certain and infallible; but
when we apply them, our fallible and uncertain faculties are very apt to…
fall into error. We must, therefore, in every reasoning form a new judg-
ment, as a check… on our first judgment or belief; … By this means all
knowledge degenerates into probability (T1.4.1.1, 180).
Even expert “Algebraists” and “Mathematicians” check their proofs to increase

their confidence, which shows that this process “… is nothing but the addition
of new probabilities” (T1.4.1.2, 181). Once we allow that the ubiquitous checking
process of mathematical reasoning is simply probabilistic reasoning, Hume in-
sists that “a new species of probability” emerges that can “correct and regulate”
the previous probability judgment (T1.4.1.5, 182). Unfortunately, this new form
of regulation leads to further uncertainties, launching a devastating regress. As
Hume puts it:
When I reflect on the natural fallibility of my judgment, I have less con-

fidence in my opinions, than when I only consider the objects concern-
ing which I reason; and when I proceed still farther, to turn the scrutiny
against every successive estimation I make of my faculties, all the rules
of logic require a continual diminution, and at last a total extinction of
belief and evidence (T1.4.1.6, 183).
We are confronted, then, with a seamless regress argument that threatens to

undermine the epistemic merit of all cognitive endeavors. When Hume later
reflects on this insidious regress argument, he confesses that he wants to cast
aside all reasoning “… and can look on no opinion as more probable or likely
than another” (T1.4.7.8, 268–269). I have labeled the view “epistemic egalitar-
ianism” and have argued that Hume does indeed embrace it (Meeker 2008,
2013).
Others disagree. These debates about Hume’s second claim turn on sever-
al key points. Some argue that Hume clearly denies that anyone holds such
a view and point to his contention that those who try to refute “total scepti-
cism” have really “… disputed without an antagonist” (T1.4.1.7, 183). Oftentimes
commentators buttress this anti-scepticism reading by arguing that Hume’s
probabilistic reasoning in this regress is so bad that it is implausible that any
philosopher would accept it. Because Hume’s probabilistic inference is not
perfectly perspicuous, I believe that we have quite a bit of leeway in trying to
reconstruct it in a respectable manner. To cite just one example: I have con-
tended that Hume’s argument could be read as paralleling to some extent the
C.I. Lewis/Reichenbach debate about whether for anything to be probable,
something needs to be certain.6 Turning to the textual challenge, when Hume

denies the existence of “total sceptics” I take him to be referring to the view
that one should abandon all beliefs. I agree that Hume scoffs at such doxastic
eliminativism because he thinks that belief is unavoidable. But I deny that he
thereby abandons epistemic egalitarianism.
While these controversies are important, we might be able to make inter-
pretive progress more quickly if, as mentioned, we look at the first argument.
As David Owen has astutely noted: “… the eventual solution Hume presents
later on in I.4.I is a solution to the challenge posed by the second argument,
i.e., the probability of all judgments reduces to zero, and not the first argument,
i.e., that all knowledge degenerates into probability. So it looks like the argu-
ment remains intact…” (Owen 1999: 182). Given that the controversy about the
second claim involves debates about whether the argument remains intact, we
can avoid such distractions by focusing on the interpretive issues raised by the
first argument. So we now turn to this first claim.
3 Reflection and the Neglected Trickle Down Argument
Donald Ainslie has recently argued that “the nature of the checking process”
leads Hume to the first claim and then “the [second] sceptical argument gets its
force because the process of checking iterates” (Ainslie 2015: 23). If the regress
derives all of its force from the reflective nature of the checking process, then
one can undermine the epistemic egalitarian reading by identifying a source of
justifiable certainty that needs no checking. Identifying such a source could al-
low one to contend that the claim “all knowledge degenerates into probability”
should not be read as a denial of knowledge as Hume understands it. Rather,
it should be read as claiming that “there is knowledge but it is destroyed by
reflection” (Kail 2015: 627; Kail’s emphasis). Consider again in this context
mathematical computation. On this view, our calculations begin with intui-
tively certain mathematical claims, which arise from a comparison of ideas.
Once we, or our colleagues, check the calculation, we base our justification for
continuing to accept the results on fallible or probabilistic evidence. The falli-
bility of the checking processes we use transfers to the products of the process.
Demonstrative knowledge is thus “fragile” and “frequently destroyed” (Kail
2015: 628). But what is frequently destroyed is not always destroyed. Fragile
6 See Meeker (2013: chs. 3 and 8) for some brief discussion. For an excellent overview of the
epistemological issues raised in these contexts, see Atkinson and Peijnenburg (2006).
50 Meeker
d emonstrative knowledge can escape the ravages of reflective uncertainty by

resting on intuition, which gives us special access to the contents of our ideas.
Such access provides us with a certainty not subject to the probabilistic rea-
soning of the checking process.7
Although the checking process plays an important role in Hume’s reason-
ing, there is more to the story of how intuition is undermined. Interestingly,
Ainslie’s book, which focuses almost exclusively on Treatise 1.4, neglects the
other important argument for this first claim, which should be considered in
its entirety:
Now as none will maintain, that our assurance in a long numeration ex-
ceeds probability, I may safely affirm, that there scarce is any proposition
concerning numbers, of which we can have a fuller security. For ’tis eas-
ily possible, by gradually diminishing the numbers, to reduce the longest
series of addition to the most simple question, which can be form’d, to
an addition of two single numbers; and upon this supposition we shall
find it impracticable to shew the precise limits of knowledge and of prob-
ability, or discover that particular number, at which the one ends and the
other begins. But knowledge and probability are of such contrary and
disagreeing natures, that they cannot well run insensibly into each other,
and that because they will not divide, but must be either entirely present,
or entirely absent. Besides, if any single addition were certain, every one
wou’d be so, and consequently the whole or total sum; unless the whole
can be different from all its parts. I had almost said, that this was certain;
but I reflect, that it must reduce itself, as well as every other reasoning,
and from knowledge degenerate into probability (T1.4.1.3, 181).
For Hume the uncertainty inherent in complex calculations trickles down

to even the simplest ones. Importantly, this uncertainty does not spread via
reflection. The lack of a sharp boundary between uncertain calculations and
allegedly certain additions is enough to defeat the knowledge of “intuitive” ad-
ditions. To be sure, it is “easily possible” to see that complex computations and
the simplest additions do “run insensibly into each other.” But this possibil-
ity need not be actualized for there to be a defeater. Facts about our fallibility
7 For a sketch of this position, see Kail (2015: 627). Ironically, contemporary philosophers who
disparage Hume’s views have a similar interpretation: “… the epistemic transparency of
meanings or contents of thoughts, widely taken for granted in Hume’s day, is now hotly con-
tested; views as straightforward as Hume’s are no longer regarded as tenable” (Brown 2004:
129). On my interpretation, Hume’s argument entails that we lack the epistemic transparency
ascribed to him by Kail and Brown.
and cognitive limitations are sufficient to undermine any justifiable certainty we

might have thought we had. So the first step of Hume’s regress need not rely on
any reflection requirement.
Note that Hume holds this argument in high esteem for at least two reasons:
its strength and its scope. With regard to the former, he is tempted to assert
that the conclusion is certain, which suggests the very strong claim that one
can know that nothing is justifiably certain. Because knowledge requires justi-
fiable certainty, though, he realizes that this argument would be self-defeating.
Nevertheless, his language suggests that the conclusion is as certain as any can
be without actually attaining such a lofty status. Adding to the impressive na-
ture of such a strong argument is its wide-ranging applicability. That is, Hume
argues that we can apply this reasoning to “every other reasoning” to show the
degeneration of knowledge to probability. Thus, we can apply it to every kind
of demonstrative reasoning or reasoning that aspires to the exalted status of
knowledge.
Is Hume’s high opinion of this argument justified? That’s a difficult question
to answer. One way to gauge the strength of this argument is to examine dif-
ferent types of allegedly exalted epistemic states to see if they are vulnerable
to this type of defeater. If we find other similar areas to which we can apply
this trickle down reasoning, or can find others who have reasoned in a similar
manner, then we have good reason to believe that Hume is on to an impor-
tant issue. As it happens, this trickle down argument is remarkably similar to
Timothy Williamson’s influential anti-luminosity argument, which attacks a
Cartesian thesis about the kind of access we have to some “core” mental states
such as being in pain or being appeared to greenly. According to Williamson,
some maintain that such mental states are luminous because one is always in
a position to know whether one is in the state (Williamson 2005a: 434). Clear-
ly a luminous condition is an exalted epistemic state just as knowledge is for
Hume. Moreover, note that attacking luminosity would be tantamount to at-
tacking the claim that we have infallible access to the contents of our ideas.8
Williamson’s argument focuses not on the border between simple and com-
plex computations but on the border between feeling cold and feeling hot.
Very roughly speaking, if feeling cold is luminous, then in a situation in which
one is very gradually warming, there would be an exact point at which one
can know that one has transitioned out of this state. Because there is no such
precise boundary, feeling cold is not luminous. Similarly, we can read Hume as
constructing a reductio of the idea that the condition of intuitively grasping a
8 For more on how Williamson’s anti-luminosity argument is related to various types of exalted
epistemic states, see Meeker and Poston (2010).
52 Meeker
mathematical claim provides the justifiable certainty necessary for knowledge.

Assume that we know (are justifiably certain of) simple sums by intuition. If
we have such knowledge then it is possible to identify the precise point at
which additions require a non-intuitive (and thus probabilistic or uncertain)
justification. But it is not possible to identify such a point. Thus we have no
knowledge of simple sums. We can take Williamson’s summary of his strategy
and parenthetically add the Humean parallels to gain a clearer picture of their
similarities:
The strategy is to construct a sorites series between a case in which the

condition [intuitively grasping a mathematical claim] clearly obtains and
one in which it clearly fails to obtain, and then to argue that such a series
cannot exist for a luminous [or epistemically certain] condition. Lumi-
nosity [or Epistemic certainty] must fail close to the boundary between
cases where the condition [intuitively grasping a mathematical claim] ob-
tains and cases where it does not, just on the obtaining side.9
In sum, the structural similarity between these two arguments suggests that
we have good reason to take Hume’s trickle down reasoning seriously.10
Potential objections to this reading of Hume fall into two predictable cat-
egories: philosophical and textual. Philosophically, one could pull a G.E. Moore
and admit that while complex calculations are uncertain and there is no obvi-
ous boundary separating dubitable long computations and intuitively certain
simple ones, there clearly are some simple demonstrative truths about which
there are no doubts. We are as certain that 1+1=2 as we are that we have two
hands. Simple arithmetical truths are beyond doubt even if we cannot say pre-
cisely where such simple truths end and the complex dubitable ones begin.
Moreover, some principles of logic are just as certain as mathematical prin-
ciples, if not more so. Since the time of Aristotle the law of non-contradiction
9 Williamson (2005b: 230). Here is a similar statement: “… a luminous condition obtains

in any case close enough to cases in which it obtains … a luminous condition spreads
uncontrollably through conceptual space, overflowing all boundaries. It obtains every-
where or nowhere, at least where we are in a position to wonder whether it obtains. For
almost any condition of interest, the cases in which it obtains are linked by a series of
imperceptible gradations to cases in which it does not obtain, where at every step we
are in a p
osition to wonder whether it obtains. The condition is therefore not luminous”
(Williamson 2000: 13).
10 For a much more detailed comparison of the similarities between Hume and Williamson,
see Meeker and Poston (2010). There we also criticize recent attempts to preserve some
kind of privileged access to our mental states in light of these types of arguments. In the
next Section i attack attempts to preserve some kind of certainty for the demonstrative
sciences.
(lnc) has been regarded as the most certain of principles. Surely this law is so
obviously indubitable and intuitively certain that it is not possible to rationally
renounce it. And surely Hume would not have been so daft as to deny such an
obvious position. Paralleling this Moorean attack, one could also argue, textu-
ally, that Hume acknowledges domains containing intuitively certain truths,
such as Euclidian geometry (e.g., ehu 4.1). In the next Section i show the
weaknesses of these objections and argue that we are left in a very worrisome
position.
4 The Radical Nature of Trickle Down Arguments
4.1 Stating the Argument

To answer the objections just mentioned, it will be helpful to test Hume’s
reasoning in the context of another contemporary debate. We can compare
Hume’s claim that there is no certain knowledge to Quine’s claim that there
is no analytic knowledge. The two are of course not making the exact same
point. But they are similar enough. Consider Quine’s famous point that “… no
statement is immune from revision. Revision of even the logical law of the ex-
cluded middle has been proposed as a means of simplifying quantum mechan-
ics” (1980: 43). Just as Quine argued that empirical/synthetic statements from,
say, physics, can affect the justification of allegedly analytic statements so too
Hume argued that empirical considerations (or considerations about matters
of fact concerning the limitations of our cognition) can affect the justification
of the demonstrative sciences that arise from a comparison of ideas.11 Interest-
ingly, just as some dispute the textual basis for attributing this view to Hume,
so also some dispute the textual basis for assigning this view to Quine. Some
distinguish between a “radical” Quine who admits that any logical laws can be
revised and the “logic-friendly” Quine who exempts logic from his attack on
the analytic/synthetic divide.12 This interpretive controversy notwithstanding,
Arnold and Shapiro (2007) argue that one who subscribes to radical Quinean
ideas must allow that all principles are open to revision, even the most certain
of all: non-contradiction. By the same token, I suggest that the radicalism of
Hume’s trickle down argument likewise entails that even lnc is up for grabs.
More to the point, I now argue that the logical radicalism of Hume and Quine
is so radical that it leads one to the precipice of epistemic egalitarianism.
11 See Meeker (2007, 2011, and 2013:ch. 2).

12 Arnold and Shapiro note this interpretive debate (2007: 276) and claim to read Quine as a
radical (2007: 278).
54 Meeker
We can sketch a simple version of the argument:
(1) It is rationally permissible to accept the radical Humean/Quinean the-

sis about the uncertainty/revisability of demonstrative sciences such as
logic.
(2) If (1), then, (2b) possibly, it is rationally permissible to deny the law of
non-contradiction.
(3) If (2b), then, possibly, it is rationally permissible to accept that all propo-
sitions are true.
(4) Possibly, it is rationally permissible to accept that all propositions are
true.
If the possibility in (4) were actual, then in an important sense all beliefs would
be epistemically equal, which certainly looks like a form of egalitarianism. For
if it is rational to believe that every proposition is true, then it is rational to
believe that all beliefs are the same with respect to epistemic merit. You are,
after all, guaranteed to be correct no matter what you believe. And what could
provide more rational support for a proposition than a guarantee that it is cor-
rect? More directly, if it is rational to believe that every proposition is true, then
clearly the proposition that all beliefs are the same with respect to epistemic
merit is likewise rationally acceptable. Granted, this argument only leads to
the possible rational acceptability of epistemic egalitarianism. But, as we’ll see,
this possibility is quite worrisome.
4.2 David Lewis, Quine and Hume

What to make of this argument? It is clearly valid: if we accept (1) then we
can infer (4). But what about the premises? David Lewis seemed to accept (2)
when he claimed that “Many philosophers hold an unprincipled and unstable
position: they have been persuaded by Quine and Putnam that logic is in prin-
ciple open to revision, they are prepared to contemplate revisions of logic that
seem to them to require only small and esoteric changes, yet they still think it
absurd to countenance true contradictions.”13 Perhaps we can understand the
resistance to (2) that Lewis characterizes along the following lines. Logic is in
principle revisable, but it is very unlikely that there are, say, true contradictions.
Of course, in the rhetorical heat of battle very unlikely positions are often dis-
missed as though they are absurd. This understanding of the resistance is still
vulnerable to Hume’s challenge though. For even very unlikely positions are
still by definition at least somewhat likely and debates about them essentially
13 Priest quotes this passage from a letter that Lewis wrote in defense of the advisability of
publishing Priest’s original book on dialetheism. See Priest (2006: xix).
involve probabilistic reasoning. Mimicking the reasoning of Hume’s trickle

down argument, Lewis points out that once one allows the uncertainty/revis-
ability of empirical matters to infect any part of logic, there is no way to draw
a precise limit between those peripheral principles of logic that are uncertain
and those central principles that are certain and unrevisable. Logic principles
stand or fall together. Infecting any part of the logical body is tantamount to
infecting the whole body. Lewis is correct. Accepting the antecedent of (2) but
denying its consequent is implausible.14
That’s why Lewis rejects (1). And by rejecting (1) he can consistently reject
the consequent of (2). More specifically, in “response” to dialetheists such as
Graham Priest who believe that some contradictions are true, Lewis has con-
tended that lnc is something that
we know for certain, and a priori, and without any exception for especial-
ly perplexing subject matters… That may seem dogmatic. And it is… con-
trary to the rules of debate–I decline to defend it. Furthermore, I concede
that it is indefensible… They have called so much into question that I
have no foothold on undisputed ground
(Lewis 1982: 434).15
By arguing that we know lnc for certain without any possible exception, Lewis
implies that, evidentially speaking, it is not possible to rationally deny lnc.16
So Lewis is comfortable with jettisoning (1), or at least the consequent of (2),
without a debate. His point is that it is difficult to see what kind of eviden-
tial reasoning could possibly lead to the rejection of lnc because evidential
14 A similar point is made by Arnold and Shapiro: “… the radical Quinean is led rather quick-
ly and rather directly into something in the neighborhood of Priest’s dialetheism” (2007:
278).
15 Later Lewis declined to “debate” this issue on similar grounds: “My feeling is that since
this debate instantly reaches deadlock, there’s really nothing much to say about it. To con-
duct a debate, one needs common ground; and in this case, the principles not in dispute
are so very much less certain than non-contradiction itself that it matters little whether
or not a successful defence of non-contradiction could be based on them” (2004: 176).
16 Interestingly, one could contend that (1) is true on other construals of rationality. Susanna
Rinard maintains that it can be pragmatically rational to believe in contradictions and
that: “The idea that there is anything inherently wrong with believing contradictions is
just a symptom of evidentialist thinking” (2017: 137). Blaming resistance to believing con-
tradictions on evidentialism seems to presuppose that we have overwhelming evidence
that lnc is true, even if there are some circumstances in which it is pragmatically rational
in some sense to believe contradictions. This view still seems to deny (1), or at least the
consequent of (2), insofar as rational permissibility is understood roughly in terms of
evidential or epistemic rationality.
56 Meeker
considerations presumably presuppose lnc.17 Despite Lewis’ protestations, it

seems that philosophers do debate the acceptability of lnc on something like
common ground. Otherwise, it is difficult to account for those who have re-
cently defended lnc without simply dismissing opposition to it. For example,
instead of maintaining that dialetheism is certainly false, some provide exten-
sive reasons for lnc and conclude that arguments attacking lnc “are not yet
conclusive” (Zalta 2004: 433) or face “serious challenges” that need to be “met”
(Littman and Simmons 2004: 334). But these inconclusive conclusions seem
to intimate that the arguments have at least some epistemic merit and that
this “is a view that all of us must reckon with” (Shapiro 2004: 337). While it
may be important to be aware of historically important positions that are ra-
tionally unacceptable, I take it that positions that we are required to “reckon
with” are rationally acceptable on some level. In short, then, while many find
dialetheism ultimately unconvincing, their arguments against it imply that
they tacitly accept (1) and the consequent of (2).
Moreover, recall Hume’s point that [certain] knowledge and [uncertain]
probability cannot run insensibly into each other. Lewis effectively argues or
assumes that there is some precise boundary that prevents the uncertainty of
most philosophical debates from infecting our certain knowledge of lnc. To-
wards the end of his life, however, Lewis conceded that he could “reason about
impossible situations” although he did not “really understand how that works”
(Lewis 2004: 176). Interestingly, he claimed to reason rationally about subtly
impossible situations and relegated dialethic reasoning to the realm of bla-
tantly impossible reasoning. As he puts it: “I agree with [Priest] about the many
uses to which we could put make-believedly possible impossibilities… The
trouble is that all these uses seem to require a distinction between the subtle
ones and the blatant ones (very likely context-dependent, very likely a matter
of degree) and that’s just what I don’t understand” (Lewis 2004: 177). Instead of
a precise boundary between lnc and other philosophical debates, we are left
with a vague distinction between rationally debatable, subtle impossible situ-
ations and blatant impossible situations that cannot be fruitfully discussed.
Once again, this is exactly what one would expect if the radical interpretation
of Hume’s trickle down argument is correct.
4.3 The Specter of Trivialism

Now consider (3). One of the most common objections to dialetheism is the
accusation that accepting a contradiction means that one must accept all
17 One can find similar arguments in Frege, Paul Boghossian, Hartry Field, and Crispin
Wright. For a brief discussion of their arguments, see Resnik (2004: 182).
propositions, a view called trivialism (see, e.g., Kroon 2004). Unfortunately, dis-
cussions of trivialism are even more recent and more scarce than ones about
dialetheism. Nevertheless, the trivialist debate is simply an extension of the
debate about the epistemic status of logic. And as with the unfolding dialec-
tic about lnc, we can see a similar pattern already emerging. Although Priest
tries to insulate himself from trivialism by, among other maneuvers, famously
spurning the classical principle of explosion, he does not reject (3). We can
see this in several ways. In response to an argument advancing trivialism via
a different route than explosion, Priest claims only that such a view is “[m]ost
implausible”18 but not rationally indefensible. More generally, in trying to pres-
ent something of a “transcendental” argument against trivialism, Priest admits
that he “may well be … wrong” (Priest 2000: 189). Such qualifications seem to
preclude inferring that one has established the rational unacceptability of triv-
ialism in any circumstance. Interestingly, one need not be a dialetheist to make
the case that “nothing in our logical notions rules out [trivialism]” (Estrada-
González 2012: 176). These discussions provide reason to believe that the case
against trivialism does not approach certainty, especially if one accepts the
radical Humean/Quinean thesis. Our treatment of dialetheism also illustrates
how the uncertainty of probable reasoning not only leaps the boundary to lnc
but also trickles down to a principle that some see as more certain. That is, if all
beliefs are probable and not certain, then we need to abandon the certainty of
lnc and anti-trivialism. Overall, Hume’s trickle down reasoning applies to log-
ic in a way that anticipates some of the nuanced steps of the current debates
that produce views that seem almost to imperceptibly shade into each other.
Once (3) is in the fold, we can return to the question of epistemic
egalitarianism. Priest has noted that one can distinguish between sceptics and
trivialists because “The trivialist will subscribe to everything; the sceptic will
subscribe to nothing” (Priest 2000: 189). On my view, Hume is not a t rivialist.
Nor is he a sceptic who suspends belief. But his trickle down argument does,
at a minimum, seem to entail the possibility of rationally accepting every
proposition. If such a possibility were actual, then obviously one could “… look
on no opinion as more probable or likely than another” (T.1.4.7.8, 268–269).
The acceptability of every opinion would essentially eviscerate the notion of
evidence understood as a means by which to distinguish truth or falsity or even
as a means to proportion one’s belief. So the trickle down argument seems to
lead effectively to the acceptability of the “extinction of evidence.”
18 Actually, this comment appears in a paper Priest co-authored with two others (see Degu-
chi, Garfield and Priest 2013: 357).
58 Meeker
4.4 Interpretive Issues Revisited

Having disposed of the philosophical objection to the radical interpretation of
Hume’s trickle down argument, we can return to the textual one. To be sure,
Hume does at times speak of the “certainty” of aspects of Euclidian geometry
(ehu 4.1). In so doing, though, he has not given us any reason to suppose that
he has abandoned his trickle down argument. Instead, I suggest, he is using
the term differently or being sloppy. After all, he begins T1.4.1 by noting that
the rules of the demonstrative sciences – which presumably include lnc and
non-trivialism – are “certain” in an objective sense but he quickly argues that
they are “uncertain” in a subjective sense. That is, he allows that such laws are
necessary or “infallible” truths but denies that we have infallible access to such
truths. Given the ambiguity of the term “certain” or its various cognates for
Hume, we should look to his explicit discussion of these issues as a controlling
interpretive authority. We can put the point in the form of a concrete inter-
pretive principle: If a philosophical text explicitly affirms a universal generaliza-
tion and explicitly recognizes that the universality of this generalization applies
to its own position or the reasoning supporting this position, then one is justified
in attributing this position to the text unless there is an explicit retraction of that
universal generalization. It is difficult to imagine a more explicit or unambigu-
ous statement than the following: “I had almost said, that this was certain; but
I reflect, that it must reduce itself, as well as every other reasoning, and from
knowledge degenerate into probability” (T1.4.1.3, 181). Clearly here Hume is
affirming a universal generalization: all alleged knowledge/certain reasoning
degenerates into probability. Moreover, he recognizes that the universal nature
of this generalization means that it applies to the argument/reasoning that
produced it and, though he thinks highly of it, agrees that the reasoning itself is
only probabilistic. Because none of the other uses of “certainty” are an explicit
retraction, we can conclude that Hume is using the term objectively elsewhere
or he is just being sloppy.
Lest one think that this principle was an ad hoc one devised to arrive at the
desired interpretive conclusion about Hume, consider the following passages
from Priest: “In a word, all beliefs, certainly all beliefs of any degree of sub-
stance, are fallible” (Priest 2006: 262) Tellingly, he adds in a note: “And, yes, that
belief is fallible, too” (Priest 2006: 262). Despite this explicit statement of a uni-
versal generalization that is clearly meant to be self-referential, Priest claims
elsewhere in the same book: “Though some have been tempted to deny it [i.e.,
the existence of inconsistent obligations], that there can be – indeed, are –
such situations is beyond doubt” (Priest 2006: 183). One could interpret the
first set of quotations as affirming that all statements are fallible or doubtable
and the last one as claiming that some statements are not doubtable. In this
case, it is obvious, as my interpretive principle suggests, that the universal gen-
eralization should take precedence over the other claim, even if the two are in
conflict with one another. By the same token, Hume’s explicit affirmation of
the trickle down argument clearly overrides any use of an ambiguous term in
other contexts.
One question remains: how might Hume go from the claim that it is possible
that epistemic egalitarianism is permissible to the view that it is permissible,
or perhaps somehow required? Answering this question forces us to confront
Hume’s argument for the second claim. Although I do not have time to mount
a complete interpretation of this argument in light of what we have discussed,
let me sketch a plausible path to an even more worrisome endpoint. Hume’s
second argument talks about a series of belief evaluations. Unfortunately, the
discussion is left at a very abstract level. But it appears that it can apply to
different types of beliefs. So consider beliefs about “core” logical principles
such as lnc or mathematical beliefs such as 1+1=2. Even in Hume’s time many
recognized that philosophers debated every principle. Richard Popkin point-
ed out a long time ago that Hume “avidly” read Bayle (1955: 69), who, among
other things, expressed “complete skepticism… [about] mathematics” (1955:
66). And Bayle’s famous predecessor, Montaigne, attacked our knowledge of
logic with liar-like paradoxes (Popkin 2003: 53).19 Moreover, much to Berkeley’s
consternation, mathematicians at this time were working with an inconsistent
theory of infinitesimals (see Arnold and Shapiro 2007: 288). So it is not surpris-
ing when Hume claims: “There is nothing which is not the subject of debate,
and in which men of learning are not of contrary opinions. The most trivial
question escapes not our controversy, and in the most momentous we are not
able to give any certain decision… Amidst all this bustle ’tis not reason, which
carries the prize, but eloquence; and no man needs ever despair of gaining
proselytes to the most extravagant hypothesis, who has art enough to represent
it in any favourable colours” (TIntro 2, xiii-xiv). What, precisely, did Hume make
of such fundamental disputes among philosophers? On some plausible princi-
ples about epistemic disagreement, we should, roughly speaking, consider two
views as being supported by equal evidence if they are advocated by different
groups with equal epistemic credentials. If Hume adopts a view in this neigh-
borhood, then we have equal evidence for trivialism and non-trivialism; and if
we have as much evidence that the notion of evidence should be extinguished
as we do that it is helpful, we lack sufficient evidence to consider any belief
19 Moreover, Montaigne also attacked our knowledge of geometry (see Popkin 2003: 53).
60 Meeker
as backed by any evidence.20 If this is not epistemic egalitarianism, I do not

know what is. In short, the possibility of rationally accepting any philosophical
position should lead us to consider the actual arguments some have used to at-
tack what others have thought to be “certain.” Refusing to acknowledge such a
possibility allows some to ignore the actual debates, historically and presently,
and dismiss the epistemic peerhood of other philosophers. Recognizing such a
possibility requires us to get our hands dirty.
5 Conclusion
To wrap up, I have maintained that Hume’s trickle down argument shows that,
for Hume, there is no knowledge because all claims to knowledge are defeated,
not by reflection, but by our fallibility and cognitive limitations. If we did have
certain knowledge then we would be able to isolate the precise limit at which
certainty stops and uncertainty begins. Hume’s example is mathematics but I
have argued that we can apply the same reasoning to the demonstrative sci-
ence of logic. Debates about dialetheism and, more recently, trivialism, seem to
follow the pattern of Hume’s reasoning and highlight how what we may think
of as obviously certain logical principles are still debatable and thus uncertain
to some extent. Moreover, conjoining this radical holism consequence with
certain plausible views about epistemic disagreement allows us to see how one
might argue that epistemic egalitarianism is required rather than possibly per-
mitted. For once we allow for this possibility and examine the actual debates
in this area, it seems that we are rationally permitted to accept trivialism and
thus believe any proposition. But then we seem to have undercut the notion of
evidence; and we have arrived at a view functionally equivalent to epistemic
egalitarianism. That is, in a very real sense we seem to be rationally permitted
to accept the conclusion of Hume’s second claim that considerations of logic
extinguish evidence. A worrisome conclusion indeed.
20 One can raise various objections here. As Kevin McCain has put it to me: “It’s not clear that
we should think that the groups supporting these two views have equal epistemic creden-
tials. If nothing else, there are more people (with independent backgrounds) in the non-
trivial group than the trivial one. Presumably, the opinion of x number of highly qualified
people can be outweighed by the number of x + n highly qualified people when n is high
enough. Shouldn’t we think this is what we find when it comes to this case?” I doubt that the
number of philosophers in a group matters much to the issues of epistemic disagreement or
peerhood. But because I am simply at this point sketching a way to connect Hume’s first and
second conclusions, I lack the time to deal with these interesting and complicated issues.
Bibliography
Ainslie, D. 2015. Hume’s True Scepticism. New York: Oxford University Press.
Arnold, J. and S. Shapiro. 2007. “Where in the (World Wide) Web of Belief is the Law of
Non-Contradiction?,” Noûs 41(2): 276–297.
Atkinson, D. and J. Peijnenburg. 2006. “Probability without certainty: foundationalism
and the Lewis–Reichenbach debate,” Studies in History and Philosophy of Science
Part A 37(3): 442–453.
Brown, B. 2004. “Knowledge and Non-Contradiction.” In Priest et al. (2004), 126–155.
Bueno, O. and M. Colyvan. 2004. “Logical Non-Apriorism and the ‘Law’ of Non-Contra-
diction.” In Priest et al. (2004), 156–175.
Couvalis, G. 2011. “A Modern Malignant Demon? Hume’s Scepticism with Regard to
Reason (Partly) Vindicated.” In Craig Taylor, Stephen Buckle (eds.), Hume and the
Enlightenment, 105–111. London: Pickering & Chatto.
Deguchi, Y., J.L. Garfield and G. Priest. 2013. “Two Plus One Equals One: A Response to
Brook Ziporyn,” Philosophy East and West 63 (3): 353–358.
De Pierris, G. 2015. Ideas, Evidence, and Method: Hume’s Skepticism and Naturalism con-
cerning Knowledge and Causation New York: Oxford University Press.
Durland, K. 2011. “Extreme Skepticism and Commitment in the Treatise,” Hume Studies
37 (1): 65–98.
Estrada-González, L. 2012. “Models of Possibilism and Trivialism,” Logic and Logical
Garrett, D. 2015. Hume. New York: Routledge.
Huemer, M. 2002. Epistemology: Contemporary Readings New York: Routledge.
Hume, D. 1990. A Treatise of Human Nature, edited by L.A. Selby-Bigge 2nd ed., rev., ed.
P.H. Nidditch. Hong Kong: Oxford University Press.
Hume, D. 1999. An Enquiry concerning Human Understanding, ed. L Tom. Beauchamp
Oxford: Oxford University Press.
Hume, D. 2000. A Treatise of Human Nature, edited by David Fate Norton and Mary J.
Norton. New York: Oxford University Press.
Kail, P.J.E. 2015. “Review of Kevin Meeker, Hume’s Radical Scepticism and the Fate of
Naturalized Epistemolgy,” Dialectica 69 : 623–630.
Kroon, F. 2004. “Realism and Dialetheism.” In Priest et al. (2004), 245–263.
Lewis, D. 1982. “Logic for Equivocators,” Noûs 16: 431–441.
Lewis, D. 2004. “Letters to Beall and Priest.” In Priest et al. (2004), 176–177.
Littman, G. and K. Simmons. 2004. “A Critique of Dialetheism.” In Priest et al. (2004),
314–335.
Meeker, K. 2007. “Hume on Knowledge, Certainty, and Probability: Anticipating the
Disintegration of the Analytic/Synthetic Divide?,” Pacific Philosophical Quarterly 88
(2): 226–242.
62 Meeker
Meeker, K. 2008. “Was Hume Mathematically Challenged?,” Paper presented at 35th

International Hume Conference in Iceland.
Meeker, K. 2011. “Quine on Hume and the Analytic/Synthetic Distinction,” Philosophia
39(2): 369–373.
Meeker, K. 2013. Hume’s Radical Scepticism and the Fate of Naturalized Epistemology.
New York: Palgrave Macmillan.
Meeker, K. and T. Poston. 2010. “Skeptics without Borders,” American Philosophical
Quarterly 47 (3): 223–237.
Morabia, A. (2013). “Epidemiology in History: Hume, Mill, Hill, and the Sui Generis
Epidemiologic Approach to Causal Inference,” American Journal of Epidemiology
178(10): 1526–1532.
Owen, D. 1999. Hume’s Reason. New York: Oxford University Press.
Popkin, R. 1955. “The Skeptical Precursors of David Hume,” Philosophy and Phenomeno-
logical Research 16 (1): 61–71.
Popkin, R. 2003. The History of Scepticism: From Savonarola to Bayle. Revised and Ex-
panded Edition. New York: Oxford University Press.
Priest, G. 2000. “Could Everything be True?” Australian Journal of Philosophy 78 (2):
189–195.
Priest, G. 2006. In Contradiction: A Study of the Transconsistent, 2nd ed. New York: Ox-
ford University Press.
Priest, G., J.C. Beall and B. Armour-Garb. 2004. The Law of Non-Contradiction: New Phil-
osophical Essays. New York: Oxford University Press.
Quine, W.V.O. 1980. “Two Dogmas of Empiricism.” In From a Logical Point of View, 2nd
ed. Cambridge, Massachusetts: Harvard University Press.
Resnik, M. 2004. “Revising Logic.” In Priest et al. (2004), 178–194.
Rinard, S. 2017. “No Exception for Belief,” Philosophy and Phenomenological Research
94 (1): 121–143.
Shapiro, S. 2004. “Simple Truth, Contradiction, and Consistency.” In Priest et al. (2004),
336–354.
Williamson, T. 2000. Knowledge and Its Limits. Oxford: Oxford University Press.
Williamson, T. 2005a. “Precis of Knowledge and its Limits,” Philosophy and Phenomeno-
logical Research 70: 431–435.
Williamson, T. 2005b. “Contextualism, Subject-Sensitive Invariantism, and Knowledge
of Knowledge,” The Philosophical Quarterly 55: 213–235.
Zalta, E. 2004. “In Defense of the Law of Non-Contradiction.” In Priest et al. (2004),
418–436.
Chapter 5
Skepticism as a Way of Life

Baron Reed
1 Factual Disputes and Conflicts in Value
Consider a deep and longstanding disagreement in philosophy: what does it

mean to be free and to have responsibility for our actions? The opposing views
in this debate are well-entrenched. On one side, some philosophers say that we
have a robust sort of free will that underwrites a very substantial responsibility
for our own actions; although we may be tempted to undertake an action, what
we do is ultimately up to us. On the other side are philosophers who say that
freedom is a matter of being integrated in the right way into the same causal
network that governs every event in the universe; our responsibility is at best
conditional and qualified.
Much of the discussion in this area has focused on whether the robust free-
dom characteristic of the first sort of view can exist in a world that operates in
the way science indicates. Even if determinism is not true, it is still difficult to
see how things like beliefs, desires, and intentions fit into a physicalistic under-
standing of the universe. But despite this difficulty, the debate has not ceased.
There are at least two reasons for this. First, even if it could be definitively prov-
en that the robust sort of freedom is impossible for beings like us, at least some
of those who have defended it would not thereby be persuaded to join their op-
ponents. They would instead continue to say that the robust conception tells
us what freedom is—it is simply an unfortunate fact that we are incapable of
having it.1 Second, and related, some philosophers on both sides of the debate
have come to think of the disagreement as being driven by a difference, not in
beliefs about some fact of the matter, but in values. Philosophers in the rival
camps disagree because they have different ideas about why freedom matters
and how it connects to other things that matter, like autonomy or meaningful-
ness in life.2 Showing that robust freedom is impossible doesn’t put an end to
the debate if it turns out to be essential to the value system that animates our
lives. For this reason, then, any attempt to think through the free will debate
that does not address the question of value is hopelessly incomplete.
1 See, e.g., Strawson (1986) and Pereboom (2001).

2 See Dennett (1984) and Kane (1996: 14−16).

64 Reed
Something similar, I believe, can be said about the equally old and deep
disagreement between skeptics and anti-skeptics. Consider the ways in which
G.E. Moore responds to skepticism about the external world. In one version of
his response, Moore argues that we can reject the philosophical assumptions
underlying the skeptical argument because none of them could be more cer-
tain than the perceptual knowledge he has of the world around him.3 We don’t
even need to identify the specific flaw in the skeptical argument; the difference
in epistemic status between it and the perceptual beliefs it targets is enough to
allow us to dismiss the argument.4 In the second version, his famous “proof of
an external world,” Moore argues:
(1) Here is a hand.

(2) Here is another hand.
(3) Hands are external objects. Therefore,
(4) There is an external world.
Although it is directed in the first instance against the idealist, it is clear in

Moore’s commentary on what the proof is supposed to accomplish that he also
took it to have anti-skeptical implications.5 As with the first version, Moore does
not tell us in much detail how he knows the first two premises to be correct. He
simply says that it would be “absurd” to deny that he knows he has hands.
It is fair to say that both versions of the Moorean response to skepticism
have been controversial. Even those philosophers who are not sympathetic to
skepticism have felt uncomfortable at what seems to be a dogmatic dismiss-
al of the skeptical challenge. In effect, Moore poses a stark choice to us: we
can either reject the legitimacy of the philosophical pursuit of an account of
knowledge or leave ourselves vulnerable to a total, isolating skepticism. Many
philosophers have tried to come to grips with this dialectical stand-off by
showing precisely where either Moore or the skeptic (or perhaps both) goes
wrong. If one side or the other is implicitly relying on some epistemic principle
that can be abandoned or modified, we can see our way to rejecting skepticism
without lapsing into a problematic dogmatism.
Although a great deal of interesting work in epistemology has come out of
this way of approaching the problem, I do not think it will ever be sufficient by
itself to bring the dispute over skepticism to a satisfying resolution—the same
dynamic we saw in the debate over freedom and responsibility is at play here
as well. Suppose a definitive case could be made for the claim that skepticism
3 Moore (1959a). See also Lycan (2001).

4 For objections to this sort of Moorean reply, see Frances (2005: 167) and Reed (2007: 259−261).
5 See Coliva (2018: 469).
Skepticism as a Way of Life 65
arises out of a robust conception of knowledge that is impossible to satisfy.

Even if this were so, some of the philosophers who have defended it might still
argue that this robust conception is correct about what knowledge is. It is just
an unfortunate fact that we are incapable of having it.6 And they might be driv-
en to this conclusion because they take the robust conception of knowledge to
be of significant value and perhaps to be essentially connected to other aspects
of our intellectual lives that are also valuable, like understanding, philosophi-
cal assurance, or the ethics of belief.7
If this is correct, then any consideration of skepticism that fails to address
the question of value in epistemology will be inadequate.8 In what follows,
then, I will trace what happens when the debate between skeptics and anti-
skeptics is shifted from a dispute over the truth of epistemic principles to a
disagreement over epistemic values and the intellectual lives they make pos-
sible. Understanding the debate in this way would seem quite natural to the
ancient skeptics and their opponents. The apraxia, or inaction, objection to
skepticism was perhaps the most prominent response the ancient dogmatists
used.9 A modern version of the argument, voiced by David Hume and devel-
oped by Myles Burnyeat, also focuses on the supposed inability of the skeptic
to “live his skepticism.” After showing how the Pyrrhonists replied to this sort
of argument, I go on to develop a series of new ethical objections to the Pyr-
rhonist way of life. Although it is not impossible, a Pyrrhonist life proves to
be far from desirable. This does not mean, however, that the Moorean wins
the debate. I conclude by showing how the Academic skeptical tradition has
resources that point toward a way of life that is healthier and more admirable
than dogmatism permits.
6 See, e.g., Fumerton (1995). See Sosa (2009: 173–177) on why the “desire for a fully general,
legitimating, philosophical understanding of all our knowledge is unfilfillable.”
7 See Fumerton (1995: Ch. 6; 2016) on philosophical assurance. See Clifford (1877) on the ethics
of belief.
8 See, for example, Nozick (1981). His response to skepticism rests on rejecting the closure prin-
ciple, thereby blocking the rational connection between being unable to know that one is
not dreaming and one’s purported everyday knowledge. When considering what his tracking
view has gained for us, Nozick says that it “illuminates and explains how knowledge is pos-
sible,” though it “does not tell us how to find out whether a particular belief that p is knowl-
edge” (1981: 287). Without the ability to tell whether a particular belief is knowledge, however,
it may prove to be impossible to attain some of the other goods knowledge has been thought
to provide, like understanding or guidance in action. By contrast, Williams (1996) not only
rejects foundationalism, because he takes it to be the basis for skepticism, but he also exerts a
great deal of effort in trying to convince us that a theory of knowledge in general is unnatural
and not something we should want to have (1996: 45−46). In other words, he argues that foun-
dationalism is not merely false, it arises from an intellectual endeavor we ought to abandon.
9 Vogt (2010: 165).
66 Reed
2 The Apraxia Objection
The dispute between skeptic and dogmatist in ancient philosophy was framed
by broadly ethical concerns. Two claims from Stoic epistemology drew the at-
tention of Arcesilaus, the first Academic to move that school in the direction
of skepticism:
(i) The sage will never assent to mere opinions—i.e., beliefs that are inad-
equately grounded and could turn out to be false; and
(ii) The cognitive (or apprehensive) impression is our connection to the
truth.10
Arcesilaus agreed with the first claim, about what was proper for a sage, but he
thought we were incapable of meeting the conditions laid out in the Stoics’ ac-
count of the cognitive impression.11 The conclusion that Arcesilaus was thought
to have drawn was that knowledge is impossible.12 This conclusion was then
combined with some other Stoic doctrine to form the basis of the apraxia ob-
jection—viz., that the skeptic’s arguments threaten to make action impossible.
There are two versions of the apraxia objection, depending on which
additional Stoic doctrine is used.13 According to the first version, skepticism is
incompatible with rational action. This is so because knowledge is necessary for
10 The early history of skepticism is somewhat messy. The later Academics often regarded
Socrates as a skeptic himself and so would not have taken Arcesilaus to be introducing
an entirely new stage into the Academy. Most scholars would disagree with this under-
standing of Socrates. There is also a lack of clarity over where the familiar skeptical ar-
guments were first formulated. Many of them are known to us through the writings of
Sextus Empiricus (from the second or third centuries ce), but what we know as the Pyr-
rhonist school was founded, not by Pyrrho himself in the late fourth century bce, but by
Aenesidemus in the first century bce. Aenesidemus did so after leaving the Academy,
and we simply don’t know whether—or to what extent—he carried with him some of
the intellectual legacy of Arcesilaus and Carneades. There is some disagreement about
the extent to which Pyrrho’s views were skeptical in nature, but it is safe to say that his
primary concern was ethical; see Bett (2000) for a reading of Pyrrho as a dogmatist rather
than a skeptic and Perin (2018) on the two readings of Pyrrho.
11 Reed (2002). Arcesilaus argued that, for each purportedly cognitive impression, a false
impression indiscernible from it could be found; if this is the case, then no impression can
be genuinely cognitive.
12 A more reasonable conclusion—more reasonable insofar as it does not run the risk of be-
ing self-undermining—would be that knowledge as the Stoics understood it is impossible.
See Ioppolo (2018: 38) for this interpretation of Arcesilaus. See Reed (2007) on this sort of
conditional reading of skeptical conclusions more generally.
13 See Perin (2010: 99–100) and Vogt (2010).
rational action. According to Origen, the Stoics held that “A rational animal…
has reason which passes judgment on impressions, rejecting some of these
and accepting others, in order that the animal may be guided accordingly.”14
Presumably, action without the benefit of guidance from reason would be pos-
sible, even though irrational. In contemporary epistemology, a similar sort of
view—that knowledge is the norm of action or of practical rationality—has
been defended by a variety of philosophers.15
On the second version of the apraxia objection, skepticism is incompatible
with action of any kind. According to Plutarch, Chrysippus held that “without
assent there is neither action nor impulsion [i.e., the movement of thought to-
ward action].”16 Because the skeptic’s arguments aim to undermine our ability
to assent, they will also undermine our ability to act. The classic expression of
this form of the apraxia objection can be found in Hume’s Enquiry Concerning
Human Understanding:
[A] Pyrrhonian … must acknowledge, if he will acknowledge any thing,

that all human life must perish, were his principles universally and
steadily to prevail. All discourse, all action would immediately cease; and
men remain in a total lethargy, till the necessities of nature, unsatisfied,
put an end to their miserable existence. [Enquiry xii.23]
More recently, Burnyeat cites Hume approvingly and argues that the skeptic’s
methodology of attempting to find an equipollence between different theories
is impossible without belief—after all, the skeptic would come to suspend
judgment only if he really believed that those theories are counterbalanced.
Both versions of the apraxia objection are only as compelling as the
additional Stoic doctrines on which they depend. It is not surprising that the
skeptics rejected both of them. I will return to the Academics’ strategy for
resisting the apraxia objection below. First, I will examine the more familiar
Pyrrhonian way of life, which can be seen as their attempt to remove the threat
of apraxia.
14 Long and Sedley (1987: 53A).

15 See, e.g., Hawthorne and Stanley (2008) for the knowledge norm of action. See Greco
(2012) for a contemporary version of the apraxia objection, grounded in the claim that
behavior must be rationalized by beliefs.
16 Cf. Plutarch, On Stoic self-contradictions, in Long and Sedley (1987: 53S). See also Cicero’s
Academica ii.108.
68 Reed
3 Pyrrhonism as a Way of Life
According to Sextus, Pyrrhonism
is an ability to set out oppositions among things which appear and are
thought of in any way at all, an ability by which, because of the equipol-
lence in the opposed objects and accounts, we come first to suspension
of judgment and afterwards to tranquility. [PH I.4]
It is to be understood, not as a body of doctrine—there are no beliefs all Pyr-

rhonists are expected to hold—or a theory, but rather as an ability to produce
a life marked by tranquility. That state of tranquility is to be reached through
the deliberate engineering of suspension of judgment.
At first glance, it may seem as though the Pyrrhonist will enjoy the peaceful-
ness of outer space, as her mental life becomes a void in the way Hume feared,
but the Pyrrhonists identified an alternative criterion for action. Instead of be-
lief, a Pyrrhonist may rely on “what is apparent,” where this will include:
(a) Guidance by nature, as when one eats when hungry and drinks when
thirsty;
(b) Necessitation by feelings, as one may break off a friendship when angry or
begin a relationship when moved by romantic attraction;
(c) Handing down of laws and customs, which will structure many social in-
teractions, as when one gets married or educates one’s children; and
(d) Teaching of kinds of expertise, as may happen when one pursues a career
as a doctor or a teacher.17
Appearances from each of these sources allow the Pyrrhonist to make her way,
capable of facing the many difficulties that arise in everyday life, while unbur-
dened by the unsatisfiable desire to know how the world really is.
As I noted above, Burnyeat thinks that the Pyrrhonian way of life is impossi-
ble because appearances are incapable of playing the role Sextus says they do.
Although I am inclined to think that Burnyeat is imposing on the Pyrrhonists
an unfriendly interpretation of what appearances are, I will not take up that
question here.18 Even if we grant to the Pyrrhonists the viability of living by
17 Cf. PH I.11. Something similar can be found, in a provisional way, in part three of Des-
cartes’s Discourse on the Method.
18 See Burnyeat (1980, 1984). For an influential interpretation of Pyrrhonian appearances
opposed to Burnyeat’s, see Frede (1979, 1984). For criticism of Frede’s interpretation, see
Perin (2010: Ch. 3). See Brennan and Roberts (2018: 128–137) for a helpful guide to the
complex dispute between Burnyeat and Frede.
following what is apparent as a substitute for belief, in the way sketched above,
there are four additional reasons to think that it would not be an e thically at-
tractive way of life.
First, where the first two sources of appearances may be rather primitive—
e.g., the sensation of hunger or the simple emotion of anger—the two latter
sources will have to be conceptually much richer. Take, for example, the cus-
toms we follow in raising our children. We teach them to be generous, kind to
others, polite to adults, and so forth. A Pyrrhonian parent may lack the belief
that each of these character traits is genuinely a moral virtue—i.e., identified
as such by the correct moral theory—but the lessons she imparts to her chil-
dren may nonetheless be identical in content to those of, say, an Aristotelian
or a Christian parent. In this sense, her life and the life of her children will be
shaped by dogmatic views in much the same way as the dogmatist’s family life
will be. If the Pyrrhonist is motivated by a desire for tranquility and a liberation
from dogmatic theorizing, it is hard to see how this has been achieved. Belief
is not the only vector by which dogmatism can be passed from one person to
another—appearances, it seems, can also be a medium of transmission.
Second, Sextus tells us that appearances “depend on passive and unwilled
feelings and are not objects of investigation” (PH I.11). This aspect of Pyrrhonism
is essential if the school is to avoid a kind of self-contradiction. One might ask
them if they know that suspension of judgment will lead to tranquility? If so,
aren’t they committed to a philosophical theory after all? Sextus’s response
is to say that the transition from suspension of judgment to tranquility was
discovered by accident.19 They pursue tranquility as a matter of practice, and not
because they have intellectually reasoned to the view that it will arise in these
circumstances. In this way, they are simply following the guidance of nature. The
problem, however, is that there are many cognitive appearances to which we are
naturally susceptible, or that are passed on as a matter of law and custom, and
yet that are also entirely objectionable. Racism, sexism, superstition—these
19 Sextus makes the point by means of an analogy: Apelles the painter was trying to capture
the appearance of foam on a horse’s nostrils after it has been running hard. Despite his
best efforts, every attempt failed. Finally, in a fit of frustration, he threw the sponge he
used for cleaning his brushes at his painting of the horse. It landed on the horse’s muzzle
and produced exactly the effect he had desired. In something like this way, the Pyrrhonist
is someone who began philosophical inquiry with the aim of finding the right sort of life
through discovering the truth. As every attempt ends in failure, because every view can
be shown to be no better than its rivals, the Pyrrhonist comes to suspend judgment, and
tranquility follows as a matter of nature, “as a shadow follows a body” (PH i.28–29).
70 Reed
are often “passive and unwilled feelings” of superiority or hatred or fear, and yet
they absolutely should be subject to investigation and critique. Pyrrhonism, if
widely adopted, would produce a stultifyingly conservative society, as the laws
and customs passed down from one generation to the next might drift a little as
tastes change but could never be consciously improved or replaced.
Third, “what is apparent” is not so easy a criterion of action as Sextus seems to
think. The appearances we follow will very often be in conflict with one another.
For example, one might feel hungry but be in circumstances where eating the
available food would be impolite or immoral or even illegal. Which of these ap-
pearances is the one to follow? Or suppose a friend asks you for a favor: assis-
tance with moving from one house to another. You might feel the customary
obligation to help a friend, but that might be opposed by the feeling of dread
at spending an entire weekend in backbreaking labor. What are you to do? If
these feelings are entirely passive and not open to investigation, you might give
way to the dread, saying to your friend that the feeling has necessitated this out-
come. Such an excuse is unlikely to placate your friend, however, as it seems
to be a claim offered in Sartrean bad faith: in hiding behind your dread, you
are trying to obscure your own responsibility for this action. The Pyrrhonist’s
attempt to defer to what is natural, then, can too often look like nothing more
than a false denial of agency.
Fourth, even setting aside these issues—which would have been just as
troubling in the Hellenistic period as they are now—there are cultural dif-
ferences between Sextus’s time and ours that preclude the viability of Pyr-
rhonism. As a matter of methodology, we simply don’t have the wide variety
of equally plausible views that were available to the ancient skeptic. Where
Sextus might oppose an Aristotelian understanding of matter with one taken
from the Stoics, we have nothing to counterbalance with the quantum theory
of the atom. Consensus does not, of course, extend to every corner of science,
but there is a very large area of basic agreement nonetheless. As a matter of
the aim of skeptical inquiry, removing our ability to make use of this scientific
consensus would not bring anything like tranquility. Rather, it would simply be
a profound intellectual loss.
A change in religious culture from the ancient world to the modern points
to another way in which tranquility would be highly unlikely to follow on the
heels of suspension of judgment. The polytheistic religions of the ancient world
tended to be relatively accommodating and syncretistic, making room for one
another as needed. Seeing immigrants to your city building a temple to their
foreign gods did not really threaten your relationship with your own favored
god. With the rise of the monotheistic religions, however, this strategy of ac-
commodation was no longer welcome, as having the correct t heological beliefs
became essential to your salvation. Being in a position where you did not know
the proper way to worship the divine was now terrifying, not liberating.20
For us in the modern world, then, skepticism has threatened to land us in
that most untranquil state: doubt. In these circumstances, most of us would
rather resist skepticism than cultivate it. Given the Moorean choice between
dogmatism and skepticism, philosophers have tended to choose dogmatism,
even with its inability to provide intellectual satisfaction. But this outcome is
not inevitable. If we can reject the Moorean choice—as I shall argue we can—
it may be possible for us to develop a kind of skepticism we can live with.
4 A Return to Academic Skepticism
The central problem with the Pyrrhonist outlook is that it can find no way of
responding rationally to our imperfect intellectual condition. As a result, the
Pyrrhonists retreat to what is only natural and thus supposedly beyond the
scope of any rational evaluation. Given the ideal of the sage as someone who
never holds mere opinions, this is their only way of trying to abide by it. As we
have seen though, the Pyrrhonists are not, ultimately, able to evade criticism
for their ways of thinking about the world.
For the Academics, by contrast, the response to the apraxia objection does
not involve a retreat from rationality. The differences between the Pyrrhonist
and the Academic outlooks are—with one crucial exception, which I will ex-
plain below—fairly well captured by Sextus:
For the Academics say that things are good and bad not in the way we do,
but with the conviction that it is plausible that what they call good rather
than its contrary really is good (and similarly with bad), whereas we do
not call anything good or bad with the thought that what we say is plau-
sible—rather, without holding opinions we follow ordinary life in order
not to be inactive. [PH I.226]
20 See Pascal on the terror of doubt: “On contemplating our blindness and wretchedness,
and on observing the whole of the silent universe, and humanity with no light aban-
doned to itself, lost in this nook of the universe not knowing who put us there, what we
have come to achieve, what will become of us when we die, incapable of all knowledge,
I become frightened, like someone taken in his sleep to a terrifying, deserted island who
wakes up with no knowledge of what has happened, nor means of escape” (Pensées, 229).
Uriel da Costa, Jewish by descent but raised by necessity as a Christian in seventeenth-
century Iberia, tried to rejoin the Jewish community in Amsterdam to which Spinoza’s
family belonged, but tragically he committed suicide as he found it impossible to straddle
the divide between the Christian and Jewish worlds; see Nadler (1999: 66–73).
72 Reed
There are two central elements to the Academic outlook. First, Carneades
proposed an alternative to the Stoics’ account of the cognitive impression. In its
place, he suggested that we make use of merely plausible (probable, persuasive)
impressions (Cic. Ac. ii.99, 101, 110). Although there is “no distinctive mark of
truth or certainty” (Cic. Ac. ii.103), it is nonetheless possible to act on the basis of
these plausible impressions—as, indeed, even the Stoics must, for no one who
sets out on a journey by sea or marries and has children can be certain that all
will turn out for the best, though it may seem plausible that it will (Cic. Ac. ii.109).
The second element—and the one that Sextus doesn’t get quite right—has
to do with the attitude an Academic will take toward these plausible impres-
sions. According to Clitomachus, Carneades’s closest follower, the latter drew a
distinction between two ways of withholding assent (Cic. Ac. ii.104). The first
is to withhold assent in the fully dogmatic sense; the Academic, like the Pyr-
rhonist, will withhold assent to everything in this sense. The second way of
withholding assent is to not even give any sort of response at all, not saying
“yes” or “no” to anything. The Academic does not always withhold assent in this
sense. To put the point more straightforwardly, the Academics will sometimes
assent in this limited way; they will do so when an impression is plausible and
“unimpeded”—i.e., not in conflict with other plausible impressions. Holding
on to assent in this limited way allows them “to move and act” (Cic. Ac. ii.104),
removing the threat of the apraxia objection. It also allows the Academics to
continue to take part in philosophical investigation as a rational endeavor,
though without the danger of becoming too strongly committed to their own
views. In this way, the Academics take a middle course between the extremes
of Stoic dogmatism and Pyrrhonist quietism.
These two kinds of assent can be called the Stoic model and the Academic
model, respectively.21 Just as the Stoic sage will assent to cognitive impres-
sions in the fully dogmatic, Stoic way, so the Academic will assent to plausible
impressions in the limited, Academic way. But the Academics do not simply
retreat to their model of assent because it is the best one can do in a hopelessly
flawed world. Rather, they use it to formulate a new intellectual ideal—a new
21 The difference between the two models can be seen in relation to the dogmatism para-
dox; see Kripke (2011) and Harman (1973: 148). That paradox arises because it seems as
though, if you know that p, then you know that any evidence to the contrary you might
encounter would have to be misleading and could therefore be ignored—and yet ignor-
ing evidence seems to be a problematic thing to do. If our beliefs are stable, we would
like it to result from a consideration of all the relevant evidence. Given their model of
assent, the Stoics face the dogmatism paradox; without an explanation of how a purport-
edly cognitive impression could be wrong, they run the risk of becoming dogmatic in a
problematic way. For the Academics, on the other hand, the fallibility of the plausible
impression provides no basis for rejecting counterevidence.
c onception of what it is to be a sage: “where there are persuasive <impres-

sions>, the wise person won’t lack something appropriate to do or say.”22 Notice
that the Academic sage is not merely acting or speaking through the guidance
of nature. What she does and says is not merely natural but appropriate, where
the latter is a term of normative evaluation.
The development of a skeptical sage has gone relatively unnoticed—
certainly, in comparison to the attention paid to the Pyrrhonist way of life—for
at least two reasons. First, the textual record is regrettably spotty. Our best
sources—Cicero and Sextus—were writing well after the central disputes
between Stoics and Academics occurred. Sextus’s aim is also suspect insofar
as he often presents the Academics as having “asserted that things cannot be
apprehended” (PH I.3)—a position that threatens to be self-contradictory, if
assertions are supposed to be cognitively well-grounded. He may do this in
the attempt to mark the distinction between Pyrrhonism and Academic skep-
ticism in a way as favorable to the former as possible.23
Second, Cicero’s Academica—the most important source for Academic
skepticism—survives only in two fragments of different versions of his dia-
logue. Moreover, the dialogue itself is presented as occurring at a specific point
in the development of the Academy which is itself quite difficult to understand.
After Carneades’s death, his legacy—and that of the Academy more broadly—
became contested by four of his followers.24 Clitomachus apparently defended
the view attributed to Carneades here: the sage will assent in only a limited
way to plausible impressions. Philo of Larissa thought that, according to Car-
neades, the sage will assent to plausible impressions strongly enough to “have
opinions” (Cic. Ac. ii.59).25 Antiochus moved toward a dogmatic interpretation
of Academic philosophy, drawing on what he took to be a shared Platonism;
22 Cic. Ac. ii.110. (The angle brackets indicate an editorial addition to the text for the sake of
clarity; I have added italics for emphasis.) Similarly, Arcesilaus, the first of the Academic
skeptics, allows that “one who suspends judgment about everything will regulate choice
and avoidance and actions in general by ‘the reasonable’” (Sextus, M 7.158; Long & Sedley
1987: 69B).
23 Sextus’s efforts have largely been successful in this regard. See Klein (2002) for a present-
day echo of this way of distinguishing the Pyrrhonists and Academics.
24 See Brittain’s introduction to Cicero (2006), xxiii–xxxi, and Striker (1997).
25 Cicero himself sometimes says that he thinks Carneades advanced this sort of view for
dialectical purposes without actually accepting it (Ac. ii.78). (Similarly, some scholars
think that Carneades also developed the account of the plausible impression in a dia-
lectical way, as an alternative that could be set in opposition to Stoic epistemology, but
without actually endorsing it; see, e.g., Allen (1994).) Cicero says of himself that he holds
opinions, but he also says that he is not wise and thus not the subject of discussion (Ac.
ii.66). Another character in the dialogue, Catulus, records the opposite view, saying that,
for Carneades, “the wise person will assent to something he hasn’t apprehended—that is,
hold opinions—but in such a way that he understands that it is an opinion and realizes
74 Reed
in practice, it meant he defended Stoic epistemology against the objections

raised by Arcesilaus and Carneades. And, finally, Aenesidemus thought that
Carneades had betrayed the skeptical impulse; he left the Academy to recreate
(or, perhaps, simply to create) the Pyrrhonist school.26
As complicated as all this is, the occasion for the conversation Cicero pre-
sented in the Academica is the reception of Philo’s so-called Roman Books. Al-
though these do not survive, scholars have tried to reconstruct the view he
advanced there. Perhaps the most intriguing possibility is that Philo moved
beyond skepticism altogether, claiming not merely that the sage will assent in
some limited way to plausible impressions, but also that doing so will allow her
to have a fallible sort of knowledge.27
Philo may have thought that he was steering a middle course between the
more stringent skepticism of Carneades on the one hand and the dogmatism
of the Stoics on the other. He failed to convince any of his contemporaries to
move in the direction of fallibilism; Cicero remained in the skeptical Academy,
and Antiochus took up the banner of infallibilist dogmatism.
Although the split between skepticism and dogmatism was taken to be de-
cisive in the ancient world, there is some reason to think that the line between
them can be at least partially effaced. According to one of the ways in which fal-
libilism is presently developed, the different degrees of epistemic justification
can be taken to underwrite different degrees of knowledge.28 All f allibilists al-
low that one can have knowledge that is less secure than certainty. This entails
that there are at least two different cognitive states: certainty and the lesser sort
of knowledge. But justification comes in a wide array of degrees—think, for
example, of how your perceptual belief that the animal in the distance is a gold-
en retriever becomes increasingly well justified as it walks closer to you—and
there is no reason for a fallibilist to deny that knowledge may similarly come in
a variety of degrees.
that nothing is apprehensible” (Ac. ii.148). Cicero associates this interpretation of Car-
neades with Philo and Metrodorus (Ac. ii.78).
26 And here we have another complication caused by the paucity of sources. As I mentioned
above, we simply do not know whether, and to what extent, Aenesidemus carried with
him arguments that had first been formulated in the Academy. In choosing Pyrrho as the
founder of his school—a figure who predated Arcesilaus—it may be that Aenesidemus
was trying to establish his school as the original, and purer, form of skepticism. Although
that is often how the Pyrrhonists are now seen, the reality is much more complicated and
obscure. For more on Aenesidemus and his relation to other philosophers, see Castagnoli
(2018).
27 See Brittain (2008). For other ways of interpreting Philo, see Striker (1997) and Tarrant
(2018).
28 Reed (2013). In addition to different degrees of knowledge, a fallibilist may also distin-
guish between different kinds of knowledge.
Knowledge pluralism, the claim that there are many different cognitive
states, allows us to say that a person can be knowledgeable for some practical
purposes and yet fail to be so for others.29 A student in the police academy
may know well enough that a particular sort of bomb can be safely disarmed
by cutting the blue wire when she is studying for a test, and yet she might be
properly reluctant to act on that knowledge when faced with a live bomb in
the field. A similar point holds for the different ways these cognitive states can
function with respect to varying theoretical purposes. Where an instance of
knowledge may be suitable for one sort of inquiry—say, inferring that a sa-
fari to see okapis won’t be economically viable because no okapis have been
seen in recent years—it may be inadequate for others, such as concluding that
okapis are now entirely extinct.
It should be clear that a neo-Academic epistemology of this sort is not
subject to the four objections to Pyrrhonism laid out above: (i) the a ppearances
are conceptually loaded; (ii) they may prevent problematic biases from being
overturned; (iii) the Pyrrhonists provide no guidance on how to resolve con-
flicts among appearances; and (iv) suspension of judgment threatens to cut us
off from valuable intellectual achievements. These objections arise from the
fact that Pyrrhonism collapses all epistemic distinctions—no belief is more or
less rational than any other. Following appearances without examining them
is the only way of preserving one’s capacity for action, though it does so at the
cost of significantly impairing one’s agency. Action remains possible, but only
because all pretense to rational action has been abandoned.
Neo-Academic epistemology, by contrast, preserves our ability to hold
one belief to be more rational, more justified, than another—even in cases
where both beliefs fail to count as knowledge in the sense relevant to one’s
current practical and theoretical circumstances. Because beliefs can be ra-
tionally related to one another, as well as to one’s perceptual experiences,
rational intuitions, and newly acquired testimonial evidence, an agent re-
tains the rational capacity to weigh conflicts among them. These beliefs can
be rationally ordered, at least in principle and even in spite of the fact that
any rational ordering we produce is liable to be corrected when new evidence
comes in. And, most importantly, nothing prevents us from continuing to ben-
efit from the genuine intellectual gains that have been made in science, history,
and other areas of inquiry. Indeed, the self-correcting nature of these disci-
plines is better explained by neo-Academic epistemology than by its rivals.30
29 See Reed (2013; unpublished).

30 The label “neo-Academic epistemology” is intended to mark a contrast with neo-
Pyrrhonism—see Fogelin (1994) and Lammenranta (2018)—and to indicate that we are
better off drawing inspiration from the Academic tradition than from the Pyrrhonists.
76 Reed
5 The Skeptical Sage
Neo-Academic epistemology, which is both fallibilistic and skeptical in nature,

turns out to provide a more compelling account of our cognitive abilities than
either Pyrrhonism or Moorean common sense. In identifying a wide range of
cognitive states, from perfect certainty down to a mundane and defeasible sort
of knowledge, the neo-Academic view preserves our capacity to respond to our
evidence rationally, without lapsing into either the negative dogmatism of Pyr-
rhonism or the positive dogmatism of Moore.31
This unique ability to resist dogmatism forms a central part of the Academic
conception of a skeptical sage. As we have already seen, this conception can be
found in Cicero’s Academica (e.g., at ii.110). It is later taken up and developed
at great length in the early seventeenth century by Pierre Charron, Montaigne’s
closest philosophical associate.32 In his treatise On Wisdom, Charron says that
there are two central dispositions that make it possible for us to be wise. The
first is freedom from error arising from the world or from one’s own passions
(1697: bk. ii, Ch. 1); the second is “a full, entire, and generous Liberty of Mind,”
consisting of liberty of judgment and liberty of the will (1697: bk. ii, Ch. 2, p.
16). Liberty of judgment does not arise from doxastic voluntarism—Charron
does not appear to be committed to thinking our beliefs are under the direct
control of the will—but from considering everything “carefully and without
Passion,” which will allow one to “find somewhat of Reason and Probability on
every side” (1697: 18). This leads the skeptical sage to grow comfortable with
doubt (1697: 33–34) and to take on a “Universality of Soul” that marks her as “a
Citizen of the World” (1697: 38).33
This expansive connection to the world at large stems from the skeptical
sage’s recognition that “others are at least in a possibility of being in the right”
(1697: 35). As Pierre Bayle will note, late in the seventeenth century, the err-
ing conscience has the very same warrant as the conscience that is guided by
the truth.34 Given this, as Charron says, it is clear that “nothing can be more
extravagant, than to imagine that a Man who is not one of Us cannot possibly
31 Ironically, the charge of negative dogmatism is usually lodged against the Academics; see
Sextus (PH I.3, 226). As we have seen, though, it is the Pyrrhonists who are vulnerable to
the charge of recommending the uncritical following of appearances.
32 See Maia Neto (2014: Ch. 2) and Paganini (2018) for more on Charron’s Academic concep-
tion of skeptical wisdom.
33 These values are at the heart of Stefan Zweig’s powerful rediscovery, in the early years of
World War ii, of Montaigne, “the uncontested herald” (2015: 39) of “the freedom of the
soul” (2015: 37).
34 Bayle (2005: 220). See also Laursen (2018: 360–363).
be a Good Man” (1697: 54). This freedom from dogmatism, then, not only par-
tially constitutes the liberty of mind so essential to the flourishing of the in-
dividual sage, it also motivates the sort of toleration for others that is utterly
foundational to the possibility of an open, democratic society.
Recall that the dispute between skeptic and dogmatist was earlier framed
as involving the sort of clash in values that tends to be at the heart of the most
intractable philosophical disagreements. Academic skepticism, and neo-
Academic epistemology more broadly, are attractive because they embody a
powerful set of values, including freedom of the mind, intellectual humility,
and toleration. But that is not the only reason to favor the skeptical way of life,
properly understood. The skeptical outlook also provides the most satisfying
way of understanding the true nature of these philosophical, value-laden dis-
agreements in general.
To see this, notice that straightforwardly factual disagreements ideally re-
solve with convergence on a single correct answer. Although there may be
good evidence on both sides of the factual dispute—e.g., I say the bird is a
robin because of its size and shape, while you say it is a northern flicker be-
cause of its distinctive white rump and yellow coloring on the underside of its
wings—careful consideration leads to the discarding of some of the evidence
(i.e., mine) as inconclusive. Your take on the situation is the true one, and mine
should be abandoned.
But this is typically not the case with value disagreements. Even when we
converge on a single answer to a disputed question, the ideal here does not
require us to discard the considerations that support the rejected answer. The
values that motivated one of the parties may be outweighed by the values en-
dorsed by the other party, but that doesn’t mean they are no longer valuable.
They may still matter, even if they matter less in the circumstances than the
opposing values do. This is why the wisest of judges will attempt to locate the
proper balance among conflicting values in difficult cases, acknowledging,
for example, that freedom of speech is still of great importance even when it
needs to be curtailed in cases where the speaker is inciting others to immedi-
ate violence.
The methodology of the wise judge here is adopted by the skeptic in philo-
sophical disagreements. Unlike the dogmatist, who attempts to utterly reject
the opposing view,35 the Academic skeptic seeks to gather the considerations
on either side. Values that are outweighed are nevertheless preserved in the
35 See, e.g., Reid’s claim that “Nature hath given us a particular emotion—to wit, that of ridi-
cule—which seems intended for this very purpose of putting out of countenance what is
absurd, either in opinion or practice” (1872: 567).
78 Reed
resulting balance. Toleration is often thereby made easier, as the distance be-
tween opposing sides may be at least reduced to different weightings of con-
siderations that are valued to some extent by all parties to the disagreement.
This is, in a practical sense, the “largeness of soul” Charron took to characterize
the skeptical sage. By opening our own minds to the perspectives of others in
this way, we enable ourselves to enter into a larger community of rational peers
and truly become—together—citizens of the world.36
Bibliography
Allen, J. 1994. “Academic Probabilism and Stoic Epistemology,” Classical Quarterly 44:
85–113.
Bayle, P. 2005 [1686–1688]. A Philosophical Commentary on the Words “Compel Them to
Come In”…, ed. by J. Kilcullen and C. Kukathas. Indianapolis: Liberty Fund.
Bett, R. 2000. Pyrrho, his Antecedents, and his Legacy. Oxford: Oxford University Press.
Brennan, T. and C. Roberts. 2018. “Sextus Empiricus.” In D. Machuca and B. Reed (eds.),
Skepticism: From Antiquity to the Present, 125–144. London: Bloomsbury.
Brittain, C. 2008. “Philo of Larissa,” Stanford Encyclopedia of Philosophy, Edward N.
Zalta (ed.), https://plato.stanford.edu/archives/fall2008/entries/philo-larissa/.
Burnyeat, M. 1980. “Can the Sceptic Live His Scepticism?” In M. Schofield, M. Burn-
yeat, and J. Barnes (eds.), Doubt and Dogmatism: Essays in Hellenistic Epistemology,
20–53. Oxford: Clarendon Press.
Burnyeat, M. 1984. “The Sceptic in His Place and Time.” In R. Rorty, J. Schneewind, and
Q. Skinner (eds.), Philosophy in History, 225–254. Cambridge: Cambridge University
Press.
Castagnoli, L. “Aenesidemus.” In D. Machuca and B. Reed (eds.), Skepticism: From An-
tiquity to the Present, 67–80. London: Bloomsbury.
Charron, P. 1697 [1603]. Of Wisdom: The Second and Third Books, tr. by G. Stanhope.
London: Gillyflower et. al.
Cicero. 2006. On Academic Scepticism, tr. by C. Brittain. Indianapolis, IN: Hackett.
Clifford, W.K. 1877. “The Ethics of Belief,” Contemporary Review 29: 289–309.
Coliva, A. 2018. “Moore and Mooreanism.” In D. Machuca and B. Reed (eds.), S kepticism:
From Antiquity to the Present, 467–480. London: Bloomsbury.
36 For helpful comments, I am grateful to the audience at the Orange Beach Epistemology
Workshop, the audience at the Collège de France workshop on skepticism, Pascal Engel,
Jacques Vollet, Jean-Baptiste Guillon, and especially Jennifer Lackey. I am also grateful to
Kevin McCain and Ted Poston for their generosity and patience.
Dennett, D. 1984. Elbow Room: The Varieties of Free Will Worth Wanting. Cambridge,
MA: MIT Press.
Descartes, R. 1985. The Philosophical Writings of Descartes, vol. I, tr. by J. Cottingham, R.
Stoothoff, and D. Murdoch. Cambridge: Cambridge University Press.
Fogelin, R. 1994. Pyrrhonian Reflections on Knowledge and Justification. Oxford: Oxford
University Press.
Frances, B. 2005. Scepticism Comes Alive. Oxford: Clarendon Press.
Frede, M. 1979. “The Skeptic’s Beliefs.” In his Essays in Ancient Philosophy, 179–200. Ox-
ford: Oxford University Press.
Frede, M. 1984. “The Sceptic’s Two Kinds of Assent and the Question of the Possibility
of Knowledge.” In R. Rorty, J. Schneewind, and Q. Skinner (eds.), Philosophy in His-
tory, 255–278. Cambridge: Cambridge University Press.
Fumerton, R. 1995. Metaepistemology and Skepticism. Lanham, MD: Rowman &
Littlefield.
Fumerton, R. 2016. “Rising above the Animals: The Search for Intellectual Assurance.”
In M.A. Fernandez Vargas (ed.), Performance Epistemology: Foundations and Appli-
cations, 151–166. Oxford: Oxford University Press.
Greco, D. 2012. “The Impossibility of Skepticism,” Philosophical Review 121: 317–358.
Harman, G. 1973. Thought. Princeton: Princeton University Press.
Hawthorne, J. and J. Stanley. 2008. “Knowledge and Action,” Journal of Philosophy 105:
571–590.
Hume, D. 1975. Enquiry Concerning Human Understanding, 3rd ed. Oxford: Clarendon
Press.
Ioppolo, A.M. “Arcesilaus.” In D. Machuca and B. Reed (eds.), Skepticism: From Antiq-
uity to the Present, 36–50. London: Bloomsbury.
Kane, R. 1996. The Significance of Free Will. Oxford: Oxford University Press.
Klein, P. 2002. “Skepticism.” In P. Moser (ed.), The Oxford Handbook of Epistemology,
336–361. Oxford: Oxford University Press.
Kripke, S. 2011. “Two Paradoxes of Knowledge.” In his Philosophical Troubles: Collected
Papers, vol. 1, 27–51. Oxford: Oxford University Press.
Lammenranta, M. 2018. “Neo-Pyrrhonism.” In D. Machuca and B. Reed (eds.), Skepticism:
From Antiquity to the Present, 565–580. London: Bloomsbury.
Laursen, J.C. 2018. “Bayle.” In D. Machuca and B. Reed (eds.), Skepticism: From Antiquity
to the Present, 355–368. London: Bloomsbury.
Long, A.A., and D. Sedley (eds). 1987. The Hellenistic Philosophers, vol. 1. Cambridge:
Cambridge University Press.
Lycan, W. 2001. “Moore against the New Skeptics,” Philosophical Studies 103: 35–53.
Maia Neto, J. 2014. Academic Skepticism in Seventeenth-Century French Philosophy: The
Charronian Legacy 1601–1662. Dordrecht: Springer.
80 Reed
Moore, G.E. 1959a. “Four Forms of Scepticism.” In his Philosophical Papers, 196–226.
London: George Allen & Unwin.
Moore, G.E. 1959b. “Proof of an External World.” In his Philosophical Papers, 127–150.
London: George Allen & Unwin.
Nadler, S. 1999. Spinoza: A Life. Cambridge: Cambridge University Press.
Nozick, R. 1981. Philosophical Explanations. Cambridge, MA: Harvard University Press.
Paganini, G. 2018. “Pierre Charron.” In D. Machuca and B. Reed (eds.), Skepticism: From
Antiquity to the Present, 247–259. London: Bloomsbury.
Pascal, B. 1999. Pensées, tr. By A. Levi and H. Levi. Oxford: Oxford World’s Classics.
Pereboom, D. 2001. Living without Free Will. Cambridge: Cambridge University Press.
Perin, C. 2010. The Demands of Reason: An Essay on Pyrrhonian Scepticism. Oxford:
Oxford University Press.
Perin, C. 2018. “Pyrrho and Timon.” In D. Machuca and B. Reed (eds.), Skepticism: From
Reed, B. 2002. “The Stoics’ Account of the Cognitive Impression,” Oxford Studies in An-
cient Philosophy 23: 147–180.
Reed, B. 2007. “The Long Road to Skepticism,” Journal of Philosophy 104: 236–262.
Reed, B. 2013. “Fallibilism, Epistemic Possibility, and Epistemic Agency,” Philosophical
Issues(Epistemic Agency) 23: 40–69.
Reed, B. Unpublished. “Doubt.”
Reid, T. 1872 [1785]. Essays on the Intellectual Powers of Man. Collected in The Works of
Thomas Reid, Vol. 1, ed. by W. Hamilton. Edinburgh: Maclachlan and Stewart.
Sextus Empiricus. 2000. Outlines of Scepticism, ed. By Annas, J. and J. Barnes. Cam-
bridge: Cambridge University Press.
Sosa, E. 2009. Reflective Knowledge (Apt Belief and Reflective Knowledge, Vol. ii). Oxford:
Clarendon Press.
Strawson, G. 1986. Freedom and Belief. Oxford: Oxford University Press.
Striker, G. 1997. “Academics Fighting Academics.” In B. Inwood and J. Mansfeld (eds.),
Assent and Argument: Studies in Cicero’s Academic Books, 257–276. Leiden: Brill.
Tarrant, H. 2018. “Philo of Larissa.” In D. Machuca and B. Reed (eds.), Skepticism: From
Vogt, K.M. 2010. “Scepticism and Action.” In R. Bett (ed.), The Cambridge Companion to
Ancient Scepticism, 165–180. Cambridge: Cambridge University Press.
Williams, M. 1996. Unnatural Doubts. Princeton, NJ: Princeton University Press.
Zweig, S. 2015 [1960]. Montaigne, tr. by W. Stone. London: Pushkin Press.
Part ii
New Skeptical Challenges
∵
Chapter 6
Disagreement Skepticism and the Rationality

of Religious Belief
Jonathan Matheson
1 Introduction
The Equal Weight View is a view about the epistemic significance of disagree-
ment that is thought to have significant skeptical consequences. In this paper
I do two things: (i) apply the Equal Weight View to cases of religious disagree-
ment, and (ii) evaluate some consequences of that application for the rational-
ity of religious beliefs. With regard to (i), I argue that the Equal Weight View
implies that awareness of the current state of disagreement over religious
propositions, such as God exists or God doesn’t exist, gives us a defeater for any
non-skeptical attitude toward such propositions. With regard to (ii), I examine
what this entails about the rationality of religious beliefs. In particular, I exam-
ine how troubling the theist should find these consequences, and argue that
they are not as troubling as is typically thought.
2 The Epistemology of Disagreement
The epistemology of disagreement concerns how discovering that another

person disagrees with you about some claim should affect what you believe
about that claim. Discussion of the epistemic significance of disagreement
typically takes place in two ways: examining peer disagreements and examin-
ing more general everyday disagreements. While there are a number of distinct
conceptions of epistemic peers in the literature, the central feature of all of
these accounts is that epistemic peers are a kind of epistemic equal. Epistemic
peers about p are in an equally good epistemic position with respect to p – they
are equally likely to be correct about p.
Three questions lie at the heart of the debate concerning the epistemic sig-
nificance of peer disagreement, and the various views on the epistemic signifi-
cance of disagreement can be distinguished in terms of how they answer each
of these questions.1
1 This follows Matheson (2015a) and (2015d).

84 Matheson
(Q1) Does evidence of a peer disagreement give me a defeater for my

belief?
(Q2) If yes to (Q1), then how strong is this defeating reason?
(Q3) If yes to (Q1), then under what conditions is this defeating reason
itself defeated?
2.1 The Equal Weight View

The Equal Weight View gives a “yes” answer to (Q1). It claims that in gaining
evidence that a peer disagrees with you, you gain evidence that you have made
a mistake. Finding out that someone who is in an equally good epistemic po-
sition on the matter disagrees with you about that matter gives you a reason
to think that there is a deficiency in your evidence or your processing of it, or
more generally, it gives you a reason to believe that you have made a mistake.
While the fact that you disagree with your peer is also a reason for your peer
to believe that he has made a mistake (and for you to believe that they have
made a mistake), this does not prevent this information about your peer from
being a defeater for your belief. One of you has made a mistake, and the expla-
nation that it is you is no worse (or better) than that your peer has made the
mistake.
With regards to (Q2), the Equal Weight View claims that this defeating rea-
son is quite strong. Since peers are in an equally good epistemic position on the
matter, the evidence that you have that you have made a mistake is just as strong
as the evidence you have that you are correct. So, according to the Equal Weight
View, the evidence that you have made a mistake and the evidence that your
peer has made a mistake is equally strong, and this results in each of the peer
opinions being given equal weight. While there are a number of ways that
two opinions may be given “equal weight,” according to the Equal Weight View,
the significance of giving both peer opinions equal weight is that each peer
should “split the difference” and adopt the mean doxastic attitude between
the two originals. That is, if you believe p, and gain evidence that your peer dis-
believes p, then this evidence calls for you to split the difference with your peer
and suspend judgment about p. On a more fine-grained account of doxastic at-
titudes, if you have a 0.8 degree of belief that p, and gain evidence that your peer
has a 0.4 degree of belief that p, then this evidence calls for you to adopt a 0.6 de-
gree of belief that p.
What about (Q3)? According to the Equal Weight View, the call to split the
difference coming from the discovery of a peer disagreement is only defeated
by considerations independent from the disagreement itself. Along these lines,
David Christensen (2009: 758) has given the following principle:
Disagreement Skepticism and Rationality of Religious Belief 85
INDEPENDENCE: In evaluating the epistemic credentials of another’s

expressed belief about P, in order to determine how (or whether) to
modify my own belief about P, I should do so in a way that doesn’t rely
on the reasoning behind my initial belief that P.2
While INDEPENDENCE is focused on establishing the epistemic position of

potential peers, the Equal Weight View extends this reasoning to apply to any
potential defeater-defeaters. So, according to the Equal Weight View, while it
is possible to possess a defeater that defeats the call to split the difference, such
a defeater must come from considerations outside of the disagreement itself.
As such, considerations such as that I disagree, that I had reasoned differently,
that my intuitions were what they were, and the shared first-order evidence3 in
fact supports my belief, all do not qualify as defeaters to the call to split the dif-
ference. On the other hand, considerations such as that my peer is drunk, my
peer is biased on this topic, my peer has been taking medication, or that the
majority of experts agree with me, all could qualify as defeaters to the call to
split the difference (at least if I am justified in believing that they obtain with
respect to the peer disagreement at hand).
2.2 Moving beyond Peer Disagreements

While much of the literature on the epistemic significance of disagreement
focuses on disagreement between epistemic peers, a complete account of
the epistemic significance of disagreement must move beyond this and say
something about the everyday disagreements that we discover.4 Regarding ev-
eryday disagreements, the Equal Weight View claims that you should weigh
each opinion on the matter in accordance with what your evidence supports
about that subject’s epistemic position on the matter. The greater you are jus-
tified in believing one’s epistemic position on the matter, the more you are
to weigh her opinion on the matter. Here, however, a caveat is needed. Two
agreeing and independent opinions about p will carry more weight than two
agreeing opinions about p that are not independent (all else being equal). For
instance, evidence of two individuals who both believe incompatibilism but
who went to the same school, took the same classes, had the same advisor, etc.
2 Principles similar to Independence can be found in Elga (2007) and at least implicitly in
Frances (2010).
3 First-order evidence for p is evidence that directly pertains to p. It is to be contrasted with
higher-order evidence regarding p which is evidence directly pertaining to the first-order
evidence for p.
4 For more on such a move, see Matheson (2015b), Matheson (2014), Frances (2010), and King
(2011).
86 Matheson
will give less support for that metaphysical claim than evidence of two indi-
viduals who each believe without such a shared history. The less independent
two individual’s opinions are, the more likely they are to be influenced by the
same non-epistemic factors. So, the less independent any two opinions are, the
greater the likelihood of an alternative, non-epistemic, explanation for their
agreed opinion, and so the combined evidential weight of their opinions is less
than it would otherwise be.
Regarding many controversial propositions in science, politics, and philoso-
phy, we are not justified in believing of any non-skeptical doxastic attitude5
that our higher-order evidence best supports adopting it toward the disputed
proposition.6 While it may be plausible that some non-skeptical attitude is
often better supported by the evidence than the others, we are often in the
dark regarding which attitude that is. It is very difficult to gauge the epistemic
position of various people, it is difficult to find out what attitude a number
of people have toward a given proposition, and it is difficult to assess the in-
dependence of various individual’s attitudes on a topic. The precise state of
disagreement on such controversial issues is simply too difficult to determine.
However, regarding propositions that we know enjoy an overwhelming con-
sensus (i.e. that the Earth is not flat, that 2+2=4, that Obama was President in
2015) these difficulties can be overcome – a non-skeptical “winning” attitude
can reasonably be declared. Regarding such claims, the difficulty in determin-
ing the precise state of the disagreement does not prevent us from justifiably
declaring a non-skeptical “winning” attitude. It is clear that belief is the “win-
ning” attitude regarding whether Obama was the President in 2015 regardless
of how the small details work out. However, regarding claims that we know to
be more deeply contentious (i.e. that free will and determinism are compat-
ible, that God exists, that property taxes should be raised) these difficulties
prevent us from having an undefeated non-skeptical attitude toward them –
here the race is too close to call. Regarding such deeply contentious claims,
such details will decide the “winning” attitude and thus cannot properly be
ignored in assessing the “winning” attitude.
There is an analogy here with political elections.7 While there are difficul-
ties in calling an election winner, sometimes we have enough information to
make such a call without having all the relevant information. Such cases paral-
lel our ability to declare “winning” non-skeptical attitudes regarding relatively
uncontroversial propositions. In these cases even incomplete information is
5 By “any non-skeptical doxastic attitude” I mean any doxastic attitude other than withholding
judgment.
6 This is argued at length in Carey and Matheson (2013) and Matheson (2015a).
7 See Carey and Matheson (2013) and Matheson (2015) for more on this analogy.
sufficient to declare victory. Other times, the races are tight and are too close
to call given the information at the time. Such cases parallel our inability to
declare a “winning” non-skeptical attitude regarding deeply contentious prop-
ositions. While we may know who the contenders are (what attitudes, what
propositions), and we may be able to rule out some possibilities (like distant
third-party candidates), we aren’t in an epistemic position to declare a “win-
ning” non-skeptical attitude at this time. When the controversy is deep, the
details will matter and we simply lack the right kind of access to those details.
Given our awareness of the extensive disagreement surrounding a claim, any
justification that we had for a non-skeptical attitude toward such a proposition
is thus fully defeated.
My task here is not to defend the Equal Weight View, such defenses can be
found elsewhere.8 Rather, my task here is to examine the consequences of the
Equal Weight View for the rationality of religious belief. With this understand-
ing of the Equal Weight View in hand, we turn now to this task.
3 Disagreement & Religious Belief
The Equal Weight View has it that any justification we have for a non-skeptical
attitude toward a deeply contentious proposition is fully defeated by our
awareness of the controversy surrounding it. The Equal Weight View thus has
consequences for contentious claims in science, politics, philosophy, and re-
ligion, though my focus in what follows will be on religious disagreement. By
“religious disagreement” I mean simply disagreement concerning a religious
claim. While there are a number of such disagreements, I will focus on the
disagreement concerning God’s existence as a paradigm case of religious
disagreement.
God’s existence is controversial. There are intelligent, informed, open-
minded individuals (those in a good epistemic position on the matter, or at
least as good as it gets for creatures like us) that believe that God exists and
there are intelligent, informed, open-minded individuals that disbelieve that
God exists. Regarding God’s existence, the disagreement is so extensive that
the precise state of it cannot be determined; the higher-order evidence does
not clearly favor belief or disbelief. So, we are not justified in believing of either
belief or disbelief that it is best supported by our higher-order evidence. Like
other deeply controversial claims it appears that the Equal Weight View has it
that our awareness of the extensive disagreement surrounding God’s existence
8 See Elga (2007), Christensen (2007), Feldman (2006), and Matheson (2015a).
88 Matheson
d efeats any justification we have for a non-skeptical attitude (belief or disbe-

lief) toward the claim that God exists.9
3.1 Religious Disagreement: a Special Case?

Now it might be thought that religious disagreement is special – that religious
disagreement is importantly unlike other disagreements in science, politics,
and philosophy more generally, and that such a difference renders religious
beliefs immune from the defeating effects of discovered disagreement. Some
of the more plausible candidate differences are (i) that religious disagreement
is in some sense fundamental, (ii) that religious experience is importantly un-
like other kinds of evidence, and (iii) that there is an important asymmetry in
epistemic position between believers and non-believers.
While each of the alleged distinctions between religious disagreements and
other disagreements deserve more attention than I can give here, there is rea-
son to worry that these distinctions are up to the proposed task. In brief, the
problem is this: the claim that any of these distinctions is both legitimate and
epistemically relevant is itself deeply contentious. Further, the contentious-
ness of such claims prevents them from being used as reasons to treat religious
disagreement as special. To see this, consider the following claim:
C: eligious disagreement is fundamental, and in virtue of that, our

R
awareness of the disagreement does not give us a (undefeated)
defeater for our justification for adopting a non-skeptical attitude
about God’s existence.
The problem is that C is itself a deeply controversial claim; there is significant

disagreement concerning C. One may worry that not many people have con-
sidered C. However, we have good reason to believe that were many to do so
there would be significant disagreement. That is, we have good reason to think
that were more people to consider C, they would be deeply divided over its
9 Some may maintain the epistemic election regarding God’s existence is not so close. After all,
billions of people alive today believe that God exists, and historically, theism has been quite
dominant at least in the west. If someone is justified in believing that belief, say, is the win-
ning attitude, then the Equal Weight View will not have any skeptical consequences for that
person. However, others see atheism as trending, with theism on the decline, and point to the
large proportion of academics who are atheists. They also stress the lack of independence of
may of those who have come to theistic conclusions. I take it that the outcome of such an
epistemic election is itself significantly controversial. So, given the Equal Weight View, those
aware of the controversy are not justified in believing that belief wins the epistemic election.
truth. Mere novelty cannot save one from the defeat of disagreement.10 So, the
Equal Weight View has it that our awareness of the controversy concerning C
provides us with a defeater for any justification we have for believing C. Given
the Equal Weight View and our awareness of the disagreement surrounding C,
we are not justified in believing C. This is not to “beg the question” against C,
since C is consistent with such consequences in general. In fact, C is offered as
a reason to think that religious beliefs do not suffer a similar fate, granting that
other controversial beliefs do. Remember that here we are assuming that the
Equal Weight View is correct, and simply examining the claim that religious
beliefs are to be treated as a special case. So, the defeating effects of disagree-
ment cannot themselves be defeated by an appeal to C, since any justification
for believing C is defeated by the disagreement surrounding it (at least given
the truth of the Equal Weight View).
Further, C does not have the resources to save itself from this fate. C is not
able to defeat the defeating effects of disagreement concerning C. Even if C is
true and enjoys prima facie justification, this is of no help to defending C from
the threat of disagreement. After all, C is not itself a fundamental proposition
(on any plausible account of what counts as a fundamental proposition). C
certainly does not play a foundational role in one’s worldview. Disagreements
about C are not fundamental disagreements (on any plausible account of what
counts as a fundamental disagreement), so given the Equal Weight View, one’s
awareness of the disagreements surrounding C provides a defeater for any jus-
tification one has for believing C. So, even if C is true and prima facie justified,
the disagreements surrounding C prevent us from being able to use C to ward
off the disagreement defeater for religious beliefs. If our justification for adopt-
ing a non-skeptical attitude toward C is itself defeated and we should suspend
judgment about C, then we cannot then use C as a reason to believe that these
same skeptical consequences apply to our religious beliefs. Put differently, our
awareness of the controversy surrounding C has it that we are justified in sus-
pending judgment about whether C is true (and not justified in adopting any
competitor doxastic attitude toward that proposition). However, if we are jus-
tified in suspending judgment about C, then appealing to C cannot defeat the
defeater for our justification for our religious beliefs. So, C cannot simply save
the justification of our religious beliefs.
10 For instance, suppose that I construct a novel argument for incompatibilism. While in my
office, I know of no one who rejects this argument. That said, I know how these things
typically go, so I know that were people to think about the argument, many will disagree.
Such evidence of “nearby” disagreement has a defeating effect. For more on this point, see
Carey and Matheson (2013).
90 Matheson
Further, similar considerations apply to the other plausible candidate dis-

tinctions between religious disagreement and more general disagreement.
Whether religious experience is a special type of evidence, and whether there
is an important asymmetry in epistemic position between believers and non-
believers, are themselves deeply controversial claims.11 Further, those claims do
not provide the resources to save themselves from the skeptical consequences
of disagreement. We do not have religious experiences about the epistemic
weight of religious experience, and there is not an asymmetry of epistemic po-
sition between believers and non-believers regarding assessments of epistemic
position, even if there is such an asymmetry regarding God’s existence. So even
if religious disagreements are importantly different from more ordinary dis-
agreements, the extant controversy surrounding that issue, or better our aware-
ness of it, prevents us from using such claims as reasons to believe that our
awareness of religious disagreement does not give us a (undefeated) defeater
for any non-skeptical attitude toward those claims. So, given the Equal Weight
View, none of our initially plausible candidates give us a good reason to treat
religious disagreement differently.
In addition, there is good reason to worry that some other candidate we
have yet to examine is up to the task. The reason is that for any such candidate
the same recipe of response will apply. Given the nature of such philosophical
claims, any new candidate will likely itself be controversial, and thus given the
Equal Weight View, those aware of such an impending controversy will not
be justified in believing the relevant claim absent some reason to treat it as
special, and it is implausible that the candidate claim will also have the re-
sources to save itself from the defeating effects of disagreement. So, given that
our three plausible candidate claims fail to give us a reason to treat religious
disagreement differently, and there is good reason to be pessimistic about the
existence of some other candidate claim being successful, we should think that
the Equal Weight View applies to religious disagreements in the same way that
it applies to disagreements more generally.
3.2 An Early Worry

Now it might be thought that the above argument is too quick. The above ar-
gument shows that given the Equal Weight View, aware individuals are not
justified in believing the candidate claims, but it might be thought that the
mere truth of any of these claims is sufficient to treat religious disagreement
11 Recall that even if many people have not actually considered these claims, it is reasonable
to believe that were they to do so, we would find their opinions greatly divided on the
matter. Such “nearby” disagreement can take the place of actual disagreement if needed.
as special. After all, we needn’t be justified in believing a number of epistemic

principles in order for their truth to have implications for the status of our
beliefs.
While there is some pull to this worry, it ultimately fails. For even if the can-
didate claims are true, the problem noted above wasn’t simply that aware in-
dividuals are not justified in believing those claims (given the Equal Weight
View), but that aware individuals are justified in suspending judgment about
the truth of those candidate claims. This is an important difference. Let’s con-
sider why in more detail by considering an analogy. Suppose that Smith now
believes that he is sitting at a computer on the basis of his perceptual faculties.
Smith then hears about the skeptical brain-in-vat scenario and learns that it
is possible that he is being deceived right now. Suppose further that it is an
epistemic truth that the mere possibility that one is being deceived is not a
defeater for one’s justification for their perceptual beliefs. So, what Smith has
learned is not in fact a defeater for his perceptual belief. However, suppose that
Smith hears from two (and only two) equally reliable epistemologists, known
by Smith to be such. One epistemologist tells Smith that the mere possibility of
deception is not a defeater, but the other epistemologist tells Smith that even
such mere possibilities are defeaters. Perhaps other details are needed, but fill
them in as necessary to make it the case that Smith is justified in suspending
judgment about whether mere possibilities are defeaters (even though they in
fact are not). In such a situation where Smith is justified in suspending judg-
ment about whether he has a defeater for his perceptual beliefs he is also jus-
tified in suspending judgment about whether there is a computer in front of
him. Since Smith is justified in suspending about whether his computer belief
is defeated, Smith is not justified in either believing or disbelieving the target
proposition about there being a computer in front of me (only suspension is
justified for him). Such a justified suspension of judgment effectively under-
mines the justification Smith’s perceptual evidence gives to his perceptual be-
lief. So, a justified suspension of judgment about whether a belief is defeated
amounts to a full defeater for that belief (whether or not the candidate de-
feater is true). Applied to the case of religious disagreement, if we are justified
in suspending judgment about whether our religious beliefs are defeated (in
virtue of being justified in suspending judgment about whether the candidate
claims are true), then we are not justified in believing our religious beliefs (and
this is true whether or not the candidate claims are true).12
12 One might wonder how many people this skeptical argument applies to. After all, since
it is not the truth of the relevant claim, the effect of the defeater will not be universal.
The question of how many people are aware of the relevant controversies is an empirical
question, and I won’t suggest an answer here.
92 Matheson
3.3 A Second Worry

Another worry at this point may be that the same considerations that doomed
C apply equally well to the Equal Weight View. After all, like C, the Equal Weight
View is also a deeply contentious view. So, according to the Equal Weight View,
our awareness of this controversy constitutes a defeater for any justification
we have for believing the Equal Weight View. For these reasons, many have
charged that the Equal Weight View is self-defeating.13 While the principle aim
of this paper is not to evaluate the Equal Weight View, one might think that the
above concern must be addressed given that similar reasoning is being given
to dismiss the effectiveness of C.
The first thing to note is that the disagreement defeater does apply to both
the Equal Weight View and C. Both the Equal Weight View and C are signifi-
cantly controversial, so given the truth of the Equal Weight View, those aware
of this fact have acquired a full defeater for their justification for these propo-
sitions. So, the philosophically aware individual is not on balance justified in
believing either the Equal Weight View or C.
The second thing to note is that this fact does not show that the Equal
Weight View or C is false. Just as above we needed to consider whether C was
true independently of whether we are justified in believing it, similarly, the fact
that one is not on balance justified in believing the Equal Weight View does
not show that the Equal Weight View is false. So, since we are supposing that
the Equal Weight View is true, and examining what follows, such self-defeating
concerns do no inhibit this project. Even if we are not on balance justified in
believing the Equal Weight View, we can still examine what consequences
such a view would have.
3.4 A Third Worry

At this point, one may worry that the project of this paper is a mere philo-
sophical exercise with no real import. After all, if we are not on balance justi-
fied in believing the Equal Weight View, can’t we simply ignore it and all of it’s
alleged consequences? In short, no we cannot. As we have seen above, if we
are justified in suspending judgment as to whether we have a full defeater for
the justification of one of our beliefs, then we thereby have a full defeater
for the justification for that belief. Recall the case of Smith and the computer.
So, if the end result of the self-defeating concern for the Equal Weight View is
that we are on balance justified in suspending judgment about its truth, the
Equal Weight View would nevertheless retain, rather than lose, its defeating
13 See Elga (2011), O’Connor (1999), Plantinga (1999), Taliaferro (2009), and Weatherson
(2014). For responses, see Bogardus (2009), Christensen (2009), Elga (2011), Frances (2010),
Graves (2013), Kornblith (2013), Littlejohn (2013), and Matheson (2015a), (2015c).
consequences. So, like with C, a justified suspension of judgment regarding

the Equal Weight View itself will only have us remain in its defeating grasp.
This discussion reveals what is necessary to escape from the defeating ef-
fects of the Equal Weight View – one must be on balance justified in believ-
ing that the Equal Weight View is false or one must be on balance justified
in believing that we have a defeater-defeater to neutralize the effects of the
Equal Weight View. However, given the supposed truth of the Equal Weight
View both such routes of escape are blocked since they require appealing to
claims that are themselves recognized by us to be suitably controversial, and
thus according to the Equal Weight View we are not on balance justified in be-
lieving them. As such, we are left with the defeating effects of the Equal Weight
View. We turn now to getting clear on what those effects are and what to make
of them.
4 Some Consequences
Having clarified the Equal Weight View and its application to religious beliefs,
my primary task here is to identify and evaluate the consequences of the view
for religious belief. This is an important task since too often the Equal Weight
View has been thought to have consequences for religious belief that it simply
does not have. One thing we care about is what we should believe. We particu-
larly care about this question with regard to matters that are of deep impor-
tance to us – like religion. It has been claimed that the Equal Weight View has it
that we should not have religious beliefs; that according to such views religious
beliefs are irrational.
Thus, Richard Feldman (2006a: 213) claims,
[The Equal Weight View] is also a skeptical view, in the limited sense that
it denies the existence of reasonable beliefs in a significant range of cas-
es. This may seem to be a distressing conclusion.
Feldman reinforces this idea (2006b: 217):
[O]n many [issues in philosophy, religion, science, and public policy]

about which you have a belief, informed and intelligent people dis-
agree with you. The question I will raise concerns the reasonableness
of maintaining your point of view in the light of such disagreements.
My conclusion will be that, more often than we might have thought,
suspension of judgment is the epistemically proper attitude. It follows
94 Matheson
that in such cases we lack reasonable belief and so, at least on standard
conceptions, knowledge. This is a kind of contingent real-world skepti-
cism that has not received the attention it deserves.
David Christensen echoes this sentiment (2009: 757–758):
The most obvious motivation for [opposing] views on disagreement

flows from the degree of skepticism that [the Equal Weight View] would
seem to entail. There must be something wrong, the thought goes, with
a view that would counsel such widespread withholding of belief. If you
have an opinion on, for example, compatibilism about free will, scientific
realism, or contextualism about knowledge, you must be aware that there
are very intelligent and well-informed people on the other side. Yet many
are quite averse to thinking that they should be agnostic about all such
matters. The aversion may be even stronger when we focus on our opin-
ions about politics, economics, or religion.
4.1 Broad Notions of Rationality

To evaluate such a charge we must distinguish various senses in which a belief
may or may not be rational. Here is one:
R1: S’s belief B is rational iff B is not ridiculous.
The sense of rationality in R1 is familiar in everyday speech. To say that a be-

lief is rational in this sense is to say that a generally reasonable and informed
individual could have such a belief; that the belief in question isn’t outside
the realm of plausibility. For instance, when thinking about who will win an
upcoming sporting event, many competitor claims will be seen as rational in
this sense. When thinking about next year’s Super Bowl it is rational (in this
sense) to think that the Patriots will win, and it is rational (in this sense) to
think that the Seahawks will win. In this sense of rationality, it is clear that
reasonable people can disagree. On many questions there are numerous com-
petitor answers that all enjoy some plausibility. Further, this sense of rational-
ity is not vacuous since it still leaves room for some beliefs to be irrational. For
instance, sticking with the Super Bowl example, it is nevertheless irrational
(even in this sense) to believe that both the Patriots and Seahawks will win,
or even irrational (even in this sense) to believe that the Jaguars will have an
undefeated season leading up to a Super Bowl win.
It should also be clear that religious belief can be rational in this sense. In
fact, the particular kind of controversy that surrounds God’s existence entails
that religious belief is rational in this sense. There is significant controversy
surrounding God’s existence. The extensive controversy requires that there are
numerous individuals who are intelligent, open-minded, and informed on ei-
ther side of the issue. So, regarding the claim that God exists, both belief and
disbelief are real contenders; each of these doxastic attitudes has a significant
number of intelligent, informed, open-minded individuals who have adopted
it toward the claim that God exists, and this is enough to satisfy R1 – belief (and
disbelief) isn’t a ridiculous response to the evidence.
However, other senses of rationality go well beyond R1. Consider the
following:
R2: S’s belief that p is rational iff p is what S should believe.
R2 is much stronger than R1 in that many non-ridiculous beliefs are still not
ones that we should believe. Even if it is not ridiculous to believe that the Patri-
ots will win next year’s Super Bowl, it does not seem that this is a proposition
that we should believe. There is nothing wrong with people who fail to have
this belief (even once they have considered the proposition). If anything, there
is something wrong with people who do have this belief. Similarly, believing
that there is an even number of geese in Canada, while not ridiculous, is not
a belief that you should have. So, it takes more to satisfy R2 than it does to
satisfy R1.
What does the Equal Weight View tell us about religious beliefs with re-
spect to this sense of rationality? Does the Equal Weight View claim that we
should not have religious beliefs? No. The Equal Weight View simply does not
concern what we all-things-considered should believe. The Equal Weight View
simply does not address any moral, prudential, or even religious reasons that
there may be to believe anything, nor does it tell us how, if at all, reasons of all
these varieties are to be weighed and sorted out to give a final verdict about
what one all-things-considered should believe. The Equal Weight View is si-
lent on all these matters. So the inference from the Equal Weight View is true
and applies to religious beliefs to we should not have religious beliefs is too hasty.
The Equal Weight View is concerned with but one kind of reason that we
may have for religious belief – epistemic reasons – reason that we have to
believe something from the perspective dedicated to believing truths and
not believing falsehoods. This matters since we can have reasons to believe
something that are not epistemic reasons. To make this clear, consider Pascal’s
96 Matheson
Wager.14 Pascal has famously argued that belief that God exists is rational since
believing that God exists is a good bet to make. Betting (with belief) on God’s
existence has the potential for an infinite payoff with only negligible negative
consequences in the event that you are mistaken. In contrast, not making that
bet leaves little room for reward and has the potential for vast negative con-
sequences. While an evaluation of Pascal’s argument is outside the scope of
this paper, it should be clear that the Equal Weight View does not provide an
indictment of this argument – the Equal Weight View does not show that we
lack such reasons for believing that God exists. Pascal’s argument for believ-
ing that God exists concerns a kind of reasons for belief that the Equal Weight
View simply does not address – pragmatic reasons (reasons concerning what
is good and beneficial).
Now there is significant disagreement over the merits of Pascal’s Wager, and
thus the Equal Weight View plausibly has it that we are not epistemically justi-
fied in believing that Pascal’s Wager is a good argument. Nevertheless, the exis-
tence of pragmatic reasons (such as the reasons Pascal maintains there are for
believing that God exists) does not depend upon our being epistemically justi-
fied in their existence – pragmatic reasons are what we might call “objective
reasons.” They exist, when they do, independent of our awareness of them and
they are not prone to defeat like epistemic reasons. So, if Pascal’s reasoning is
correct (whether we are epistemically justified in believing it is correct or not),
there would be good reason to believe that God exists – reasons untouched by
the epistemic significance of disagreement. Whether believing is a good bet
is independent of whether we are justified in believing it is a good bet. So, the
skeptical consequences of the Equal Weight View fail to defeat or even address
any such pragmatic reasons to believe that God exists.
4.2 Purely Epistemic Notions of Rationality

These considerations lead us to a third conception of rationality:
R3: S’s belief that p is rational iff from the epistemic perspective, p is what
S should believe.
As we have seen, the Equal Weight View does not concern just any type of reason
to believe a proposition. The relevant sorts of reasons are epistemic reasons –
reasons to believe that the proposition is true (and not merely beneficial to
believe). Put differently, at issue are the reasons one has from the epistemic
perspective. The epistemic perspective is the perspective one occupies with
14 See Pascal (1670) for a more detailed explanation and defense of Pascal’s Wager.
the twin goals of believing truths and not believing falsehoods.15,16 These twin
goals each keep the other in check, preventing the unhealthy epistemic lives
of both the overly cautious believer and the indiscriminate believer. The Equal
Weight View concerns only epistemic reasons, and as we have seen, these are
but one kind of reasons for belief.
Nevertheless, lacking sufficient epistemic reasons for religious belief may
still seem like a significant negative consequence. After all, we care a lot about
epistemic reasons. They are nice things to have, and the Equal Weight View
may seem to imply that our epistemic reasons are not good enough to support
a non-skeptical attitude toward religious propositions. However, there is good
reason to believe that we still have not honed in on the particular kind of rea-
sons targeted by the Equal Weight View.
The idea that we should abandon controversial beliefs is challenged by
psychological evidence about group inquiry.17 In brief, there is evidence that
groups with members who genuinely disagree about a proposition do better
with respect to determining whether that proposition is true. This evidence
gives us a reason to not abandon our beliefs in the face of disagreement, and it
is a reason coming from the epistemic perspective. In contrast to Pascal-type
reasons, the reasons here are entirely about getting true beliefs and avoiding
false ones. The psychological data indicates that both:
A. Groups are better at discovering the truth when there is dissent and
debate within the group.
B. Groups are better at discovering the truth when dissenting parties
genuinely hold their dissenting beliefs in the debate.
The evidence suggests that groups with genuine dissent do better at avoid-
ing both belief bias and confirmation bias.18 Genuine diversity of opinion
amongst members of a group has been found to help in both uncovering a
15 The representativeness of this view of the epistemological perspective can be seen in

David (2001). David provides an extensive list of quotations to this end including episte-
mologists such as Alston, BonJour, Chisholm, Descartes, Foley, Goldman, Lehrer, Moser,
Plantinga, and Sosa.
16 This is not to say that truth is the only thing of value from the epistemic perspective, only
that everything that has value from the epistemic perspective has it due to its relation to
truth. For a nice discussion of this and related issues, see Pritchard (2014).
17 See Moshman and Geil (1998), Sunstein (2002), Mercier (2012), Dawson et al. (2002), and
Schulz-Hardt et al. (2002). See Dunn (forthcoming) for bringing these considerations to
bear on the disagreement debate.
18 See Moshman and Geil (1998), and Sunstein (2002).
98 Matheson
more balanced body of evidence as well as in properly evaluating that body of

evidence.19 There being individuals who genuinely believe these competing
views provides better checks and balances and improves our epistemic situa-
tion on the matter. That is, it is plausible that there are even reasons pertaining
to epistemic goods to maintain a belief in the face of controversy.20
For instance, there is good epistemic reason for physicists to believe their
competing views of the quantum world. In disparate views having genuine de-
fenders the resulting inquiry is mostly likely to lead us to the truth of the matter.
When competitor views each have genuine defenders, we gain a better body of
evidence and are more likely to evaluate that body of evidence correctly (we
are less prone to blindspots and biases). These epistemic payoffs are not just for
the group as a whole, but also for the individual members of the group. The dis-
agreement need not last forever. In fact, in the studies the disagreements within
groups vanished fairly quickly. There are difficult questions concerning when
the benefits of continued disagreement wear off, and I won’t address those
here. The purpose here is simply to show that there is another kind of epistemic
reason out there, one that is often ignored.21,22
While such reasons are properly considered epistemic, since they are rea-
sons that one has simply from occupying the epistemic perspective, they are
not the kind of epistemic reasons addressed by the Equal Weight View. The rea-
sons to maintain one’s beliefs in the face of disagreement are concerned with
fulfilling your epistemic goals (believing truths and not believing falsehoods)
in the long run – we might call them diachronic epistemic reasons. They are
19 See Mercier (2012), and Dawson et al. (2002).

20 For more on this argument see Matheson (2015b).
21 The relationship between epistemic reasons possessed by a group and epistemic reasons
possessed by an individual is interesting, deserving of more attention, and bound to be
contentious. The core of my argument here is that the Equal Weight View doesn’t say any-
thing about whether an individual has any such epistemic reasons. I am not arguing that
we have such reasons for religious belief, just that the Equal Weight View does not deny
that we do. Additional premises may be added to the Equal Weight View to get that more
skeptical conclusion, but my focus here is simply on what consequences the Equal Weight
View itself has for the rationality of religious belief.
22 An additional kind of epistemic reasons to believe that God exists that parallels the rea-
sons offered Pascal. Believing that God exists could be a great means to arriving at a num-
ber of true beliefs in the future (and avoiding false ones), true beliefs that would perhaps
be unavailable to those without such a belief. Such reasons are epistemic reasons, since
they are reasons that one has from the epistemic perspective – they are reasons that one
has in virtue of having the twin goals of believing truths and not believing falsehoods.
That said, such reasons are importantly different from other kinds of epistemic reasons
that we have – the kind of epistemic reasons epistemologists have typically been con-
cerned with.
distinct from reasons concerned with fulfilling your epistemic goals (believing
truths and not believing falsehoods) right now – what we might call synchronic
epistemic reasons. Diachronic epistemic reasons do not provide the kind of
epistemic justification required for knowledge (at least not knowledge right
now), whereas synchronic reasons do. While they differ in this respect, we have
seen reason to view both types of reasons as epistemic reasons.
With this distinction in hand, we can see that the Equal Weight View con-
cerns only the synchronic epistemic reasons that you may have for your reli-
gious beliefs. The Equal Weight View does not address your long term epistemic
goals and how the doxastic attitudes you have now may affect your ability to
fulfill those goals. Since the Equal Weight View concerns only your synchronic
epistemic reasons, and those reasons do not exhaust your epistemic reasons,
such views do not make pronouncements about what you should and should
not believe from the epistemic perspective as a whole. Rather, the Equal Weight
View is concerned with one type of epistemic reasons that you may have for
your beliefs, so the Equal Weight View simply fails to make pronouncements
about what you should believe from the epistemic perspective. This is not an
argument that there are good diachronic epistemic reasons for religious belief
(nor is it a denial), but this shows that there are more epistemic reasons out
there then what the Equal Weight View considers.23
So, the sense in which the Equal Weight View has it that religious belief is
irrational is only this fourth sense of rationality:
R4: S’s belief that p is rational iff from the synchronic epistemic perspec-
tive, p is what S should believe.
The Equal Weight View concerns only epistemic reasons for belief, and of
those, only one particular type – synchronic epistemic reasons. Regarding
these reasons, the Equal Weight View has it that they are fully defeated by our
awareness of a significant controversy over the relevant proposition.
23 It might be thought that I have mischaracterized the epistemic perspective, even if I have
done so in good company. For instance, it might be thought that the epistemic perspec-
tive ought to be characterized in terms of ascertaining knowledge. An evaluation of this
alternative conception of the epistemic perspective goes beyond the scope of this paper,
but it should be noted that even on such a conception the distinction between one’s syn-
chronic and diachronic epistemic goals applies. For the reasons considered above, some
beliefs do better at getting at knowledge now, and other beliefs can do better at getting at
knowledge in the long run. So, even on such a conception of the epistemic perspective, it
makes sense to distinguish these two epistemic projects.
100 Matheson
5 Evaluating the Epistemic Consequences
So, at last we have seen the epistemic consequences of the Equal Weight View
for religious beliefs. Given awareness of the extensive disagreement concern-
ing religious propositions, the Equal Weight View has it that we lack (undefeat-
ed) synchronic epistemic reasons that support a non-skeptical attitude toward
those propositions. As such, the Equal Weight View has the consequence that
such individuals currently lack knowledge of religious propositions as well.
However, we have seen that this does not entail that religious belief is ridicu-
lous, that we (all-things-considered) should not have such beliefs, or even that
we should not believe such things from the epistemic perspective (with the
sole goals of believing truths and not believing falsehoods). Given all of this,
the epistemic consequences of the Equal Weight View for religious belief are
much more minimal than they have been thought to be.24
Having finally narrowed in on the epistemic consequences (or lack thereof)
of the Equal Weight View for religious belief, we turn now to evaluating how
significant or worrisome these consequences are. The first thing to note is that
these epistemic consequences are contingent, they are not inevitable. Any
state of controversy is a contingent matter, and while it may be reasonable to
believe that religious controversies will stick around for the foreseeable future,
the Equal Weight View does not mandate that they do. The kind of defeater
for religious beliefs that comes by way of disagreement is not inevitable. The
second thing to note is that these epistemic consequences are contained; they
do not affect everyone. The disagreement defeater only comes by way of our
awareness of the extensive controversy surrounding religious propositions.
The mere existence of religious disagreement has no defeating effects. So,
those who are unaware of the controversy surrounding their religious beliefs
will not be susceptible to the disagreement defeater.
In addition, there is no special problem here for religious beliefs.25 The Equal
Weight View does not single out religious beliefs in particular. While there is
a great deal of controversy about religious beliefs, religious beliefs are by no
means the only type of beliefs that are significantly controversial. The con-
sequences of the Equal Weight View apply equally to controversial scientific
claims, controversial political claims, and controversial philosophical claims
in general. So, in this way, the Equal Weight View does not have it that religious
beliefs are any worse off, rationally speaking, than one’s favored controversial
24 Note that this is a comparative claim. Many will likely find these consequences of the
Equal Weight View to still be quite substantial. My goal is merely to show that the conse-
quences of the Equal Weight View are much less strong than typically thought.
25 For more on this point, see van Inwagen (1996).
scientific, political, or general philosophical beliefs. In fact, it is plausible

that the Equal Weight View has it that religious beliefs are much better off, in
terms of our synchronic epistemic reasons, than many controversial claims
in science, politics, or philosophy in general. For instance, the higher-order
evidence for theism appears to be much greater than the higher-order evi-
dence for the truth of string theory, that a flat tax should be implemented, or
for an agent-causation view of free will. While, like theism, none of these views
have enough higher-order evidence in their favor to support a judgment that
“belief” would win the relevant epistemic election, the higher-order evidence
in favor of theism appears to be much more impressive than the higher-order
evidence in favor of these other views.26 Returning to the political analogy, we
can compare the results of the relevant epistemic elections. While “belief” fails
to win in the epistemic election regarding each of the following:
A) God exists.
B) String theory is correct.
C) A flat tax should be implemented.
D) An agent-causation account of free will is correct.
It is quite plausible that “belief” does better in the epistemic election regard-
ing A than it does in the epistemic elections regarding B-D. Theism is rather
impressive in terms of the number and quality of its defenders. While there are
a number of complications that must be considered in determining the exact
state of the relevant controversies,27 theism appears to perform relatively well
amongst the significantly controversial claims in science, politics, and philoso-
phy in general. Given that, the Equal Weight View would have it that we have
better synchronic epistemic reason to endorse A than we do to endorse B-D.
While given the Equal Weight View, these reasons fail to make us on balance
synchronically epistemically justified in believing A (and thus fail to allow us to
know A); belief that God exists would have more synchronic epistemic reasons
going for it than either of B-D, even if we still lack on balance positive synchronic
epistemic reasons to so believe. So, not only is there no special problem here for
religious belief, given the Equal Weight View, religious belief actually appears
to fare better than many other controversial propositions in science, politics,
and in philosophy in general. Those who believe B, C, or D do so on the basis
of less synchronic epistemic reasons than those who believe A. This grounds
an important comparative claim, that religious believers are no less rational
(in the given sense) than believers of B, C, or D. In fact, it looks like religious
26 For a related discussion, see Kelly (2011).

27 For more on this point, see Carey and Matheson (2013) and Matheson (2015a).
102 Matheson
b elievers actually have more synchronic epistemic reasons for their belief (even
if those reasons do not suffice to make them on balance synchronically epis-
temically justified to believe A). Given all of this, the epistemic consequences
of the Equal Weight View for religious beliefs are much less severe than they
have been made out to be, and they may not even seem to be all that bad.
Bibliography
Bogardus, T. 2013. “Disagreeing with the (Religious) Skeptic,” International Journal for
Philosophy of Religion 74: 5–17.
Carey, B. & J. Matheson. 2013. “How Skeptical is the Equal Weight View?” In Diego Ma-
chuca (ed.), Disagreement and Skepticism, 131–149. New York: Routledge.
Christensen, D. 2007. “Epistemology of Disagreement: The Good News,” Philosophical
Review 116: 187–218.
Christensen, D. 2009. “Disagreement as Evidence: The Epistemology of Controversy,”
Philosophy Compass 4: 756–767.
Conee, E. 1987. “Evident, but Rationally Unacceptable,” The Australian Journal of Phi-
losophy 65: 316–326.
David, M. 2001. “Truth as the Epistemic Goal.” In M. Steup (ed.), Knowledge, Truth, and
Duty: Essays on Epistemic Justification, Virtue, and Responsibility, 151–169. Oxford:
Dawson, E., T. Gilovich & D. Regan. 2002. “Motivated Reasoning and Performance on
the was on Selection Task,” Personality and Social Psychology Bulletin 28: 1379–1387.
De Cruz, H. (2017) “Religious disagreement: An empirical study among academic phi-
losophers,” Episteme: A Journal of Social Philosophy 14: 71–87.
Dunn, J. (forthcoming). “Peer Disagreement and Group Inquiry.”
Elga, A. 2007. “Reflection and Disagreement,” Noûs 41: 478–502.
Elga, A. 2011. “How to disagree about how to disagree.” In R. Feldman & T. Warfield
(eds.), Disagreement, 175–186. New York: Oxford University Press.
Feldman, R. 2006a. “Reasonable Religious Disagreements.” In L. Antony (ed.), Philoso-
phers without Gods: Meditations on Atheism and the Secular Life, 194–214. New York:
Feldman, R. 2006b. “Epistemological Puzzles about Disagreement.” In Stephen Hether-
ington (ed.), Epistemology Futures, 216–236. Oxford University Press.
Frances, B. 2008. “Spirituality, Expertise, and Philosophers.” In J. Kvanvig (ed.), Oxford
Studies in Philosophy of Religion, 44–81. New York: Oxford University Press.
Frances, B. 2010. “The Reflective Epistemic Renegade,” Philosophy and Phenomenological
Research 81: 419–463.
Graves, S. 2013. “The Self-Undermining Objection in the Epistemology of Disagree-

ment,” Faith and Philosophy 30: 93–106.
Kelly, T. 2011. “Consensus Gentium: Twenty-First Century Reflections on the ‘Common
Consent’ Argument for the Existence of God.” In K. Clark and R. Van Arragon (eds.),
Evidence and Religious Belief, 135–156. Oxford University Press.
King, N. 2011. “Disagreement: What’s the Problem? Or A Good Peer is Hard to Find,”
Philosophy and Phenomenological Research 85: 249–272.
Kornblith, H. 2010. “Belief in the Face of Controversy.” In R. Feldman and T. Warfield
(eds.), Disagreement, 29–52. New York: Oxford University Press.
Kvanvig, J. 2005. “Truth is not the Primary Epistemic Goal.” In E. Sosa and M. Steup
(eds.), Contemporary Debates in Epistemology, 285–296. Oxford: Blackwel.
Kvanvig, J. 2008. “Pointless Truth,” Midwest Studies in Philosophy 32: 199–212.
Littlejohn, C. 2013. “Disagreement and Defeat.” In D. Machuca (ed.), Disagreement and
Skepticism, 169–192. New York: Routledge.
Matheson, J. 2014. “Disagreement: Idealized and Everyday.” In J. Matheson & R. Vitz
(eds.), The Ethics of Belief: Individual and Social, 315–330. New York: Oxford Univer-
sity Press.
Matheson, J. 2015a. The Epistemic Significance of Disagreement. New York: Palgrave.
Matheson, J. 2015b. “Disagreement and the Ethics of Belief.” In James Collier (ed.), The
Future of Social Epistemology: A Collective Vision, 139–148. Lanham, Maryland: Row-
man and Littlefield.
Matheson, J. 2015c. “Are Conciliatory Views of Disagreement Self-Defeating?” Social
Epistemology 29: 145–159.
Matheson, J. 2015d. “Disagreement and Epistemic Peers.” Oxford Handbooks Online in
Philosophy. DOI: 10.1093/oxfordhb/9780199935314.013.13.
Mercier, H. 2011. “Reasoning Serves Argumentation in Children,” Cognitive Development
26: 177–191.
Moshman, D. & M. Geil. 1998. “Collaborative reasoning: Evidence for collective ratio-
nality,” Thinking & Reasoning 4: 231–248.
Nickerson, R. 1998. “Confirmation bias: A Ubiquitous Phenomenon in Many Guises,”
Review of General Psychology 2: 175–220.
O’Connor, T. 1999. “Religious Pluralism.” In M. Murray (ed.), Reason for the Hope Within,
165–181. Grand Rapids: Eerdmans.
Pascal, B. 1670. Pensées. Translated by W.F. Trotter. London: Dent, 1910.
Pittard, J. 2016. “Religious Disagreement,” The Internet Encyclopedia of Philosophy,
http://www.iep.utm.edu/rel-disa/, 02/18/2016.
Pittard, J. 2014. “Conciliationism and Religious Disagreement.” In M. Bergmann & P.
Kain (eds.), Challenges to Moral and Religious Belief: Disagreement and Evolution,
80–100. Oxford University Press.
104 Matheson
Plantinga, A. 1999. “Pluralism: A defense of religious exclusivism.” In P.L. Quinn & K.

Meeker (eds.), The Philosophical Challenge of Religious Diversity, 172–192. New York:
Pitchard, D.H. 2014. “Truth as the Fundamental Epistemic Goal.” In J. Matheson and
R. Vitz (eds.), The Ethics of Belief: Individual and Social, 112–129. New York: Oxford
University Press.
Schellenberg, J.L. 2007. The Wisdom to Doubt: A Justification of Religious Skepticism.
Ithaca, NY: Cornell University Press.
Schulz-Hardt, S., M. Jochims, & D. Frey. 2002. “Productive conflict in group decision
making: genuine and contrived dissent as strategies to counteract biased informa-
tion seeking,” Organizational Behavior and Human Decision Processes 8: 563–586.
Sunstein, C. 2002. “The Law of Group Polarization,” Journal of Political Philosophy 10:
175–195.
Taliaferro, C. 2009. “Philosophers without Gods: A Review Essay,” Philosophia Christi
11: 224–230.
van Inwagen, P. 1996. “It is Wrong, Always, Everywhere, and for Anyone, to Believe Any-
thing, Upon Insufficient Evidence.” In J. Jordan & D. Howard-Snyder (eds.), Faith,
Freedom, and Rationality, 137–154. Hanham, MD: Rowman and Littlefield.
Weatherson, B. 2014. “Disagreements, Philosophical and Otherwise.” In D. Christensen
& J. Lackey (eds.), The Epistemology of Disagreement: New Essays, 54–76. New York:
Chapter 7
Pragmatist Anti-Skepticism: at What Price?

Scott Aikin and Thomas Dabay
1 Pragmatism and Anti-Skepticism
Anti-skepticism comes in a variety of forms. There is the heroic-quixotic quest

of refuting skeptics on their own grounds. There is the program of showing
skepticism to be self-defeating. Additionally, there is the argument that the
skeptic cannot live: the argument from apraxia. These programs, though popu-
lar, are regularly inconclusive. The direct heroic-quixotic reply often proves too
little,1 and the self-refutation program is easily turned into a case for skepti-
cism.2 Finally, apraxia touches on only the most thinly-stuffed straw men, as
almost all skeptical programs are conjoined with either Sextus Empiricus’s Pyr-
rhonian fourfolds (PH i.17), Montaigne’s gay and sociable reason (Essays iii, 13,
1044), Cartesian provisional morality (csm i. 122–4), or Humean diffidence and
modesty (1978: Appendix).3
A more fruitful program of anti-skepticism begins with the charge of high
redefinition against the skeptics. The charge runs as follows: the skeptic’s stan-
dards for knowledge are set unnecessarily high, and once the standards are
appropriately set, knowledge is not only manifest, but it is abundant. The prag-
matist tradition has two programs of redefinitive anti-skepticism focusing on
the justification and truth-conditions of knowledge, respectively.
The first is the pragmatist’s fallibilist outlook on knowledge: it is possible
to know without having truth-entailing reasons. So, one may be sufficiently
justified for knowledge, but nevertheless be wrong. The skeptic, as the high
redefinition charge goes, incorrectly assumes that knowledge requires infallible
1 See Margolis (1973), Rescher (1974), (2003), Williams (1991), DeRose (1999), and McGinn
(2003). See also Nagel’s account of “heroism” about objectivity in The View from Nowhere
(1986: 68–69).
2 See Sextus Empiricus (PH ii.192), McPherran (1978), Bailey (1990), and Aikin & Simmons
(2009). The basic form of this argument is that because the premises for the argument for
skepticism are derived from our shared conceptions of knowledge, the self-refutation prob-
lem is not just a problem for skeptics, but for our conceptions of knowledge. Consequently,
the self-refutation argument against skepticism purports to show that our conception of
knowledge is incoherent.
3 See Frede (1987), Ribeiro (2002) and Thorsrud (2009) for models of Pyrrhonian praxis.

106 Aikin and Dabay
justification. Once we are knowledge fallibilists, the pragmatist argues,

skepticism evaporates.
The second program of pragmatist anti-skepticism is to redefine what it is
for our beliefs, claims, and theories to be true. The skeptical tradition starts
with the thoughts that there is a difference between appearance and reality,
and that we can inquire into appearances to the best of our abilities and still
get reality wrong. Truth and our best reasons are distinct, and this is where the
skeptical challenges arise. In contrast, the pragmatist holds that truth is a func-
tion of how our beliefs serve our interests, how well our plans are achieved,
and to what extent we are able to avoid perceptual surprise. Truth, then, is a
function of inquiry, and so inquiry can, at least in principle, approach all the
truths we care to inquire about.
As a consequence of these two lines of thought, the redefinitive programs
of pragmatist anti-skepticism are focused on the redefinition of two concepts
at the core of knowledge: justification and truth. The anti-skeptical payoff of
the conjoined programs of pragmatist fallibilism and the pragmatist theory
of truth is clear. On the pragmatist theory of truth, many of the skeptical chal-
lenges are no longer formulable as intelligible reasons to doubt, and areas
wherein there is acknowledged ignorance are so only for contingent reasons.
On the pragmatist view of fallible knowledge attributions, we may know many
things without having to eliminate every possible defeater. Having enough evi-
dence to make a commitment sufficiently probable is enough for knowledge-
level justification. And as a consequence of this twofold redefinitive program,
a muscular anti-skepticism arises, along with a robust theory of inquiry and a
model for intellectual progress. This all sounds like excellent news. As a conse-
quence, Hilary Putnam accurately observes that “pragmatism has been charac-
terized by its anti-skepticism” (1995: 20).
Our thesis is that this pragmatist anti-skeptical program, despite its clear
benefits, also has significant costs. We will highlight three here. The first is that
the pragmatist version of fallibilism and the pragmatist theory of truth are in
tension with each other. This, we call the bad teammates thesis. The second
is that fallibilism, at least as formulated by the pragmatist as part of an anti-
skeptical program, is itself incoherent. This, we call the incoherence of fallibil-
ism thesis. The third is that the pragmatist theory of truth proves too much as
an anti-skepticism – it proves that we are omniscient. This, we call the proves
too much thesis.
Our plan in what follows will be to sketch out the motives of pragmatist fal-
libilism and the pragmatist theory of truth. We will then outline the cases for
the three theses. We will then close with the question of whether these costs of
pragmatist anti-skepticism are too high – that is, we will ask whether it would
instead be better to revert to skepticism or to rehabilitate heroic anti.
Pragmatist Anti-Skepticism: At What Price? 107
1.1 Fallibilism
The charge that skepticism is based on high redefinition is a form of semantic
reply to skepticism – one challenges and then changes (or clarifies) the
meanings of the terms of the debate. In the case of fallibilist redefinitive
anti-skepticism, the term to be reconsidered is that of justification and then,
thereby, knowledge. Most fallibilists and infallibilists alike agree to the follow-
ing two claims:
S knows that p ONLY IF S has justifying reasons for p

S has justifying reasons for p ONLY IF S has reasons that establish that p,
as opposed to p’s contraries, is true.
Reasons, insofar as they show that one proposition among many that could be
true is true, must be discriminative – they must distinguish the true from the
false.
What separates fallibilism from infallibilism is how each understands what
it means for a reason to establish that something is true. Infallibilism requires
that S’s justifying reasons must imply the falsity of p’s contraries. So S’s reasons
must make it so that it is epistemically impossible for those reasons to obtain
and any of p’s contraries be true. The infallibilist’s reasoning is that if we can
make sense of the best, then why should we accept anything less?
In contrast, the fallibilist holds that one may know a proposition is true on
the basis of evidence that is “good enough,” i.e. evidence that, although not
truth-entailing, does make p more likely true than false. And so, one’s knowl-
edge-qualifying reasons for p may not eliminate every contrary of p as false.
The reasoning behind the fallibilist’s thought is that in life we are constantly
making do with the merely good enough, so why would this be any different
when it comes to knowledge and justification?
The consequences that these different ways of thinking have for knowledge
are profound. If one is an infallibilist about knowledge, then one must hold (as-
suming that having justifying reasons is a necessary condition for knowledge):
Infallibilism: S knows that p only if every contrary to p is inconsistent with

S’s evidence.
It is this principle that runs the famous programs of skeptical doubt-makers –

dreams, demons, Matrix computers, and mad scientists galore. On the other
hand, one may be a fallibilist and hold that, to the contrary:
Fallibilism: Possibly, S knows that p even though S’s evidence is consistent

with the truth of some contrary to p.
108 Aikin and Dabay
Hopefully, this characterization of fallibilism helps us to understand what fal-

libilism is, but it is important to note that this characterization is a negative
one (Fallibilism is little more than the negation of Infallibilism) and so it can
only go so far.
To give a positive characterization, we must say a few words about the
modal operator in Fallibilism above and what justifies us in using it. Since
Fallibilism is more or less the negation of Infallibilism, it is worth noting that
the concepts at the core of Infallibilism are logical ones. What matters for In-
fallibilism are the truth values of and the entailment relations between items
in a subject’s total evidence set, and the resulting (in)consistency of those
items with the proposition under consideration as well as that proposition’s
contraries.
However, if we take these same logical concepts to be at the core of Fallibil-
ism, then we run into problems when we try to make sense of our epistemic re-
lations to necessary truths. To see why, imagine that I am good but not great at
constructing natural deduction proofs, and that I know this fact about myself.
When constructing a proof for a particularly complicated theorem, I am liable
to make minor mistakes that I do not notice in the moment, but that I almost
always identify once I double-check the proof. Now imagine my epistemic po-
sition after I complete the proof of this theorem but before I double-check it.
In an everyday sense I know that the theorem is true, but only fallibly so. I know
it because I have the strongest possible evidence for it (i.e. a logical proof); but
until I have double-checked the proof, my knowledge remains fallible because
I must acknowledge that I might have made mistakes (even in the case where
I have not actually made any).
However, this everyday sense of fallible knowledge is not what we get if we
interpret the phenomenon at the core of Fallibilism to be a logical one. Since
by hypothesis, the target of my knowledge is a theorem, any contraries to it
will be inconsistent with the basic principles of natural deduction. Addition-
ally, since I am able to construct proofs in the first place, I know the basic
principles of natural deduction (meaning that they are included in my overall
evidence set). Therefore, no contraries to the claim that the theorem is true
will be logically consistent with my evidence, which just means that I do not
fallibly know that the theorem is true—I infallibly know it. This is surely too
strong a knowledge claim given my tendency for making mistakes, but the only
other option left is to deny that I have knowledge of any sort before I double-
check the proof.
Infallibilists will no doubt think this is the appropriate interpretation of my
epistemic position, but accepting this interpretation stacks the deck in favor
of skepticism in precisely the manner that the fallibilist is trying to avoid. If I

can make mistakes in a proof, the skeptic will note that surely I can make mis-
takes while double-checking a proof. So imagine that I have double-checked
the proof and found no errors. At this point, the skeptic needs only walk me
through an analogous line of thinking as above to demonstrate that my only
options will be to insist that I infallibly know that the theorem in question is
true after double-checking my proof or that I do not know whether the theo-
rem is true at all. The former response makes too strong a claim given my (ad-
mittedly weaker) tendency to make mistakes while double-checking proofs,
but the latter response opens the door to a regress of a thoroughly predictible
sort—I am forced to triple-check the proof, then choose between insisting on
my having infallible knowledge or conceding that I do not have knowledge at
all, leading to me quadruple-checking the proof, and so on. At this point, it is
hardly necessary for the skeptic to note that no proof is so obvious and no logi-
cian so astute for it to be literally logicallly impossible for the logician to make
a mistake while constructing the proof, and so this line of thinking generalizes
to any person’s epistemic relation to any necessary truth.
This final conclusion leaves the fallibilist in the unenviable position of
not being able to give a genuinely fallibile account of a person’s relation to
necessary truths—such an account must either be infallibilist or skeptical.
Luckily for the fallibilist, this conclusion is premised on the thought that the
phenomenon at the core of Fallibilism is a logical one. Given that both we
and the fallibilist would like to maintain the thought that knowledge, even of
necessary connections, can be fallible, the most convincing fix is to interpret
Fallibilism in epistemic (as opposed to logical) terms.4 What makes the logical
phenomenon in Fallibilism possible is something along the lines of the follow-
ing epistemic phenomenon:
K-Fallibilism: Possibly, S knows that p and some of p’s contraries are epis-
temically possible for S, given S’s evidence.
A proposition is an epistemic possibility for a subject when that subject does

not know, given the subject’s evidence, that the proposition is false. This notion
is captured in natural language with the expression for all you know. The advan-
tage of characterizing the fallibilist outlook in terms of epistemic possibilities,
4 The literature on defining fallibilism precisely is rich with debate over the appropriate terms
for definition; however, it is clear that some epistemic, as opposed to logical, term must play
the role. See as exemplary Reed (2001), (2012), (2013), Harper (2010), Hetherington (2013).
110 Aikin and Dabay
as opposed to logical terms such as entailment or consistency, is that even if a

subject does believe a proposition whose contraries are all inconsistent with
her evidence, it does not follow that the subject knows about these contraries
and their inconsistencies. Ironically, this lack of knowledge on the subject’s
part is precisely what allows fallibilists to acknowledge that the subject has fal-
lible knowledge, because without the link between the subject and these con-
traries there is nothing to motivate the dilemma from above, of either inflating
the subject’s merely fallible knowledge into infallible knowledge and deflating
it to something less than knowledge. It is this feature of epistemic possibility,
and by extension K-Fallibilism, that allows us to articulate fallibilist accounts
of knowledge of necessary truths.
Now that we have given a positive account of fallibilism, one that has led us
to substitute K-Fallibilism for Fallibilism for being more in line with the spirit
behind the fallibilist outlook, it will be helpful to articulate an epistemic ver-
sion of Infallibilism as well:
K-Infallibilism: S knows that p only if none of p’s contraries are epistemi-

cally possible for S, given S’s evidence.
Since we will be focusing on K-Fallibilism going forward, it will be helpful to

use K-Infallibilism as its foil and not the original Infallibilism.
Now that we have finished characterizing the fallibilist outlook, we should
highlight how the history of pragmatism’s epistemology is rich with avowals of
views in this vein. A few exemplary cases include:5
C.I. Lewis: Fallibilism “is the only thing that would save those who appreci-
ate the continually changing character of this spectacle from skepticism.”
(1929: 273)
Cheryl Misak: “It is a good thing proof is not required.” (2013: 147)
Hilary Putnam: “It is perhaps the unique insight of American pragmatism

(that one may be) both fallibilistic and anti-skeptical.” (1994: 152)
Nicholas Rescher: “The cognitivist who rejects skepticism and purports

to know some fact need not insist that he has intrinsically irrefutable and
logically conclusive evidence that this obtains.” (2003: 40)
5 See also Apel (2001: 453), Friedman (1999: 733), De Waal (2005: 16), Hildebrand (2008: 46),
Cooke (2006: 31), Hookway (2012: 37), Stout (2007: 11), Lewis (1946: 182), Murphy (1990: 11),
Rescher (2001: 89), Brown (2008: 36) for statements of the pragmatist-fallibilist outlook and
its anti-skeptical purport.
William James: “But please observe, now, that when as empiricists we

give up the doctrine of objective certitude, we do not give up the quest or
hope of truth itself.” (WB 23)
John Dewey: “Knowledge … is a name for the product of competent in-

quiries. Apart from this relation, its meaning is so empty that any content
… may be arbitrarily filled in.” (LW 12: 16)
C.S. Peirce: “A person, it is true, in the course of his studies, may find rea-
son to doubt what he began believing, but in that case he doubts because
he has a positive reason for it, and not on account of the Cartesian max-
im. Let us not pretend to doubt in philosophy what we do not doubt in
our hearts.” (CP 5: 265)
In the wake of the pragmatist tradition’s avowal of fallibilism, the pragmatist is

able to pose the following Fallibilist Master Argument against skepticism:
1. Skepticism is true only if K-infallibilism is true.

2. Knowledge is widespread.
3. K-infallibilism is inconsistent with widespread knowledge.
4. Therefore, Knowledge Infallibilism is false.
5. Therefore, Skepticism is false.
Anti-skepticism is part of the fallibilist program, and so, as Nicholas Rescher

quips: “fallibilism is our destiny” (2003: 36).
1.2 The Pragmatist Theory of Truth

Skeptical doubts require that there be a gap between appearance and real-
ity. The smaller the gap, so anti-skeptical reasoning goes, the less the skep-
tic has to work with. So, if you want to be anti-skeptical, reduce the gap. In
essence, this is the impetus for the idealists’ and phenomenalists’ revolt
against realism, and so we see the same with the pragmatist theory of truth.
Peirce’s clarification of our meanings begins with indicators of truth: “our
idea of anything is our idea of its sensible effects” (CP 5: 401). James, simi-
larly, holds that “the true … is only the expedient in our way of thinking” (wwj:
438) and that beliefs “become true just in so far as they help us to get into
satisfactory relation with other parts of our experience” (wwj: 382). Dewey
famously identified truth with the conditions of satisfaction or warranted
assertion:
[T]his is the meaning of truth: processes of change so directed that they

achieve an intended consummation. (LW 1: 128)
112 Aikin and Dabay
[Pragmatism requires] consequences as necessary tests of the validity of

propositions, provided these consequences are operationally instituted
and are such as to resolve the specific problem evoking the operation.
(LW 12: 4)
Once we see the pragmatist program of reconstructing truth along the idealist-
phenomenalist line, the skeptical program then becomes a non-starter.6 C.I.
Lewis famously observes:
The traditional arguments of the skeptic, that knowledge … is dubious or

impossible in light of the subjectivity of sense, is without valid founda-
tion. That our possession of any considerable array of common concepts
depends on the presence to our minds of a common reality is … a com-
monplace (1929: 116)
These thoughts together yield the Pragmatist Theory of Truth’s Master

Argument against Skepticism:
1. Skepticism is intelligible only if truth is distinct from the products of in-

quiry and success of practice.
2. Truth is not distinct from the products of inquiry and success of practice.
3. So, skepticism is not intelligible.
As C.I. Lewis concludes, skeptics “talk nonsense” (1946: 20).
1.3 Formal Prospects for Pragmatist Anti-Skepticism

Pragmatists are committed to fallibilism and the pragmatist theory of truth.
Moreover, these commitments have anti-skeptical consequences. For the re-
mainder of this paper, we will evaluate these commitments and their con-
sequences in the vein of what Robert Brandom calls “analytic pragmatism.”
The pragmatist program is continuous with the Enlightenment project of
developing a secular worldview and articulating ways for humans to reason,
communicate, and live together without domination. Indispensible to this
project are the tools of modern logic, which—as Brandom puts it (1994: xix
and 2009: 11)—are the organs of semantic self-consciousness. Thoughts are
made clearer and explicit when rendered formally and with the commentary
6 Others who take this line with the pragmatist theory of truth and then turn to its anti-skepti-
cal consequences are: Haack (2005: 250), Cooke (2005: 660), Gale (2010: 106), Davidson (1986:
314), Koopman (2009: 116), Hodges and Lachs (2000: 31), Misak (2004: 53), Myers (1999: 643),
Putnam (1994: 55), Brandom (2009: 98), Rescher (2001: 97), and Apel (2001: 453).
of semantic ascent. Consequently, we hold that the pragmatist program not

only is amenable to the tools of analytic philosophy but also positively flour-
ishes in the analytic idiom.7
In the remaining section, we identify the two formal commitments that
make instances of each of these views of a family. The rest of the paper will
then be an investigation of the logical consequences of these two formal com-
mitments, first considering them in conjunction and then in isolation.
Given fallibilism about knowledge, a subject can know a proposition p on
evidence E, even though it is epistemically possible that p is false. For the pur-
poses of testing consistency, we will need some notation for the view. The fol-
lowing should suffice (with “Kp” standing for “p is known by a subject” and
“E◊p” standing for “p is epistemically possible for a subject”). Consequently, the
Knowledge Fallibilist will hold the following:
K-Fal: Kp & E◊~p
We must do the same for the pragmatist theory of truth. In essence, the theory
is a special instance of the epistemic theory of truth, which identifies truth
with ideal justification. Therefore, the truth of a proposition entails that it is
possible to know it. As Nicholas Rescher identifies the core of the pragmatist
view: “being and being knowable in principle can be plausibly identified” (2001:
98). Taking “◊p” to stand for “p is metaphysically possible,” this gives us:
ptt: p ‫◊ ﬤ‬Kp
Roughly: if p is true, then it’s in principle knowable that p. So truth is not iden-
tified with just the things we now know, but with the things we may yet know
or might come to know (assuming the evidence is or would be there).
Going forward, we will use these formulations as the core thoughts behind
our arguments that there are prices the pragmatist must pay, as these costs can
be perspicuously revealed only through formal presentation.
2 The Three Costs
Our contention is that three unhappy consequences follow from the formal
commitments of fallibilism and the pragmatist theory of truth—the bad
7 This view has received criticism from Tom Rockmore (2005) and has been defended by Aikin
and Talisse (2008).
114 Aikin and Dabay
teammates thesis, the incoherence of pragmatist fallibilism thesis, and the

proves too much thesis. We address each in turn.
2.1 Bad Teammates

Given the above formulations of fallibilism and the pragmatist theory of truth, it
seems they should form a dynamic duo, a one-two punch, an unbeatable tag team
of anti-skepticism. But teams work well only when they are cohesive, when the
members do not step on the toes of the others in their joint venture. The problem
is that pragmatism’s joint anti-skeptical commitments form less a dream team
and more a nightmare of conflict. In this section, we begin with a naïve presenta-
tion of this conflict, one that delivers a charge of outright contradiction. Admit-
tedly, this presentation is flawed insofar as it fails to take differences between
knowing subjects into account, but nonetheless it is a useful stepping stone.
Ultimately, our charge is that the combination of fallibilism and the pragmatist
theory of truth leads to a more complex, Moorean sort of self-defeat.
To get started, consider the ways pragmatists do not define truth in terms
of what we currently have evidence for, but what we ideally have warrant to
believe:
C.S. Peirce: “The opinion which is fated to be ultimately agreed to by all

who investigate, is what we mean by the truth” (CP 5: 407)
C.I. Lewis: “Reality cannot be conceived as independent of inquiries long

term” (1929: 116)
Hilary Putnam: Truth is “rational acceptability under ideal conditions”

(1994: 55)
Nicholas Rescher: “The obvious standard for the validation of our beliefs
with respect to their … truth lies in their capacity over the long run to
convince rational inquirers of their cogency” (2005a: 355)
John Dewey: “The ‘truth’ is, by the definition, subject to the outcome of
continued inquires.” (LW 14: 56–57)
And so, when the pragmatists identify truth with some positive output of in-
quiry, it is not with the obviously fallible states of inquiry now—that would be
a manifest contradiction. Rather, the pragmatists identify some perfected in-
quiry, one more completely carried forward—one that we could perform, had
we the time, resources, and interest to do so. Consequently, the claim that true
propositions are knowable, given the notion of perfected or completed inquiry
requiring that all the relevant evidence is collected, will be in fact known under
those conditions. So, the notion of known under these idealized conditions (call
it end of inquiry) requires a new form of knowledge (with “K*p” standing for
knowledge in that sense):
End of Inquiry (ei): If ◊Kp, then K*p
Moreover, since all the evidence is in at the end of inquiry (and for all the ide-
alized forms), we see that one must take it that it is no longer epistemically
possible that the proposition known is false. (If it were epistemically possible
for the proposition to be false, then ex hypothesi the inquiry is not ideal or
complete.) And so:
K* Completeness (K*C): If K*p, then ~E◊~p
The consequence, then, is that any attribution of fallible knowledge will yield
a tension. The trouble is derivable as follows.8 Call it the Bad Teammates Argu-
ment. With Knowledge fallibilism, it is possible to know that p but for it never-
theless to be epistemically possible, given the evidence, that p is false:
1. K-Fal: Kp & E◊~p
The Pragmatist theory of truth, which connects truth and inquiry, then defines
those truths which we fallibly know.
2. ptt: p ‫◊ ﬤ‬Kp
Further, given the notion of the end of inquiry that defines those truths and
the evidence through which they are grasped, we can see that if it is possible to
know that p, then p will be known at the end of inquiry.
3. EI: ◊Kp ‫ ﬤ‬K*p
And if something is known at the end of inquiry, then the evidence for that
proposition is complete and there can be no epistemically possible defeaters
for it or counter-evidence against it:
4. K*C: K*p ‫~ ﬤ‬E◊~p
8 Versions of this argument have been banging around in the back of the minds of pragmatists
and anti-pragmatists for a long time. Lesley Friedman notes it: “[Peirce] is … both an infal-
libilist … and a fallibilist” (1999: 730). See also: Margolis (1973: 87), (1998: 537), Apel (2001: 435),
Cooke (2006: 133), and Rescher (2015: 126).
116 Aikin and Dabay
Let us then allow that we do know the proposition identified by fallibilism in

(1) and draw out the consequences.
5. Kp (1)
6. Kp ‫ ﬤ‬p (K-factivity)
7. p (5,6)
8. ◊Kp (2,7)
9. K*p (3,8)
10. ~E◊~p (4,9)
11. E◊~p (1)
The simple way into the proof is to consider the following reasoning. Imag-
ine, in an anti-skeptical mode, Theresa’s mother attributes some knowledge
to Theresa – that the cat’s on the mat (c). Theresa’s mother, being a pragmatist,
is committed to K-Fal and ptt. Being a fallibilist, she nevertheless holds that
it’s epistemically possible that the cat’s not on the mat, so: E◊~c. But, because
Theresa knows c, it is true. If c is true, then it’s knowable, and if knowable then
known at the end of inquiry. So: K*c. But if c is known at the end of inquiry,
it’s not epistemically possible that c is false. So: ~E◊~c. There’s the tension. The
trouble is that the anti-skepticism promised by the two views yields two dis-
tinct assessments of epistemic possibility. The ptt and K-Fal are bad team-
mates. So, we ask, whose fault is it? (A hint: we think it’s both!)
But before we do any more than hint, we must first acknowledge the flaw
mentioned at the start of this section. Unlike with logical or metaphysical mo-
dalities, epistemic modalities are indexed to subjects in the sense that one per-
son’s evidence can leave something epistemically possible for him even while
another person’s evidence makes it impossible for her. As such, even though
lines 10 and 11 of the derivation above look prima facie to be contradictory
when they are not in fact so. The subject for the epistemic possibility in K -Fal
is some person who is considering a proposition, while the subject for the
Pragmatist Theory of Truth (ptt) under the description of the End of Inquiry
(EI) is possibly the same person, but is not guaranteed to be so. This person is a
kind of idealization of someone who has completed all the relevant inquiries.
With this point acknowledged, we must now reformulate Lines 10 and
11 above to take the subject-indexing into account. Line 11 is a truth about what
is epistemically possible for a subject with fallible knowledge. These are the
subjects normally indexed in these cases with a lowercase “s.” So, conspicuously
presented, 11 should read:
11’: E◊s ~p
Line 10, on the other hand, is a truth about what is epistemically possible for
subjects at the end of inquiry, or under ideal conditions of epistemic access.
This is a significantly different kind of subject to whom we attribute an epis-
temic possibility in 11 (and 11’). We will reference this subject with a lowercase
sigma “σ.” So represented, 10 should instead read:
10’: ~E◊σ ~p
Properly presented and subject-indexed, 10’ and 11’ are now clearly not a
contradiction.
Obviously, this means that our objection cannot be that there is an outright
contradiction between K-Fal and ptt. Instead, our objection is that Lines 10’
and 11’ cause a sort of cognitive dissonance within anyone who accepts K-Fal
and ptt while also fallibly holding that she knows some p. In such a case, the
person sees p’s being false as epistemically possible for her (from 10’), but sees
it as epistemically impossible for those in the ideal epistemic position (from 11’).
However, the thought comprised of the conjunction of 10’ and 11’ has a kind of
Moorean, self-defeating quality to it: one attributes a stronger epistemic posi-
tion to others than one can attribute to oneself, but one does so only on the
basis of evidence one has for the self-attribution.9 Such a commitment is at
best unwarranted, at worst incoherent, and as a consequence the pragmatist
theory of truth and fallibilism are bad teammates.
The best way to capture the Moorean thought here is to put the claim into
the mouth of a subject, in the pragmatist anti-skeptical mode, claiming knowl-
edge. In such a case, like Moore cases, the s-variable for the empirical subject
will be replaced with a self-reflexive indexical i-constant, to stand for the sub-
ject claiming knowledge. So, 11’ should have an instantiation instance for the
subject to a constant:
11”: E◊i ~p
Additionally, as with Moore’s paradox and those phenomena of related form

and consequence, the problem arises less as a matter of logical contradiction
than as a performative contradiction. Moore-style claims have the following
form:
P, but I don’t believe that p.
9 See Moore (1993).

118 Aikin and Dabay
This, of course, can still be true, but it is a thought not knowable or thinkable
from the inside. The second clause, as it were, undoes the first. The same goes
for evidential versions of the Moore’s paradox phenomenon:
P, but I have no evidence for p’s truth.
Again, the second clause undoes the first, as one looks at one’s beliefs or
commitments as though they are mere announced psychological states, not
expressions of one’s rationality. They are not products of assent (see Adler 2002
and Aikin 2014).
The tension between 11” and 10’ can now be properly understood as aris-
ing from stating that something is epistemically possible for an empirical sub-
ject, but not for an epistemically ideally situated subject when the speaker is
the empirical subject in question. That is, the problem arises (as it does with
Moore’s paradox) when the conjunction is in the first person. So, the conjunc-
tion of 10’ and 11” can be true, but it is not something that can consistently be
thought by the subject instance for the constant i. This is because the claim
comes out as:
Not-p is epistemically possible for me, but not for those under ideal epis-
temic conditions.
The point, as with Moore cases, is that the second clause undoes the first –
were one in a position to properly assert the second claim, one would not have
asserted the first. Or, if the first is the best epistemic position one has on the
situation, then one has no grounds for asserting the second. Such a statement,
then, has the epistemic tension of saying something of the form “I do not
know whether p is true, but S knows that p is true.” Either way, the two can-
not fit together internal to a cognizing subject’s intellectual self-assessments.
A tension is at the core of pragmatist anti-skepticism, one between pragmatist
anti-skeptical fallibilism and the pragmatist theory of truth. This makes them
bad teammates.
2.2 Pragmatist Fallibilism is Incoherent

Given the Bad Teammates argument above, the question is which of the
two commitments contributes the bad input. Let us start with Knowledge-
Fallibilism. Our view is that the pragmatist version of anti-skeptical fallibilism
is in trouble. When paired with what we think is a plausible view of what epis-
temic possibility is for pragmatists, such fallibilism is incoherent.
The pragmatist fallibilist, when denying skepticism, is not claiming infal-

lible knowledge, but knowledge that is based on fallible justification. And so S
knows that p, but it is epistemically possible for S that p is false. And so, again,
the form for knowledge fallibilism:
K-Fal: Kp & E◊~p
But what, again, is an epistemic possibility? First, notice that the notion of
epistemic possibility is widespread. When one says to one’s colleague that It
might rain tomorrow, one is expressing a possibility. But it is a possibility of a
specific sort. One is not saying that it is logically possible that it’ll rain tomorrow.
That’s trivial. Rather, one is saying that one does not have sufficient evidence to
eliminate the chance of rain. We think we can explicate the notion of epistemic
possibility with the way inquiry proceeds—we ask questions along the lines of
what could be, and then, with inquiry, eliminate the possibilities to find the
actualities.
For example, and of particular importance to pragmatists who hold that fal-
libilism is part of their theory of inquiry: imagine that Sara has misplaced her
keys. Well, they could be on the bedside table, she thinks to herself. When she
thinks that to herself, she is expressing an epistemic possibility: for all I know,
my keys are on the bedside table. But now, let’s say she looks on the bedside
table, and there are no keys there. So she knows they aren’t there. Once she
knows the keys aren’t on the bedside table, when she goes through the places
it could be, she says:
Could they be on the bedside table? No, they couldn’t be there – I just looked
there. I know they aren’t there. But maybe on the coffee table…
What this shows is something unsurprising, particularly for pragmatists –

epistemic possibilities are functions of what we do and do not know, and so are
indicators of where inquiry can proceed. And that relation is a kind of functional
relation between them that can be captured as a conditional. Something is an
epistemic possibility only if we don’t know it doesn’t obtain. So the principle
of epistemic possibility (EP):
EP: E◊p ‫~ ﬤ‬K~p
So: Sara doesn’t think it’s possible her keys are on the bedside table if she knows
they are not. So when she looks on the table and sees no keys, she knows that
120 Aikin and Dabay
they are not there, and so it’s no longer epistemically possible that they are
there. The key for the pragmatist tradition is that the principle of epistemic
possibility is what’s behind inquiry. The conditional explains how we look for
things: we set out what epistemic possibilities there are, and eliminate them by
going and looking there.10
It’s with EP and the other operative epistemic principles that we think that
pragmatist fallibilism yields incoherence.11 Let us run the argument in reductio
fashion, following a formula developed by Dylan Dodd’s (2010) general criti-
cism of fallibilism. Let’s start with the base form for knowledge fallibilism:
1. K-Fal: Kp & E◊~p
Next, let’s define the epistemic possibility that is the right conjunct of K-Fal,
using EP:
2. EP: E◊p ‫~ ﬤ‬K~p
For the principles to be relevant, we’ll need the target propositions to match –
so we’ll make a replacement in the EP form, so that it is a condition for it being
epistemically possible for a subject that p is false:
3. E◊~p ‫~ ﬤ‬K~~p (2, ~p/p)
We can disaggregate K-Fal with the following:
4. Kp (1)
5. E◊~p (1)
Which then yields the consequent of 3:
6. ~K~~p (3,5)
10 Defenses of a rough version of this model for epistemic possibility can be found in D
eRose
(1991), and Dodd (2010).
11 Versions of the incoherence problem for fallibilism have been run in a variety of ways.
However, this particular line of argument troubled Karl-Otto Apel’s endorsement of
pragmatism, as he pauses with the concern that pragmatist fallibilism involves a “per-
formative contradiction” (2001: 454). This thought about fallibilism is as old as the view
is. Antiochus of Ascalon is reported by Cicero in the Academica to have pointed out this
problem with Academic probabilism as early as the 1st century bce (2006: Ac. 2.36). See
also Lewis (1996), Short (2006), and Talisse and Aikin (2008: 49).
Assuming we may double negate in knowledge contexts:
7. ~Kp (6)
And now, 4 and 7 together yield the contradiction that p is both known and
not known:
8. Kp & ~Kp (3,6)
Notice that this line of reasoning applies in all cases of knowledge-attribution

(assuming pragmatist fallibilist anti-skepticism is true), and so is reflective of
one’s own knowledge-assessments. So given fallibilism, one both knows and
does not know what one, by hypothesis, knows. That’s not the anti-skepticism
we were promised.
2.3 The Pragmatist Theory of Truth Proves Too Much

Given this failing of fallibilism, the pragmatist might make the following hasty
appraisal of the dialectical situation. The Bad Teammates argument establish-
es that, between fallibilism and the pragmatist theory of truth, one of them
must be incorrect; and the incoherence of pragmatic fallibilism establishes
that anti-skeptical fallibilism is incorrect. Therefore, at least the pragmatist
theory of truth is correct.
Unfortunately for the pragmatist, the Bad Teammates argument establishes
that at least one of his commitments is mistaken, leaving open the possibil-
ity that both fallibilism and the pragmatist theory of truth are incorrect—a
possibility that begins to look more like an actuality when we consider the
implications of the latter. Using a formal proof first developed by Frederic
Fitch (2003),12 we can see that the pragmatist’s commitment to the claim that
all truths are knowable entails a commitment to the claim that all truths are
actually known. The implausibility of the latter claim should make us question
the former.
To see why this is the case, we must first remember that the pragmatists’ the-
ory of truth commits them to the claim that all truths are in principle knowable:
ptt: p ‫◊ ﬤ‬Kp
12 See Theorem 5 in Fitch (2003). Commentators have developed two ways of presenting
Fitch’s proof. For the first way, which presents it as a reductio, see Kvanvig (2006: 10–14).
For the second, which presents it in terms of the collapse of knowability into knowledge,
see Rescher’s Epistemic Logic (2005b: 65) and Salerno’s New Essays (2009: 1–3). In what
follows, we adopt this second mode of presentation.
122 Aikin and Dabay
This is the first of two premises needed to derive our counterintuitive conclu-
sion that all truths are known.
The second premise follows from uncontroversial principles of epistemic
logic. Assume for the purpose of reductio that a subject S knows the following
conjunction: that a proposition p is true and that S does not know that p is true.
Because one knows a conjunction only if one knows the individual conjuncts
within the conjunction, this assumption entails that S knows that p is true and
S knows that S does not know that p is true. And because one knows a proposi-
tion only if that proposition is true, this latter point entails that S knows that p
is true and S does not know that p is true. Because this is a direct contradiction,
the original assumption cannot possibly be true. Formalizing the denial of the
possibility of this assumption gives us our second premise:
Limitation of Knowledge (LK): ~◊K(p & ~Kp)
From here, the trouble is that given ptt and LK, deriving that all truths are
known is a simple matter of first-order logic. This is our Proves Too Much argu-
ment. The first step is to substitute the embedded conjunction from LK for
p in ptt:
1. ptt: p ‫◊ ﬤ‬Kp
2. LK: ~◊K(p & ~Kp)
3. (p & ~Kp) ‫◊ ﬤ‬K(p & ~Kp) (1, (p & ~Kp)/p)
Using modus tollens, we can derive:
4. ~(p & ~Kp) (3, 2)
which is equivalent to the claim that if p is true, then p is known:
5. p ‫ ﬤ‬Kp (4)
To appreciate what this proof means in ordinary language, we must again

consider Moore’s paradox.13 Just as Moore’s paradox arises by noting the per-
formative contradiction involved in saying something of the form:
It is raining, but I don’t believe that it is raining,
13 See again Moore (1993).

line 2 expresses the performative contradiction involved in knowing some-

thing of the form:
It is raining, but I don’t know that it is raining.
The key move in the proof (see lines 3 and 4) is to realize that, given the prag-
matist theory of truth, this performative contradiction in knowing something
of this form entails the falsity of anything of this form. The result is that the
pragmatist’s seemingly plausible commitment to ptt collapses into an over-
whelmingly implausible commitment to line 5, the Omniscience Thesis, which
states that if a proposition is true, it is known.14
A great irony arises from this result. The ptt was designed as a form of an-
ti-skepticism. A desideratum of anti-skepticisms is that they reclaim a large
domain of knowledge for us – they re-establish us as knowers. In fact, this de-
sideratum is not just that we are knowers of some things, but perhaps of many
things. However, the problem with the ptt, as we have shown, is that it over-
does the job. It yields complete omniscience. We don’t just know many things, we
know all the things. This would be excellent news, except for the fact that if the
excellent news is news that we are omniscient, the proof’s conclusion is its own
counter-example. As a consequence, the pragmatist theory of truth proves too
much in its anti-skeptical role.
3 Consequences and Prospects
Our thesis is that pragmatist anti-skeptical fallibilism is incoherent. Everywhere

it goes, it is part of what yields contradiction. It does so particularly when put
into parallel service with epistemic theories of truth (which are the theories
of truth many fallibilists hold) and a pragmatic notion of epistemic possibil-
ity. And fallibilism, when set alongside axioms of knowledge-attribution and
norms of assertion, yields conflicting first-personal knowledge assessments.
Further, we have argued that once the pragmatist theory of truth is viewed
clearly, we have reason to see it proves too much as a form of anti-skepticism.
Pragmatist anti-skepticism, it seems, is deeply incoherent.
14 Although our presentation of Fitch’s proof ignores the fact that knowledge is always
knowledge for some subject at some time, similar results obtain even when the presen-
tation includes existential quantification over subjects and times. Suspicious readers
should see Kvanvig, Knowability Paradox (2006: 12–14).
124 Aikin and Dabay
The question is whether this incoherence is fatal. Here is one way to hold
it isn’t. We have many conflicting theoretical and practical commitments, but
such conflict may be an indication of our cognitive bounty and intellectual
fecundity. A few examples from the practical realm may show this. Consider
the problem that if we want to be happy or to maximize our pleasure, that
thought cannot be our operative motive in situ. We must find other, some-
times contrary, motives to follow in the meantime so as to ultimately follow
the initial one. And so the hedonistic paradox is certainly a paradox, but it is
a conflict of perspectives that self-reflective hedonists must manage. Consider
another inconsistency: we want to be successful, but that success wouldn’t be
much for us unless it involved a struggle, even some failure. There is a wealth of
parenting and educator advice now about the importance of teaching children
to fail and learn lessons from those failures. Similarly, inconsistency shows up
in metaphysics – there are no circular explanations, and there can be no unex-
plained explainers, but we hold that there are satisfactory explanations. How
can that be? But even though we have this conflicted view of explanation and
even labor under it, we nevertheless explain and pursue inquiry.15
The same thing can be said for fallibilist anti-skepticisms – fallibilism, as
we’ve argued here, yields a kind of double-mindedness about our knowledge.
We know, but don’t know. That’s incoherent, yes. But look at the practical con-
sequences of managing this double-mindedness. The conflicted thought is a
spur of inquiry, a doubt that pushes thought forward, and it is a kind of intel-
lectual conscience that opposes dogmatism. Fallibilism isn’t thinkable without
taking the goal to be infallibilism (that’s the lesson from Peirce), but the fal-
libilist’s double-mindedness prevents us from falling for the counterfeit cases
of infallibility and certainty. The same goes for the pragmatist theory of truth –
we think, when things are going well, that we’ve got things pretty well figured
out, that we know what we are doing. And that, of course, is when the wheels
fall off and the roof caves in. But that’s just life. We vacillate between satis-
factory arrangements and irritations of problematic situations. Pragmatism’s
program is designed to weather and manage these inconsistencies.
Is the tradeoff worth it? Maybe so, maybe not. Surely, if anti-skepticism is
a desideratum of a theory of knowledge, this is a price many will happily pay.
Many pragmatists have already done so, and many, we think, will continue to
15 T.L. Short (2006) and Nicholas Rescher (2001), (2003) run versions of this inconsistency
management argument. Talisse and Aikin have termed the management version of the
view “policy fallibilism” (2008: 48). In many ways, this pragmatist view is an extension
of the much older Jamesian view articulated in that the pragmatist method allows us to
acknowledge the depth of what would otherwise be impossible theoretical contraries,
but without debilitating inconsistency (wwj: 362).
do so. But the theoretical costs are high, and they require thoughtful manage-
ment. Peirce himself saw such a problem for his view looming, and he con-
cedes that his view should actually be more analogous to the skeptics:
That we can be sure of nothing in science is an ancient truth. The Acad-

emy taught it. (CP 7: 737)
Skepticism doesn’t block the road of inquiry. Rather, it’s a spur for it.
Bibliography
Adler, J. 2003. Belief’s Own Ethics. Cambridge: MIT Press.

Aikin, S. 2007. “Prospects for Skeptical Foundationalism,” Metaphilosophy 38: 578–590.
Aikin, S. 2014. Evidentialism and the Will to Believe. London: Bloomsbury.
Aikin, S. & A. Simmons. 2009. “Otherism, Skepticism, and the Problem of Self-Refuta-
tion,” The Philosophical Forum 40: 29–54.
Aikin, S. & R. Talisse. 2008. “Rockmore on Analytic Pragmatism,” Metaphilosophy 39:
155–162.
Apel, K. 2001. “Pragmatism as Sense-Critical Realism,” Transactions of the Charles S.
Peirce Society xxxvii: 443–474.
Audi, R. 2001. The Architecture of Reason. Oxford University Press.
Bailey, A. 1990. “Pyrrhonean Skepticism and the Self-Refutation Argument,” Philosophi-
cal Quarterly 40: 27–44.
BonJour, L. 2002. Epistemology. Lanham, MD: Rowman and Littlefield.
Brandom, R. 1994. Making It Explicit. Cambridge: Harvard University Press.
Brandom, R. 2009. Reason and Philosophy. Cambridge: Harvard University Press.
Brown, S. 2008. “Fallibilism and the Future of Pragmatism,” Cognitio-Estudios 5: 33–
45.
Cooke, E. 2005. “Peirce, Hookway, and Pihlström on the Conditions of Inquiry,” Trans-
actions of the Charles S. Peirce Society xli: 651–673.
Cooke, E. 2006. Peirce’s Pragmatic Theory of Inquiry: Fallibilism and Indeterminacy. New
York: Continuum.
Cicero, M. 2006. On Academic Skepticism. Trans. Charles Brittain. Indianapolis: Hackett
Publishing.
de Montaigne, M. 1987. The Complete Essays. Trans. M.A. Screech. New York: Penguin
Books. (Cited as Essays Volume.Page Number).
DeRose, K. 1991, “Epistemic Possibilities,” Philosophical Review 100: 581–605.
DeRose, K. 1999. “Responding to Skepticism,” In K. DeRose and T. Warfield (eds.), Skep-
ticism: A Contemporary Reader, 1–26. Oxford University Press.
126 Aikin and Dabay
Descartes, R. 1988. The Philosophical Writings of Descartes, 3 Volumes (CSM). Trans. J.

Cottingham, R. Stoothoff, & Dugald Murdoch, and volume 3 including A. Kenny,
Cambridge: Cambridge University Press. (Cited as CSM Volume.Page Number).
De Waal, K. 2005. On Pragmatism. Belmont: Wadsworth.
Dewey, J. 1925. Experience and Nature (LW 1). Carbondale: Southern Illinois University
Press.
Dewey, J. 1929. The Quest for Certainty (LW 4). Carbondale: Southern Illinois Univer-
sity Press.
Dewey, J. 1938. Logic: The Theory of Inquiry (LW 12). Carbondale: Southern Illinois Uni-
versity Press.
Dodd, D. 2010. “Against Fallibilism,” Australasian Journal of Philosophy. 89: 665–685.
Empiricus, Sextus. 1990. Outlines of Pyrrhonism. Trans R.G. Bury. Amherst: Prometheus
Books. (Cited as PH Book.Page Number).
Fitch, F. 1963. “A Logical Analysis of Some Value Concepts,” Association for Symbolic
Logic 28: 135–142.
Frede, M. 1987. “The Skeptic’s Two Kinds of Assent.” Essays in Ancient Philosophy, 201–
222. Minneapolis: University of Minnesota Press.
Friedman, L. 1999. “Doubt and Inquiry: Peirce and Descartes Revisited,” Transactions of
the Charles S. Peirce Society xxxv: 724–747.
Fumerton, R. 1995. Metaepistemology and Skepticism. Lanham, MD: Rowman and
Littlefield.
Gale, R. 2010. John Dewey’s Quest for Unity. Amherst, NY: Prometheus.
Haack, S. 2005. “Not Cynicism, but Synchretism,” Transactions of the Charles S. Peirce
Society. xli: 239–253.
Harper, A. 2010. “Fallibilism, Contextualism, and Second-Order Skepticism,” Philosoph-
ical Investigations 33: 339–359.
Hetherington, S. 2013. “Concessive Knowledge Attributions: Fallibilism and Gradual-
ism,” Synthese 190: 2835–2851.
Hildebrand, D. 2008. Dewey. Oxford: One World Books.
Hodges, M. & J. Lachs. 2000. Thinking in the Ruins: Wittgenstein and Santayana on Con-
tingency. Nashville: Vanderbilt University Press.
Hookway, C. 2012. The Pragmatic Maxim. Oxford Univerity Press.
Huemer, M. 2007. “Epistemic Possibility,” Synthese 156: 119–142.
Hume, D. 1978. Treatise of Human Nature, 2nd Edition. Edited by L.A. Shelby-Bigge &
P.H. Nidditch. Oxford: Oxford University Press.
James, W. 1977. The Writings of William James (WWJ). J. McDermott (ed.). University
of Chicago Press.
Koopman, C. 2009. Pragmatism as Transition. Columbia University Press.
Kvanvig, J. 2006. The Knowability Paradox. New York: Oxford University Press.
Lewis, C.I. 1949. An Analysis of Knowledge and Valuation. LaSalle: Open Court.
Lewis, C.I. 1929. Mind and the World Order. New York: Dover.
Lewis, D. 1996. “Elusive Knowledge,” Australasian Journal of Philosophy. 74: 549–567.

Margolis, J. 1973. Knowledge and Existence. Oxford University Press.
McGinn, M. 2003. “Responding to the Skeptic,” In S. Luper (ed.), The Skeptics, 149–163.
Burlington: Ashgate.
McPherran, M. 1978. “Skeptical Homeopathy and Self-Refutation,” Phronesis 32:
290–328.
Misak, C. 2004. Truth and the End of Inquiry. Oxford Univeristy Press.
Misak, C. 2013. The American Pragmatists. Oxford Univeristy Press.
Moore, G.E. 1993. “Moore’s Paradox.” In T. Baldwin (ed.), G.E. Moore: Selected Writings,
207–212. London: Routledge.
Murphy, J.P. 1990. Pragmatism: From Peirce to Davidson. Boulder: Westview Press.
Myers, R. 1999. “Pragmatism and Peirce’s Externalist Epistemology,” Transactions of the
Charles S. Peirce Society xxxv: 639–653.
Nagel, T. 1986. The View from Nowhere. New York: Oxford University Press.
Putnam, H. 1994. Words and Life. Cambridge: Harvard University Press.
Putnam, H. 1995. Pragmatism: An Open Question. Cambridge: Blackwell.
Reed, B. 2001. “How to Think about Fallibilism,” Philosohical Studies 107: 143–158.
Reed, B. 2012. “Fallibilism” Philosophy Compass 7: 585–596.
Reed, B. 2013. “Fallibilism, Epistemic Possibility, and Epistemic Agency,” Philosophical
Issues 23: 40–69.
Rescher, N. 1974. “Foundationalism, Coherentism and the Idea of Cognitive Systemati-
zation,” The Journal of Philosophy 71: 695–708.
Rescher, N. 2001. Cognitive Pragmatism. Pittsburgh: University of Pittsburgh Press.
Rescher, N. 2003. Epistemology. Albany, NY: SUNY Press.
Rescher, N. 2005a. “Pragmatism at the Crossroads,” Transactions of the Charles S. Peirce
Society xli: 355–365.
Rescher, N. 2005b. Epistemic Logic. Pittsburgh: Pittsburgh University Press.
Rescher, N. 2015. Cognitive Complications. Lanham, MD: Lexington Books.
Ribeiro, B. 2002. “Is Pyrrhonism Psychologically Possible?” Ancient Philosophy 22:
319–331.
Rockmore, T. 2005. “On Classical and Neo-Analytic Forms of Pragmatism,” Metaphi-
losophy 36: 259–271.
Salerno, J. (ed.). 2009. New Essays on the Knowability Paradox. New York: Oxford Uni-
versity Press.
Short, T.L. 2006. “Fallibilism is Omega-Inconsistent,” Cognitio 7: 293–301.
Stout, J. 2007. “On our Interest in Getting Things Right: Pragmatism without Narcis-
sism,” In C. Misak (ed.), The New Pragmatists, 7–30. Oxford: Oxford University Press.
Talisse, R. & S. Aikin. 2008. Pragmatism: A Guide for the Perplexed. London: Continuum.
Thorsrud, H. 2009. Ancient Skepticism. Berkeley: University of California Press.
Williams, M. 1991. Unnatural Doubts. Princeton: Princeton University Press.
Chapter 8
Extraordinary Skepticism
Earl Conee
1 Introduction
What can be called ordinary skepticism about knowledge is broadly Cartesian.1

Ordinary skepticism asserts the possibility of a skeptical scenario. It is a possibil-
ity in which a demon, a matrix immersion, or some other causal source induces
experiences that deceive us about our environment. We are misled into having
external world beliefs that are largely or entirely false. The other premises of an
ordinary skeptical argument link this possibility to the denial that we know any
of our possibly deceived beliefs. The linking premises say something like this.
First, we are unable to “rule out” that some such deception is occurring. Second,
we know a belief to be true only if we are able to do that ruling out.2
Richard Feldman and I have made a fallibilist case against this ordinary
skepticism.3 We have held that evidence that is strong enough to justify beliefs
for knowledge allows the possibility that the beliefs are false in the skeptical
scenarios. The justifying evidence is strong enough when it meets “the crimi-
nal standard”of proof beyond a reasonable doubt.4 Perception, memory and
inference often furnish evidence for external world beliefs that meets the crim-
inal standard. Knowledge does not require that the knower “rule out” or oth-
erwise address skeptical considerations. But if a knower considers a skeptical
scenario, the knower has evidence that can establish its falsehood by deducing
its negation from the known truth of external world beliefs.
What can be called extraordinary skepticism about knowledge also asserts
that we do not really know some wide swath of our apparent knowledge. The
1 I am very grateful for helpful comments on previous drafts from the editors, Kevin McCain
and Ted Poston, and from David Braun, Ted Sider, Jonathan Vogel, Richard Feldman, and
Thomas Grundmann.
2 For one instance of the pattern, closure arguments for skepticism are like this under their
surfaces. Instead of explicitly calling for “ruling out” the skeptical scenarios, they assert some-
thing to the effect that, by the closure of knowledge under known implication, knowledge
would place us in a position to know that the scenarios do not hold. We can understand be-
ing in this position as the closure version of “ruling out.” The closure arguments also allege
that since we are not in that position, we do not know.
3 Feldman and Conee (2004).
4 The criminal standard is further discussed below, most extensively in Section 2.

Extraordinary Skepticism 129
skeptical reasoning is extraordinary in that it does not argue for the lack of
knowledge by invoking some massive deception.5
I will discuss four versions of extraordinary skepticism. The first argues
against the existence of any fallibly justified knowledge. The second employs
localized skeptical hypotheses. The third cites a feature of Gettier problem
cases to argue that there are pervasive knowledge-preventing accidents. The
fourth argues against the truth of our beliefs.6 Although the arguments are
provocative, I will contend that none of them succeeds. The ways that they fail
show interesting things about knowledge.
2 The Threshold
Laurence BonJour argues against the existence of fallibly justified knowledge.7

This is knowledge with justification that is compatible with the falsehood of
the belief (2010: 57). BonJour accepts that he is asserting “ … a fairly strong form
of skepticism …”, since it has the consequence that “… a very high proportion
of ordinary knowledge claims and attributions are in fact false” (2010: 77). Bon-
Jour’s basic objection to fallibly justified knowledge is that any non-entailing
condition on the justification of the known proposition will set the strength of
justification that suffices for knowledge at some indefensibly arbitrary point.
I will argue in response that a particular fallible threshold – the criminal stan-
dard – can be adequately defended.
Here is my summary of BonJour’s argument against fallibly justified
knowledge.
5 This ordinary/extraordinary skepticism distinction is intended to identify one helpful divi-

sion of skeptical arguments against external world knowledge. It is not exhaustive. There are
other sorts of skeptical arguments that imply a lack of such knowledge, for instance argu-
ments that deny that there is any non-entailing justification of any strength at all. I am grate-
ful to Thomas Grundmann for noting limitations of the distinction.
6 In addition to the forms of extraordinary skepticism that are discussed here, there is one that
targets the proper basing condition on knowledge. In “The Debasing Demon,” Shaffer (2010)
argues roughly as follows. A demon might have given any and all of our beliefs bases that are
not proper for knowledge. We are in no position to know that this is not happening. But if we
are unable to know this, then we lack knowledge. I have opposed this skepticism in “Debas-
ing Skepticism Refuted” (Conee 2015) In a few words the opposing view I advocate is that
we are in a good position to know that our apparent knowledge is properly based. We have
available ample justification for this sort of knowledge by our evidence about why we believe
as we do, evidence from introspection, ordinary observation, and science. The possibility of
a debasing demon leaves this justification undiminished.
7 See BonJour (2010).
130 Conee
bsa
Pr1: If fallibly justified knowledge existed, then there would be a thresh-
old level of non-entailing justification dividing adequate justification for
knowledge from inadequate justification.
Pr2: Any threshold that fell short of entailing justification would be too ar-
bitrary to be compatible with the importance of the status of knowledge.
C: There is no such threshold and hence no fallibly justified knowledge.
The main challenge that is posed by BonJour’s argument is to identify a non-

entailing threshold that gives knowledge an importance not had by any rela-
tion to a proposition that is less well justified.
BonJour addresses just one substantial theoretical view of what a non-
entailing threshold for knowledge might be. Before we look at that view and
BonJour’s criticism of it, we should consider a preliminary comment by Bon-
Jour about the problem. He laments that accounts of the justification required
for knowledge rarely say more about the justification threshold than to call the
level “strong” or “high” (2010: 60–61). One reason he finds this problematic is
that “… nothing this vague is enough…”, given the “exalted cognitive state in
question” (2010: 60).
This purported problem is doubly problematic. First, it places knowledge
on too high a pedestal. What BonJour takes to be “exalted” about knowledge is
that it is “supremely valuable” and “the epistemic summum bonum” (2010: 58).
BonJour does not defend this supremacy of knowledge. It seems plain that the
highest epistemic status belongs, not to knowledge, but rather to certainty. A
proposition that is a certainty for someone is in a better epistemic condition
for the person than is a proposition that is known but not a certainty for the
person.8
The other difficulty for BonJour’s preliminary comment is his thought that
the vagueness of terms like “strong” and “high” makes the terminology inad-
equate to the task of specifying the knowledge-level justification threshold.
Rather than being inadequate, a vague specification may be required. Knowl-
edge may be indeterminate in the same way as is the meaning of some vague
terminology about the needed strength of justification. If so, then to say any-
thing more precise about the threshold would be to say something that the
status of knowledge does not determine.9
8 The comparison of knowledge to certainty is further discussed later in this section. A po-
tential for relational properties like knowing to be indeterminate is further discussed in
Section 5.
9 One version of this view is advocated below.
If the characterizations “strong justification” or “highly likely” are offered as

a way of specifying the threshold, then the specification is objectionable on
grounds other than vagueness. A more telling objection is that these condi-
tions are not demanding enough. Knowledge requires better justification than
that. For example, suppose that we are asked, “How do you know that Bob
attended the party?” Suppose that the following is our best answer: “We have
strong justification for thinking that he was there and it is highly likely.” Here
is a negative response to that answer: “You have good reason to think that Bob
was there but you really don’t know it for a fact.” That response seems quite
correct. If we have nothing better to go on than strong justification or high
probability, then we have inadequate support to know that the proposition is
true. So the threshold is not identified by any such description.
Next let’s consider BonJour’s discussion of one philosopher’s attempt to
characterize the threshold. He addresses Roderick Chisholm’s work on this.
BonJour writes:
Perhaps the only philosopher whom might be viewed as having seriously

tried to define the degree of justification required by the correct realiza-
tion of the weak conception is Roderick Chisholm, in the three editions
of his Theory of Knowledge (Englewood Cliffs, N.J.: Prentice-Hall, 1966,
1977, 1989), together with other works written in the same period. Ch-
isholm’s final account, in the 3rd Edition, is that a proposition p is evident
(his term for the level of justification required for knowledge) for a sub-
ject S just in case “For every proposition q, believing p is at least as justi-
fied for S as is withholding q [that is, suspending judgment with respect
to q].” But after all of Chisholm’s rather painstaking discussion in these
works, neither what the resulting level of justification really amounts to
nor, more importantly, the rationale for that level being the right one is
at all clear.
BonJour 2010: 79
This passage poses two objections to Chisholm’s 3rd edition account of

knowledge-level justification. The first is obscure. BonJour says that it is un-
clear what the justification level “amounts to.” It is unclear what clarity Bon-
Jour thinks is missing in Chisholm’s account. The account requires our having
this much justification: Our taking an attitude of suspending judgment toward
some proposition that is from our perspective manifestly neither reasonable
to believe nor unreasonable to believe is no less justified than is our believ-
ing a proposition at the threshold level. In short, the knowledge level is a
justification for belief that is at least as strong as the justification to suspend
132 Conee
judgment on a proposition concerning which our evidence is quite neutral.

When we are in that evidential situation, our suspending judgment on such a
proposition seems to be exactly the justified attitude for us to take. So the ac-
count’s requirement that we be at least as well justified in believing when we
know seems reasonably clear and very demanding. At a minimum, BonJour’s
difficulty about what that condition “amounts to” needs some further exposi-
tion, if it is to raise a serious problem for Chisholm’s proposal.
BonJour’s other objection to Chisholm’s account is more forceful. He asks
why the proposed level of justification is the knowledge level. That is a good
question. Chisholm’s 3rd edition proposal itself does not give any clue about
why that much justification is enough. Given Chisholm’s characteristic meth-
ods, he would defend the proposal by its success in classifying examples. In any
event, another of Chisholm’s proposals is better suited to answering BonJour’s
good question. Chisholm’s 2nd edition definition of knowledge-level justifica-
tion suggests a plausible answer to the question of why it is the threshold. Here
is the 2nd edition definition (with the defined terms that Chisholm uses re-
placed by his definitions of them):
CD2 h is evident for S =df. (i) accepting h is more reasonable for S than is
withholding h, and (ii) for every i, if accepting i is more reasonable for S
than accepting h, then there is no j such that accepting j is more reason-
able for S than accepting i.
chisholm 1977: 7–12
CD2 says roughly that having knowledge-level justification requires having jus-
tification that is at least the next best thing to the best justification that we can
have. This suggests the following answer to the question of why it is the thresh-
old. Knowledge is intuitively a very high epistemic status. Meeting the CD2
conditions accords with this by requiring a level of justification that is so good
that it is not susceptible to any improvement, unless the improvement moves
the justification to the highest level of all. It might be wondered why knowledge
does not require the highest level. Chisholm in effect answers that question by
offering examples of knowledge that have justification below the highest level.
The examples show a contrast between our justification for known facts of em-
pirical observation and our memories of them with our better justification for
known facts of simple arithmetic. This contrast argues that knowledge does
not require the highest level. In Chisholm’s view, unsurpassed justification is a
requirement for certainty rather than knowledge (Chisholm 1977: 11).
If we concur that some knowledge is less justified than are certainties, then
a level of justification that is the closest thing to certainty can seem just right
to require of knowledge. Given that there are known propositions for which we
cannot have certainty, having justification at the closest strength to certainty
is having justification at a level where further evidence is superfluous – further
evidence makes the proposition no better justified.
This superfluity of additional evidence seems to establish a suitably high
plateau for the strength of justification that is required by knowledge. If there
is some such level, then it is the best we can do for external world beliefs. None
of them can be as well justified for us as are some other propositions, such as
the proposition that something or other exists. If we combine this point about
the unavailability of certainty about the external world with our evidence that
we do have external world knowledge, then we have a case for the conclusion
that the CD2 account is sufficiently demanding. The next-best-thing account
backs off as little as possible to a definite level, while allowing us external world
knowledge. That amounts to an impressive case for the threshold.
CD2 is problematic though. It is doubtful that there is any such next-best
justificatory plateau.10 Examples make it reasonable to think that our justifica-
tion for known external world beliefs can be bettered. For example, for many
of us our evidence makes it is just barely coherent to suppose that recently we
have been deceptively envatted while furniture was abolished. Our evidence
makes it even less credible that either some furniture exists or some other ex-
ternal things exist. Yet we do know each disjunct. We have ample but narrower
evidence for the furniture proposition. Our evidence for the disjunction is bet-
ter support. We have two partially independent and partly mutually support-
ing lines of evidence for the furniture-or-other-external-things proposition. Yet
the latter proposition is no certainty for us either. It is not as well justified for
us as is the proposition that something exists.
If that sort of consideration is correct, then we can know propositions such
as that some furniture exists without their being evident to us according to
CD2. They are not next best things to a certainty for us. If so, then CD2 does not
solve the threshold problem for the justification that is needed for knowledge.
The criminal standard that Richard Feldman and I have advocated gives a
defensible solution to the threshold problem. This standard requires the level
of justification that is described by the legal test for conviction in a criminal
trial: proof beyond a reasonable doubt. A proposition for which we have jus-
tification that is at least that strong is well enough justified for us to know it.
Four explanatory comments are immediately in order. First, “proof” here
means what “proof” usually means. The meaning does not require proof by
formal demonstration. It is easy to think of other sorts of proof. Suppose
10 This point was made by Wayne Wasserman in Wasserman (1980).

134 Conee
that we wonder whether Smith attended the birthday party. Someone says, “I
have proof that he did. Here he is in a video of the cake-cutting.” We readily
understand how such a response is often true. Proof takes powerful evidence
that is conclusive in the absence of truly extraordinary exceptions. Here are
a couple of common types of proof that are frequently available to us. First,
clear perception of familiar facts is proof for us of what we perceive to be the
case. Q uestion: “What proof do you have that a table is in the room?” Answer:
“I see with my own eyes that one is there.” And second, our receipt of utterly
trustworthy testimony is proof for us of its content. Question: “What proof do
you have that the road was closed?” Answer: “Jones, that paragon of integrity,
told me so.” Having proof like this is common enough.
Here is a second comment about the criminal standard. In order for us to
have proof of a proposition that is beyond a reasonable doubt for us, we must
have no defeaters of the proving evidence. That is, our evidence must leave
us with no standing good reason to doubt what it supports (no “rebutting de-
feater”) and no standing good reason to doubt the efficacy of the evidence as
proof (no “undercutting defeater”).
Third, the phase “proof beyond a reasonable doubt” is somewhat vague.
The vagueness seems to be matched by the vagueness of “knowledge.” That is,
careful knowledge attributions seem difficult to make, and perhaps indetermi-
nate, whenever it is on reflection difficult to decide, and perhaps indeterminate,
whether the phrase “proof beyond a reasonable doubt” applies. If that matching
does occur, then it is an asset of the criminal standard. It supports the thought
that our careful knowledge attributions rely on meeting the standard. That re-
liance in turn supports the present view that knowledge requires meeting the
standard.
And fourth, the criminal standard avoids the problem for CD2 that no next-
to-the-best justificatory plateau seems to exist. The criminal standard is com-
patible with there being no highest level of justification for knowledge that is
short of certainty. Having doubt-free proof is compatible with gaining more
evidence that improves the support. Examples seem to bear this out. By seeing
and feeling the raindrops we can have proof beyond a reasonable doubt that
it is raining. Still, we can get better evidence of the rain by additionally hav-
ing the testimony of others, and yet better evidence by adding weather radar
displays, and so on. The perception and memory examples of external world
proof recently mentioned illustrate that the criminal standard does not require
entailing justification. Such proof misleads in skeptical scenarios. Still, such
fallibly justified perceptual and memory beliefs often are actually proven to
us beyond a reasonable doubt. The criminal standard makes available the fol-
lowing answer to BonJour’s question of why any non-entailing condition sets
the justification threshold for knowledge. The standard sets a non-arbitrary
threshold for knowledge because a central fact about knowledge is that know-
ing a proposition makes it entirely reasonable not to wonder about the propo-
sition’s truth.11 Meeting the criminal standard does that, and no lesser evidence
does so. Wondering about truth can quite reasonably cease at a point when we
have justification of a strength that proves the proposition to us and we have
no reasonable doubt about the proposition or its support.
As a response to the reconstruction of BonJour’s argument, bsa, this defense
of the criminal standard amounts to an objection to its second premise. Pr2
says that any threshold allowing fallible justification is too arbitrary to account
for the importance of knowledge. That is a mistake. Fallible justification that
meets the criminal standard is sufficient to do the important work of giving us
the basis that knowledge affords for intellectual contentment. The threshold is
thus not arbitrary.12
A philosopher like BonJour who doubts that any fallible justification is suf-
ficient for knowledge might object as follows:
Proof beyond a reasonable doubt would be enough justification for

knowledge. But fallible evidence allows the possibility of error. Evi-
dence like that allows a reasonable doubt. The doubt is that very fact of
possible error – the fact that it is possible for the evidence to exist while the
proposition is untrue. That possibility is a good reason to doubt that
the proposition is true.
This objection has a straightforward answer. Reason to doubt that a propo-

sition is true is reason to think that it is untrue or reason to think that our
evidence is defective. The possibility of a proposition having been untrue,
while we had our actual evidence, is not a reason to think it is actually untrue
or that its support is flawed.
The criminal standard sets a high threshold. In my view, reasons to doubt
include the existence of a real chance for us that the proposition is untrue. It
seems clear that a claim of this form cannot be true: “Although there is a real
chance that I am mistaken that X is true, I know that X is true.” The existence of
this chance of error depends on our evidence about when there is actual error.
Specifically, we have a real chance of being in error about a proposition when
our evidence for the proposition is no better support for its truth than our evi-
dence tells us is the support that someone actually has for some falsehood.
11 Conee (2017: 311–331).

12 The value of knowledge is further discussed in “Good to Know,” ibid. As the paper discuss-
es, the importance of knowledge to us includes the assets of meeting other conditions as
well.
136 Conee
For example, we have this sort of real chance of error about each losing tick-
et in a fair future lottery. This evidence is evidentially indiscernible by us from
our evidence supporting the proposition that asserts the loss of the winning
ticket. That evidential status gives each loss proposition a real chance for us of
being untrue. In my view this is why none of the losing propositions is known.
For many fallibly justified beliefs, however, no such chance of error exists.
One randomly selected easy illustration: We have excellent evidence for the
proposition that some sand exists. We have no evidence that anyone’s eviden-
tial situation is ever just good as it is for us in all evidentially significant re-
spects for some proposition that is actually untrue. Propositions like that are
proven to us beyond a reasonable doubt.13
Knowledge requires rationally conclusive evidence. The criminal standard
is that demanding. We must have proof and we must have no good reason to
doubt that it succeeds. When we have that, our rational interest in the truth of
the matter is fully satisfied. Often our fallible evidence is rationally conclusive
in this way. It is entirely reasonable to think that this occurs whenever it is quite
clear to us that a proposition about our environment is true. The proposed
threshold is thus not too stringent to allow us fallibly justified knowledge.
3 Piecemeal Deceptions
Martin Smith (2016) offers skeptical arguments against our knowing external
world propositions, taking them on one at a time. Smith proposes tailor-made
local skeptical hypotheses. They are argued to be exempt from the evidence
that we might be thought to have against them. Smith contends that the skep-
tical hypotheses prevent knowledge because we have no evidence against
them. It will be argued in response that we do have such evidence.
13 The two claims just discussed – that knowledge requires no real chance of error and that
evidence allows a real chance of error under the stated conditions – are independent of
the success of the criminal standard as a solution to the threshold problem. The claims
raise issues that are not essential to the minimal success of the criminal standard as
the knowledge justification threshold. The claims might be thought to set problematic
constraints on the extent of our knowledge. (They may well imply our ignorance of the
location of our currently unseen publicly parked cars, to mention one well-known sort
of example from Jonathan Vogel.) As long as the claims allow that some non-entailing
evidence meets the criminal standard, as the sand example is intended to illustrate, they
allow the criminal standard to meet BonJour’s basic challenge.
Here is an example of how the skeptical reasoning goes. Suppose that I

believe the following proposition while looking into a zoo enclosure labeled
“Zebras” and seeing animals that appear to me as my background information
leads me to think that zebras appear:
P1: The animals before me are zebras.
The general skeptical idea is that a specifically targeted skeptical hypothesis

denies the truth of an ordinary external world belief like P1. The hypothesis al-
leges that I am in an extraordinary situation in which the belief is untrue. Here
is such a hypothesis for P1:
[SH1]: Although zoos are generally trustworthy and reliable and it

wouldn’t usually be in their interests to deceive the public, the zoo that
I’m visiting is an exception to these rules and the animals before me are
mules cleverly disguised by the zoo authorities to look like zebras.
smith 2016: 49
Before we address Smith’s supplemental argument that I lack evidence against

SH1, we should see the rest of the skeptical argument using SH1. So let us as-
sume for now that I lack evidence against SH1.
Smith completes the argument by offering two principles as premises. The
first principle asserts that if a hypothesis, H, is a metaphysical possibility, then
someone, S, is in a position to know that H does not obtain only if S has evi-
dence against H:
nkwe: If H is a possible hypothesis and S has no evidence against H, then

S is in no position to know ~H (2016: 45).
The other premise in Smith’s skeptical argument is a closure principle. It asserts

that we are in a position to know any deductive consequence of s omething
that we know.
CLOSURE: If P deductively entails Q and S knows P, then S is in a posi-

tion to know Q (2016: 45).14
14 CLOSURE as it stands is doubtful. Someone might be incapable of deducing some entail-

ment of something the person knows, or incapable of understanding the entailed propo-
sition. Given either incapacity, knowing the entailing proposition seems not to place the
138 Conee
With these premises, the argument against my knowing P1 goes as follows.

Suppose for reductio that I do know P1. If so, then I can deduce that SH1 is false.
SH1 says, among other things, that the animals before me are not zebras and
P1 implies the falsity of that. By closure this places me in a position to know
~SH1. We are assuming that I have no evidence against SH1. Given that I lack
such evidence, by nkwe I am in no position to know ~SH1. Thus the reductio
assumption that I know P1 has contradictory implications about what I am in
a position to know. So I do not know P1 (2016, 45).
As Smith observes, similar piecemeal arguments could equally well refute
our knowledge of each of the ordinary propositions that we think that we
know on the basis of non-entailing evidence. Our situation concerning each
such proposition allows that some extraordinary exception makes our non-
entailing evidence misleading (2016: 50–51).
Smith concedes that SH1 might strike us as a priori unlikely, and he sym-
pathizes. He points out that nkwe requires our having evidence against a
hypothesis in order for us to be in a position to know its negation (2016: 49).
Smith does not further pursue the question of how a priori likelihood consid-
erations affect the application of nkwe. Perhaps he thinks that any impression
of a priori unlikelihood that we might have is not evidence.
But a priori impressions are evidence. In particular, suppose that I reflect on
SH1. My understanding of it indicates to me that SH1 is unlikely. I see that in
effect SH1 asserts that I am witnessing a gratuitous exception to a tendency. I
see that this is intrinsically unlikely. This reflective indication of improbability
is some a priori evidence against SH1. So at least if I do engage in reflections
with that outcome, then nkwe does not apply to me concerning SH1. Yet
the skeptical argument just described relies on the application of the prin-
ciple nkwe. The principle applies only when I lack any evidence against the
skeptical hypothesis. So my a priori evidence against SH1 makes the skeptical
argument fail.
Smith has a further argument that can be modified to avoid this objection.
The crucial difference is the presence in the further argument of an enhanced
skeptical hypothesis that includes all of the relevant evidence. Smith’s new
skeptical hypothesis is a conjunction. One conjunct, me, asserts my evidence
for P1, from perception, background information, a priori reflection, and what-
ever else. The other conjunct asserts the negation of P1:
person in a position to know the entailment. But an otherwise similar skeptical argument
would remain if we narrowed CLOSURE’s application to people who can deduce and un-
derstand the entailment, and who meet any further requirements for being in a position
to know the entailment. So we can set aside this sort of problem for CLOSURE.
SH2: ME&~P1
The other novelty in the new skeptical argument is a principle about evidence.
It says that the entailments of a proposition are not evidence against it:
enea: If a hypothesis H deductively entails a proposition P, then P is not

evidence against H (2016: 50).
enea has initial plausibility. As Smith points out, the truth of each of a prop-
osition’s entailments contributes to the whole proposition being true (2016:
50). Thus a proposition’s entailments seem to confirm it rather than to give
evidence against it.
The problem of a priori evidence against the rest of my evidence for P1 in
conjunction with ~P1 is clearly avoided by SH2, with the help of enea. Any a
priori evidence I might have against that conjunction is conjoined in me. The
evidence is therefore an entailment of SH2. So by enea that evidence is not
evidence against SH2.
What about the a priori unlikelihood of SH2 itself? SH2 in effect says that P1
does not hold, despite all of the evidence for P1 in me, which includes the a priori
unlikelihood of P1 not holding when the rest of the evidence for P1 obtains. A
priori reflection would indicate that SH2 is at least as unlikely as SH1, since SH2
asserts my a priori evidence against SH1 and SH2 also asserts what SH1 asserts.
SH2’s reflective unlikelihood is evidence that I can have against SH2.
We can modify SH2 to avoid this problem. We can conjoin within an SH
all of my a priori evidence for the unlikelihood of the SH itself. Specifically,
we can add to ME in SH2 whatever reflective evidence is available to me con-
cerning the unlikelihood of SH2 – call this whole conjunction ME*. We can
further conjoin any a priori evidence I might have that ME* is unlikely in con-
junction with both ~P1 and any further a priori evidence I might have of the
unlikelihood of the new conjunctive proposition itself – call the latter a priori
evidence A*. Now we have the new skeptical hypothesis:
SH2*: ME*&A*&~P1.
SH2* is somewhat complicated. But the idea is simple enough. SH2* asserts
ME*, which is the conjunction of ME, the totality of my evidence for P1, and
my a priori evidence that this totality is unlikely in combination with ~P1.
SH2* also asserts A*, my a priori evidence of the unlikelihood of SH2* itself.
And SH2* also asserts that ~P1 holds. Thus, conjoined in SH2* is both the evi-
dence that I have for P1, and the evidence I have against my evidence for P1
holding while ~P1 obtains. Since that is all entailed by SH2*, enea implies
140 Conee
that none of it is evidence against SH2*. Yet that evidence seems to be all that
I have against SH2*. If so, then I have no evidence against SH2*. nkwe and
CLOSURE tell us that I must have some such evidence to know P1. It follows
that I do not know P1.
Similar piecemeal skeptical arguments apply equally well to our other exter-
nal world beliefs. The result is a defense of an extraordinary skepticism.
There are two main problems for the argument. The first is that nothing in
the premises denies that we have our knowledge as evidence. So suppose that
our knowledge is evidence and suppose that I know P1. Then I have P1 itself as
evidence against SH2*. The argument that I have no evidence against SH2* ap-
plies only to the evidence that I have for P1. My evidence for P1 is conjoined in
SH2*. So by enea my evidence for P1 is not evidence against SH2*. But if I do
know P1, then P1 itself is something else that I have as evidence. P1 clearly goes
against SH2* by contradicting its third conjunct.
Is this use of P1 viciously circular? No. Invoking P1 as evidence does not
require just assuming that I know P1 in order then to claim that I have P1 as
evidence because I know it. We have strong reasons that do not derive from P1
for thinking that we know propositions like P1. Our justification for such prop-
ositions from our perceptual and memory evidence is quite good and we have
no reason to mistrust any of it. It is quite reasonable for us to think that we
thereby know such propositions. So as it stands the skeptical argument fails.
No reason has been given to deny that my knowledge of P1 is evidence that I
have against the skeptical hypothesis.
The argument can be further revised to avoid this problem. Our ordinary
knowledge can be excluded from the evidence that we need against the skepti-
cal hypotheses.
nkwe*: If H is a possible hypothesis and S has no evidence against H

from the fundamental evidence that S has for external world propositions,
then S is not in a position to know ~H.
The new idea in nkwe* is that in order to know external world propositions,
the most basic evidence we have that might enable us to know them must
itself discredit their rivals. In other words, we need excluding evidence against
any skeptical hypothesis from our fundamental external world evidence, our
non-derivative evidence on the basis of which we might know external world
propositions. Known external world propositions like P1 are not fundamental
evidence. They qualify as evidence by being known on the basis of other evi-
dence from perception and memory. So nkwe* excludes propositions like P1
from qualifying as the requisite evidence against a skeptical hypothesis.
This exclusion is questionable. Our knowledge is some of our evidence. What

makes that evidence ineligible for doing some work that evidence can do?
We can set aside that doubt about the restriction of nkwe to nkwe*. The
nkwe* claim that our fundamental evidence must favor the truth of what we
know over all incompatible rivals is plausible enough to make it part of an
interesting argument.
With nkwe* assumed, the application of enea to SH2* does what it must
do for the skeptical argument. The application of enea excludes my non-
derivative evidence for P1 as evidence against SH2*, because that evidence
is entailed by SH2*. So enea implies that none of it is evidence that I have
against SH2*. Yet that is all of the evidence that could keep nkwe* from im-
plying that I am not in a position to know ~SH2*. It follows by nkwe* and
closure that I do not know P1. This argument too can be replicated to target
any of our putative external world knowledge.
Another problem with this sort of skeptical argument is intractable. enea is
faulty. The problem is irreparable for the skeptical purposes of the argument.
Note first that as enea is stated, it excludes from being evidence implications
of some particularly powerful reductio arguments. Suppose that a proposition
is shown to entail a contradiction with no other non-logical assumptions. If
so, then the proposition is reduced to absurdity and decisively refuted. The
fact that a proposition entails a contradiction is extremely strong evidence
against the entailing proposition. Yet according to enea entailing a contra-
diction is no evidence against the entailing proposition, simply because the
proposition does entail the contradiction and consequently entails that it en-
tails a contradiction.
enea can be restricted to avoid this problem while doing its work in the
skeptical arguments. Possibilities are all that the skeptical arguments need. So
enea can be restricted to metaphysically possible hypotheses, thus avoiding
ones that entail contradictions.
Still, the existence of refutations of propositions by reductio reasoning
points to an error in the support offered for enea. Each entailment of a prop-
osition would help to make the proposition true, if it were true. It does not
follow that an entailment is on balance evidence for the proposition. We can
allow that having as evidence any proposition that is a known entailment of P
is having some evidence that P is true. As a contradictory implication shows,
an entailment may also be better evidence that P is untrue.15
15 In Jonathan Vogel’s (2014) highly valuable paper, “E & ~H” he offers convincing examples
of propositions that entail outweighing inductive evidence against themselves. See espe-
cially Vogel (2014: 92).
142 Conee
An example of this sort of better evidence occurs when I have as evidence

the following conjunction. The first conjunct is ME, the proposition assert-
ing my evidence in favor of P1, the proposition that the animals before me are
zebras. The other conjunct is P1 itself. Having as evidence this conjunction,
ME&P1, can be granted to be having some evidence for SH2 (ME&~P1) in vir-
tue of SH2 entailing ME. Yet having as evidence ME&P1 is also part of having
conclusive evidence against SH2, in virtue of ME&P1 entailing the negation of
the second conjunct of SH2, ~P1. (The other part of the evidence against SH2
is an awareness of the entailment.)
One especially pertinent way in which a proposition’s entailment may give
evidence that counts on balance against the entailing proposition is for the
entailment to give better reason to doubt that some of the proposition’s other
entailments are true than it is reason to believe the whole conjunction. The
currently relevant way for this to happen is as follows. The truth of one con-
junct of a possibly true conjunction can make it doubtful that another c onjunct
is true, because the one argues a priori against the other. Although the whole
conjunction could be true, the truth of one conjunct, one that makes its joint
truth with another “almost impossible,” makes the whole thing “almost impos-
sible.” More literally, having as evidence one of a conjunction’s conjuncts gives
excellent reason to think that some other conjunct is untrue and thus that the
whole conjunction does not hold.
Here is a simple schematic illustration. Suppose X1 is the conjunction of First
Conjunct, FC, and Second Conjunct, SC. U is the true proposition that the con-
junction of FC and SC is unlikely. Clearly and uncontroversially, anyone who
has U as evidence, say, by knowing U to be true, has evidence against X1, since
U tells the person that X1 is unlikely to be true. Suppose that X2 is the conjunc-
tion of FC, SC, and U. Just as clearly, anyone who has U as evidence also has
that same evidence against X2, since X1 is manifestly part of what must be true
for X2 to be true. It does not matter that U is logically implied by X2, but not by
X1. From the person’s perspective, X2 is at least as unlikely to be true, because
of the unlikelihood of FC and SC. X2 also asserts U. That assertion clearly does
not weaken this reason to think that X2 is untrue. U is equally good evidence
for the falsehood of any conjunction that includes FC and SC.16 This includes
X2. Yet enea disallows U from being evidence against X2. That is a mistake.
16 The evidence from U against the conjunction FC&SC can be overridden. For instance,
we might also have evidence that U is quite misleading about FC&SC. U is still evidence
against that conjunction. Overridden evidence for something is still evidence for it. In any
case, the conjoined evidence in the SHs is not overridden.
Smith writes in defense of enea:
Furthermore, it’s natural to think that P could only provide evidence

against H if the evidential probability of H given P were lower than the
prior evidential probability of H—that is, if Pr(H | P) < Pr(H). But, if Pr
is a classical probability function, we can prove that, if Pr(P) > 0 and H
entails P then Pr(H | P) ≥ Pr(H) from which [enea] will follow.
smith 2016: 50
What is said here to be natural to think is that a proposition must lower the
prior probability of another in order to be evidence against it. This thought,
however natural, is incorrect. Lowering prior probability is not the only way for
a proposition to provide opposing evidence. We have seen another way for E to
be evidence for us against H. It might be that E asserts the fact that makes H’s
prior probability low. In such a case E is evidence that we can have against H.
H might have undefined probability for us before E makes its probability low
for us. It might be that H is a priori unlikely for us precisely because as soon
as we understand what H says we have as evidence the fact asserted by E, that
H is improbable. So this low probability is as “prior” a probability for us as H
can have. As we have seen, one way for E to do this rendering low of the prior
probability for us of H is for E to report the low probability of the other con-
juncts of H occurring together. If so, then having E as evidence is having good
evidence against H. All this is true whether or not E is a conjunct in H. enea
errs by denying that.
Finally, here are a few words about how the evidentialist anti-skeptical case
that Richard Feldman and I have advocated applies to the present skeptical
arguments. The basic thought is that evidence justifies adequately for knowl-
edge when it meets the criminal standard, that is, the evidence is proof beyond
a reasonable doubt. We can have such evidence against skeptical hypotheses
like SH1-SH2*. The fundamental evidence that we can have against the SHs
is a combination of two considerations. First we have our evidence for exter-
nal world propositions like P1. That is proof beyond a reasonable doubt for
us of those propositions. The SHs give us no reason to doubt the external
world propositions, because we have no reason to think that the exceptional
circumstances that the hypotheses describe obtain. The circumstances are
possible, but we have no evidence that they are actual. Second, by reflection
we can have evidence that propositions like P1 exclude skeptical hypotheses
like SH1 – SH2*. Thus we have a modus ponens argument as proof beyond a
reasonable doubt of the negations of the SHs. Our evidence often meets the
144 Conee
criminal standard both for external world propositions like P1 and for the nega-
tions of such skeptical hypotheses.
4 Gettier Skepticism
Baron Reed (2007, 2008) gives two closely related arguments for skeptical
conclusions. The arguments allege that all fallibly justified true beliefs have
a feature that keeps them from being knowledge. The feature makes the be-
liefs relevantly like Gettier cases. It will be argued in response that what blocks
knowledge in the Gettier problem cases does not block knowledge of the rest
of our fallibly justified true beliefs.
4.1 Epistemic Performance

In setting up one of his arguments Reed describes two pairs of examples. The
first example in each pair is about a belief by someone whose name begins
with “B” – these cases are “the Bs.” In the Bs the person seems straightforwardly
to have knowledge. The second example of each pair features a “C”-initialed
person – these cases are “the Cs.” The situation in a C is virtually the same as
the situation in a B, except that the Cs are Gettier cases. Here is one of Reed’s
pairs of a B and a C:
Car Possession 1: Bartholomew has a lot of very good evidence for the
proposition that his friend, Smith, owns a Ford. He has ridden in Smith’s
Ford numerous times in the past year, has heard Smith talk about his Ford
regularly, and has even seen the title for the car in Smith’s name. Moreover,
Smith has never discussed any plans to sell or otherwise get rid of the car.
Bartholomew has also recently taken a basic symbolic logic course, and
he recognizes that the disjunction introduction rule allows him to derive
a true complex proposition by disjoining two propositions where at least
one of them is true. So, even though he has no reason to believe that it is
now snowing in Albuquerque, though in fact it is, he infers (and forms the
belief) that Smith owns a Ford or it is now snowing in Albuquerque.
Car Possession 2: Connor has a lot of very good evidence for the
proposition that his friend, Lee, owns a Honda. Connor has ridden in
Lee’s Honda numerous times in the past year, has heard Lee talk about
his Honda regularly, and has even seen the title for the car in Lee’s name.
However, Lee has very recently sold his car. Connor also has taken a basic
symbolic logic course and understands how the disjunction introduc-
tion rule works. Even though he has no reason to believe that it is now
snowing in Albuquerque, though in fact it is, he adds this as a disjunct to

his belief that Lee owns a Honda. Consequently, Connor forms the belief
that Lee owns a Honda or it is now snowing in Albuquerque.
reed 2008: 95–96
Reed comments that in the B case Bartholomew seems to have knowledge. In

the C case Connor seems not to have knowledge. Reed says that
… although [the C] beliefs are true and – it is important to note – just as

well justified as those of the B subjects, the C subjects have beliefs that
appear to be true by accident. Connor’s complex belief was true because
the disjunct he added randomly to his prior false belief happened to be
true, though he had no reason to think that it was… Knowledge is thought
to be incompatible with this kind of luck – it cannot be a mere accident
that one’s belief happens to be true rather than false.
reed 2008: 96
Reed then gives the skeptical argument that he derives from a comparison of
the Bs to the Cs:
Although this is the standard way of reading these cases, it has not
been appreciated that they permit a comparison much like the
one that u nderlies traditional skepticism. Here is how the new argu-
ment for s kepticism works: C does not have knowledge. B is just like C in
all epistemic respects. Therefore, B does not have knowledge, either.
(2008: 96)
This is Reed’s first skeptical argument:
RSA1
Pr1: In the Cs the people have a justified true belief, but they do not
know it.
Pr2: The Bs are just like the Cs in all epistemic respects.
C: In the Bs the people do not know their justified true beliefs.
To appreciate the force of this reasoning we need further details. The second
premise asserts that the Bs and the Cs are epistemically the same. This claim
conflicts with the appearance that the B people know and the C people do not
know. To understand why the Bs and Cs are supposed to be epistemically alike,
it helps to learn what, according to Reed, prevents the C people from know-
ing. To support Pr1 Reed says, “[V]irtually everyone agrees that knowledge is
146 Conee
incompatible with accidental truth” (2008: 96). Thus, Reed understands the
lack of k nowledge by the C people to result from certain accidents concerning
the truth of their beliefs.
Reed relates the accidents to what he calls the “epistemic performance”
of the C people. He describes this performance in our example pair by not-
ing, “Connor’s belief is grounded in exactly the same sort of evidence as Bar-
tholomew’s” (2008: 97). The evidential basis of a subject’s belief is part of what
Reed includes in the subject’s epistemic performance. Reed also observes that
both the reliability and the environments in which the belief-forming process-
es are occurring are as favorable for the C people as for the B people. These
shared external features lead Reed to claim that the epistemic performances
are alike, “even when we conceive of those performances in the most broadly
externalistic way possible” (2008: 97). Reed concludes the case for the Bs and
Cs being epistemically the same by relating their epistemic performances to
the accident-preventing condition that he calls “condition x.” This condition is
defined as “…what distinguishes all cases of non-accidentally true belief from
all cases of accidentally true belief” (2008: 97). Reed asserts that the satisfac-
tion of condition x “… stands entirely outside of the subject’s performance. It
is not merely external to the subject’s subjective awareness, it is also extrinsic
to her epistemic performance” (2008: 97).
Thus, the term “epistemic performance” is used by Reed to describe the way
that a belief is epistemically justified, including both any internal factors such
as mentally possessed evidence and any external factors such as belief-forming
process reliability. At least those are the factors that Reed cites in defending
the sameness of the epistemic performances of the B people and the C people.
When epistemic performance is understood in this way, it should be agreed that
the epistemic performances of the B people are the same as those of the C peo-
ple. It should also be agreed that the satisfaction of the Gettier-case-preventing
condition x is “extrinsic” to the subject’s epistemic performance, understanding
this to mean that the satisfaction of condition x is neither entailed nor causally
necessitated by the subject’s epistemic performance. The same justification-
acquiring performance under various metaphysically and causally possible
deceptive circumstances, such as those of the Cs, does not meet condition x.
These epistemic commonalities between the Bs and the Cs do not establish
that the Bs are epistemically just like the Cs. The epistemic includes whatever
especially pertains to knowledge. So the epistemic aspects of a subject, con-
cerning a proposition, include any aspects of the subject that especially pertain
to knowledge of the proposition. Satisfying condition x always especially per-
tains to knowledge by a subject of a proposition, since it is satisfying condi-
tion x that excludes the accidents that prevent knowledge in Gettier cases.
The B people satisfy condition x and the C people do not. So the Bs are not
e pistemically just like the Cs. This epistemic difference refutes Pr2, the second
premise of the skeptical argument rsa1. Pr2 asserts that the performances are
epistemically the same.
We have seen no good reason to deny that the condition x is frequently met.
When condition x is met by an adequately justified true belief, the proposition
is known.17 At least, we have seen no good reason to think that any further
condition is required. It could be claimed that there is another kind of accident
that does not make for a Gettier case but prevents knowledge in some other
way. But we have seen no example of such an accident and no argument that
it exists.
4.2 Attributability
Reed’s other version of is skeptical argument is broadly similar to the one just
discussed. The terminology and details differ. Again the general idea is that fal-
libly justified beliefs are all subject to Gettier-like accidents and this prevents
the beliefs from being knowledge.
The second argument has two new premises. They are principles that Reed
calls “Fallibilism” and “Attributabilism.”
Fallibilism: For any proposition p that S knows on the basis of justifica-

tion j, S’s belief that p could have been either (i) false or (ii) accidentally
true, even while being held on the basis of j.
reed 2007: 237
A footnote clarifies that Reed intends the Fallibilism principle to allow that
some beliefs cannot be false given their justifications. Reed mentions cogito
propositions and surely he would add beliefs in necessary truths. Reed intends
to argue for skepticism only concerning beliefs the justification for which is
compatible with their falsehood (2007: 237 n.2). It seems safe to assume that
Reed would classify all external world beliefs in this category, since he gives no
indication of an exception of that sort. Another footnote clarifies that the sense
of “accidentally true” in Fallibilism is the accidental sort of relation between a
17 The term “adequately” is added here to the usual “justified true belief” to exclude lottery
loser beliefs. One sort of true belief that can be probabilistically justified is a belief that
some particular one among a great number of equally eligible lottery tickets is not the
winner. It is a justified true belief that is not a Gettier case. Despite this, the belief is not
knowledge. It is not adequately justified for knowledge. It does not meet the criminal
standard because it is subject to a reasonable doubt. The doubt is the chance on one’s
evidence that the ticket is not a loser. So the lottery beliefs are excluded by the require-
ment for knowledge of adequately justified true beliefs, while condition x excludes Get-
tier cases.
148 Conee
justified true belief and its truth that exists in Gettier problem cases (2007: 237,
n.4). So Reed intends Fallibilism to say this:
Fallibilism*: Any external world proposition is known only if, for one
thing, its justification allows that it is false18 and, for another thing, it
could have been true with the same justification while being accidentally
true so as not to be known.
Here is Reed’s other premise for this skeptical argument:
Attributabilism: For any state of knowledge k that person S possesses, the

possession of k is attributable to S.
reed 2007: 238
Reed discusses at length what he means by “attributable” here and why he

thinks that Attributabilism is true. Concerning the meaning, Reed’s first and
most helpful elaboration is an analogy with actions. His idea is that knowing
a proposition requires that the knowledge “belongs to” its subject as does an
intentional action, in contrast to a reflexive behavior (2007: 238–239).
It further clarifies Attributabilism to note how Reed takes some philosophi-
cal views to commit themselves to Attributabilism. The commitment is taken
to occur when a view implies that knowledge is an achievement and also when a
view implies that knowledge gives credit to the one who knows (2007: 239, 242).
Reed’s notion of attributability may not yet be fully clear. We need not grasp
the precise notion though, if we take for granted a conclusion about Attributa-
bilism for which Reed argues. Here is the conclusion:
To reject or restrict Attributabilism, then, is to allow that knowledge

amounts to nothing more than the accidental possession of truth.
reed 2007: 251
If rejecting Attributabilism is the same as allowing that knowledge is no more

than the accidental possession of truth, then affirming Attributabilism is the
same as disallowing that knowledge is no more than the accidental p ossession
18 Some external world beliefs are justified by being validly inferred from others. But for all
of them it is quite plausible that their ultimate non-derivative justifying basis is fallible.
We can assume that it is this justification that Fallibilism* is about.
of truth.19 Removing the double negation, affirming Attributabilism is affirm-

ing that knowledge is something more than the accidental possession of truth.
Strictly speaking, this affirmation would be true if knowledge were the acci-
dental possession of truth, plus something else. Reed clearly does not intend
that according to Attributabilism knowledge allows accidental truth. Better if
the conclusion had said that rejecting Attributabilism is the same as allow-
ing a case that includes accidentally possessed truth to be knowledge. So un-
derstood, the conclusion tells us that to affirm Attributabilism is to deny that
knowledge is ever accidentally possessed truth:
Attributabilism*: Knowledge requires the non-accidental possession of

truth, where the accidental possession is the sort of possession that gen-
erates Gettier cases.
Attributabilism* is not a promising premise for a skeptical argument with Fal-

libilism* as the other premise. Attributabilism* tells us that knowledge is never
a Gettier-type accidental possession of truth. Fallibilism* tells us that what
justifies external world knowledge could have existed along with the acciden-
tal possession of that truth. So the two premises allow that there is external
world knowledge whenever there is a fallibly justified true belief that actually
meets condition x (called “condition c” in the present work (2007: 246)), the
condition that blocks the Gettier accidents. In other words, the premises do
not entail that knowledge is anything more than justified true belief with no
Gettier accident. Yet that requirement is no skeptical result. It is a common
post-Gettier view of knowledge. It allows that our external world beliefs are
often known.
A skeptical result would follow if it were added that condition x is rarely
or never met by fallibly justified beliefs. Reed gives no reason to accept that
additional claim. The Bs and numerous other everyday examples of appar-
ent external world knowledge seem to show that condition x is often met.
19 This inference assumes that the first “is” in the quote from Reed expresses identity rather
than inclusion. If Reed means that the rejection of Attributablism includes the allowing,
then the affirming could also include the allowing. But Reed clearly does not think that
Attributionism allows Gettier accidents in knowledge. At the other extreme, Attribution-
ism could not credibly disallow all accidents that might be involved in knowing. Acciden-
tal occurrences enable us to gain some knowledge. For instance, when we know that a
hummingbird is close, this can be partly as a result of the accident of happening to look in
the direction of a hummingbird’s quick visit to a nearby flower. In any event, we are t rying
out the interpretation according to which Attributionism is the claim that knowledge
disallows Gettier accidents.
150 Conee
The justification of the beliefs often seems to derive in the right non-accidental
way from the relevant facts.
In order to give the skeptical argument every reasonable chance of success
we should try another interpretation of Attributabilism. We can do this by
making use of Reed’s way of summarizing the skeptical result of the implica-
tions of Fallibilism and Attributabilism:
So, when the factors that might lead to accidentality are absent, this is
not something that can be attributed to the subject.
The result, then, is that the person’s cognitive performance is not good
enough to allow her to reach the truth. Her performance must be supple-
mented by something extrinsic – and therefore unattributable – to her.
Thus, knowledge itself is not attributable to the person, either. Attrib-
utabilism is therefore false, as long as we are committed to fallibilism. A
fortiori, any theory of knowledge committed to attributabilism –which is
to say, every theory of knowledge – is also false.
reed 2007: 248
Reed’s first statement here is not clearly right, but it can be granted. What he
calls “the factors that might lead to accidentality” are the factors that might
prevent condition x from obtaining. It is true that we are not usually able to
avoid all possibilities of Gettier accidents. Usually we are vulnerable to being
in some Gettier case situation where we unwittingly get reasons to believe a
true proposition that do not rightly derive from its truth. When we are not in
this situation, it is partly a result of favorable fortune rather than our inten-
tional avoidance. It need not be a sheer accident though. We can work against
being misled and have success. Generally, however, we are vulnerable to other
ways of being misled, ways that we fail to suspect. For instance, though I might
have been vigilant against perceptual anomalies, I might still have been taken
in by the distant-puddle-on-the-road illusion when it happened to occur above
a real but unseen puddle. I might have been unaware of that illusion. Using the
language that we have seen Reed to think commits a theory to Attributabilism,
we can grant that our not falling victim to any such accident is not an “achieve-
ment” by us; it is not to our “credit” that no such accident occurs. At least some-
times it includes a large dose of good luck. In any event, we can grant that the
fact of our not being in any Gettier case is not a creditable achievement.
Reed infers from this that “… the person’s cognitive performance is not
good enough to allow her to reach the truth.” Reed cannot mean quite that.
Reaching the truth is allowed by the performance. Our cognitive performance
has “reached the truth” whenever we have gained a true belief. What Reed
means is indicated by his next sentence. A “supplement” to the performance is

needed. This must be a need to satisfy condition x, the non-accidentality con-
dition on knowledge. Again, that condition is “extrinsic” to the performance,
in that the performance is neither metaphysically nor causally sufficient for
condition x to obtain. So what Reed must mean by “not good enough to allow
her to reach the truth” is that the person’s cognitive performance is not a meta-
physically or causally sufficient condition for the satisfaction of condition x.
This should be agreed. It is causally possible, and so metaphysically possible,
for the same justifying performance to occur circumstances in which the belief
is Gettier accident.20
The next step is a mistake. Reed infers, from the fact that the performance
is in this sense “extrinsic” and in need of “supplement,” that the knowledge is
not attributable to the person. That conclusion does not follow. A performance
that is not guaranteed to achieve a result may actually achieve it. For example,
a spectacular basketball shot that goes in as a result of the player’s extraordi-
nary skill is a scoring achievement and the shooter deserves credit for it, even
though it might have been blocked by an even more spectacular defensive ma-
neuver. The scoring is “extrinsic” to the shooting performance in the sense at
stake - the shooting does not causally guarantee the scoring. The shot is still a
creditable achievement.
In a similar though less extraordinary way, in examples like that of Bar-
tholomew in Car Possession 1 the subject makes a correct judgment based on
justifying evidence. This is a doxastic task that is well and successfully done.
The subject believes a true proposition by an exercise of cognitive skill. The
skill acquires a basis for belief that is ample and unimpeached justifying
20 Some external world belief justifications do not just obviously causally allow Gettier
a ccidents. For example, if we justify by rigorous scientific testing our belief that gold con-
ducts electricity, then in light of its lawful status it might seem that the belief with that
justification could not be a Gettier case in any causally possible way. First it should be
noted that there is no easy recipe for illustrations of the possibility that is neutral across
the main theories of justification, largely because the relevant process types for reliabi-
list justification are unspecified. In any event, it is causally possible for environmental
sources of evidence to have been drastically fraudulent. Using this fact, here is a causally
possible way for us to have had the same internal evidence about gold and perhaps also to
believe with the same relevant process types and thus the same reliabilist justification: An
elaborate ruse is perpetrated by people who think they are misleading us about gold. Our
careful testing of “gold” is actually performed on a simulation that we are justified to think
is gold. So our testing gives us a justified true belief that gold conducts electricity. Given
the intent to deceive us though, it is a Gettier-type accident that the justification justifies
a truth about gold. Various causally possible hoaxes like that go a long way toward show-
ing the universal vulnerability of external world belief justification to Gettier accidents.
152 Conee
e vidence. The evidence derives in a proper way from its truth. The believing
satisfies condition x. This is true despite the possibility of a Gettier accident
having occurred that would have blocked the same belief on the same grounds
from achieving that result. The possession of truth is “attributable” to subjects
like Bartholomew in the current sense, a sense that is implied by earning credit
for the achievement of knowledge.
It is doubtful that knowledge is always an achievement or does credit to the
subject. There are counterexamples. There are examples in which s omething
that happens in the environment initiates an involuntary activation of cogni-
tive capacities that results in knowledge. For instance, sometimes when we are
paying no attention to our environment – perhaps we are trying to get to sleep
with our eyes closed – we cannot help learning a fact about our surround-
ings, say, the fact that a bright light has started flashing in our presence. When
knowledge is thus foisted upon us, it does not appear to be an achievement or
otherwise credit-worthy.
Even if we grant that knowledge must be a credit-worthy achievement,
Reed’s argument does not show that our seeming knowledge usually fails to
meet that requirement. We have seen that the possibility of the same men-
tal activity having failed to gain us knowledge does not establish a lack of
achievement. In the absence Gettier obstructions to meeting condition x, we
can exercise cognitive skill and achieve knowledge. More generally, whether
or not achievement is required, we have seen no good reason to think that our
adequately justified true beliefs are often accidental in any way that prevents
us from having knowledge.
5 True Vagueness
The truth condition on knowledge is an exotic basis for a skeptical conclusion.

But David Braun and Ted Sider (2007) have argued that virtually every ordinary
sentence is untrue because it is vague. We shall see that ordinary belief contents
are subject to the Braun and Sider sort of argument too. So if they are right, then
we have virtually no true ordinary beliefs. Since knowledge requires true belief,
it follows that we have virtually no ordinary knowledge. It will be argued in re-
sponse that the vagueness can be accommodated without skeptical implica-
tions, if we make use of a metaphysically different account of belief contents.
Braun and Sider hold that virtually every English sentence includes vague
expressions. The only potential exceptions that they mention are some math-
ematical sentences and sentences with a purely logical vocabulary (2007: 3, n.
9). They do not argue explicitly for the nearly universal vagueness of natural
language sentences. But their explication of vagueness, together with their

account of candidate meanings, strongly suggests how they arrive at this
conclusion. The suggested route will soon be described. We shall see that the
reasoning extends readily to the contents of beliefs.
Braun and Sider characterize vagueness as follows:
An expression is vague if it can be unclear to a speaker informed of all

relevant facts whether the expression correctly applies … Like many, we
think that vagueness occurs when there exist multiple equally good can-
didates to be the meaning of a given linguistic expression.
braun and sider 2007: 2
Braun and Sider characterize the candidate meanings for expressions as

follows:
We assume that the properties, relations, and propositions that are can-
didates for being the meanings of linguistic expressions are precise: any
n-tuple of objects either definitely instantiates or definitely fails to in-
stantiate a given n-place relation, and any proposition is either definitely
true or definitely false.
To give one example of how English expressions relate to meanings, Braun and
Sider assert that usage, naturalness, causal relations and the like do not deter-
mine that any unique fully determinate property is meant by “red” and they
infer that no sentence containing “red” expresses any unique fully determinate
proposition (2007: 2–3).
I speculate that Braun and Sider go beyond such examples to arrive at
their view of nearly universal English vagueness as follows. They find that for
each of various typical English terms we can readily conceive of borderline
cases. For instance, “chair” is well understood, but what of a so-called bean-
bag chair, or an otherwise chair-shaped piece of wood that is missing one or
two of its legs? Also they find that we can move gradually, with no obvious
loss of applicability, from central cases of the application of an ordinary term
to more and more problematic cases before reaching cases where the term
clearly no longer applies. We can be fully informed about the relevant facts
concerning these problem cases and be unclear as to whether or not the ex-
pression applies. The properties with precise extensions that are candidates
for the meanings are the properties that apply correctly in all clear cases.
Some of these properties definitely extend to some of the borderline cases;
154 Conee
the others definitely do not extend to those cases and extend instead to oth-
ers. Nothing seems to make any of these candidate properties better for the
meanings than any of the others. If so, it follows that for each of the ordi-
nary expressions, there are multiple equally good candidates for the meaning.
Equally problematic cases can be found for any number of arbitrarily chosen
terms. A universal generalization by induction seems compelling. A potential
lack of clarity about application seems to be wholly pervasive among natural
language terms (or almost universal if some sentences in mathematical and
logical vocabularies are exceptions).
Finally, Braun and Sider argue that natural language sentences are just like
ambiguous sentences in lacking a truth-value.
To be either true or false, a sentence must have a unique meaning. Am-

biguous sentences do not have unique meanings. Therefore, they are nei-
ther true nor false. Similarly, sentences containing vague expressions do
not have unique meanings; therefore, they too are neither true nor false.
If that is how Braun and Sider reach their conclusion about a virtually univer-
sal lack of unique meaning in natural language and a resulting lack of truth-
value, then the same considerations apply equally well to mental candidates
for determining the contents of our ordinary beliefs. (Braun and Sider say in
a footnote “similar points apply to mental items” without elaborating (2007: 3.
n.9)). As an example we can consider how we mentally represent redness to
ourselves – call this “our concept of red.” We can think carefully to get as clear
as we can about the concept. Still, it is possible to imagine a borderline reddish
thing, or imagine a sequence of increasingly less red things, concerning which
it becomes unclear to us whether the concept applies, no matter what relevant
facts we assume, apart from the fact of applicability itself. This lack of clarity
gives the same support to there being no unique properties that are conceived
by our concepts. By the same sort of reasoning from unclear possible cases
it follows that there are no unique propositions that are represented by the
contents of our thoughts. The lack of uniqueness gives the same support to
the conclusion that our thought contents lack truth-value. The contents of our
beliefs are the thoughts that we accept. Thus we have the same support for the
conclusion that our ordinary belief contents are untrue.
Given Braun and Sider’s views of vagueness and truth, we seldom if ever
speak or think the truth. Yet Braun and Sider never suggest that the view has
skeptical consequences, and they deny that it is a radical view (2007: 4). They
defend the position that it lacks revisionary implications for how we think
and speak by employing three technical notions: legitimate disambiguation,
approximate truth, and ignoring vagueness. To explain the legitimate

disambiguations of a sentence Braun and Sider say that natural language
sentences have a “cloud of propositions” that are “in the neighborhood” of
each sentential tokening (2007: 4). Presumably what makes a proposition
“neighbor” a sentential token is that nothing about the tokening excludes
the proposition from being meant by the tokened sentence. Although fac-
tors that make for vagueness prevent the tokener from singling out any one
of the propositions and meaning exactly it, these neighboring propositions
are the legitimate disambiguations of the sentence token. In other words, the
legitimate disambiguations of a sentential token are those propositions any of
which the speaker might have meant by the sentence, if the factors that make
for vagueness did not prevent that. A sentential token is approximately true
just when all of the legitimate disambiguations of that token are true (2007: 4).
Finally, a sentential user ignores the vagueness of a sentence just when the user
does not take account of the vagueness and its effect on the sentence’s truth
(2007: 7).
Using these notions Braun and Sider hold that it is usually harmless to ig-
nore the vagueness of a sentence that is approximately true. They say that
insisting on truth instead would be “pointlessly fussy,” because the differences
between the legitimate disambiguations of approximately true sentences are
“rarely significant to us” and “do not matter.” As a result, someone who makes
an assertion by tokening an approximate truth “satisfies her communicative
obligations well enough” and someone who accepts a token of an approximate
truth “satisfies her intellectual obligations well enough” (2007: 5).
In the matter of intellectual obligations, Braun and Sider also mention that
the goal of successful inquiry may be known truth and not just accepted truth
(2007: 5). Knowledge is a more plausible intellectual goal for inquiry than is
true belief. For one thing it is clear that our most detached curiosity has us
interested in knowing the truth, not just believing it. Merely accepting an ap-
proximate truth, on whatever basis, is not satisfying “well enough” any intel-
lectual obligation to know the truth. The approximately true thought content
might be just guessed. Braun and Sider do not say anything else about meeting
goals of inquiry and they do not further address the epistemic implications of
their view. But they could add that we satisfy well enough an intellectual obli-
gation to know the truth, if we believe an approximately true thought content
and the believing meets the other necessary conditions for knowledge of any
proposition from among the legitimate disambiguations of the thought. The
idea would be that we often succeed at approximately true believing, with the
rest of the conditions for knowledge in place – say, having well-founded justifi-
cation for any of the legitimate disambiguations of the thought and no Gettier
problem. They might also say that usually our being in such a condition differs
156 Conee
harmlessly from knowing the truth and that it would be pointlessly fussy to
insist on knowing the truth.
Still, this extension of their view to knowledge is drastically skeptical. We do
not know any ordinary fact (noted potential exceptions aside). There would be
a difference between on the one hand knowing truths, and on the other hand
accepting approximate truths while meeting the appropriately adjusted fur-
ther conditions on knowledge. This difference would not be important to us, at
least for practical purposes. But it is an implication of this way to extend Braun
and Sider’s theory that we know approximately nothing.
Braun and Sider are well aware that according to their theory we cannot say
or think (almost) anything that is true, including their statement of the theory
itself.21 They contend that the implied lack of truth does not render the
theory unacceptable. In defense of its acceptability they cite their claim that
approximate truth satisfies well enough the obligations of communication and
inquiry and they assert the approximate truth of their theory (2007: 10–11).
Braun and Sider could also observe that there are many approximate truths
that deny skepticism. For example: “We often have ordinary knowledge, since
we often meet all of the conditions on knowledge.” But finally, when we take
into account as much of reality as we can by paying attention to vagueness and
its effect on truth, according to Braun and Sider what we find is that this appar-
ently anti-skeptical claim is untrue. The same goes for our ordinary belief con-
tents. The core reason for this is that in their view vagueness prevents us from
being in any familiar psychological relation to any single true proposition.
This is skepticism about nearly all knowledge. Even so, it might not be an
unacceptable skeptical result. Suppose that we lack nearly all knowledge by
virtue of our inability to isolate nearly any individual true proposition as the
content of a thought. Still, often we would apprehend clouds of propositions
that contain truths only and we would meet the other conditions for knowing
the propositions. As Braun and Sider say, the differences among the truths in
those clouds rarely make a difference to our concerns.
One of our concerns would be unmet. Our concern to know the truth would
not be satisfied (with the possible math and logic exceptions). Braun and Sid-
er could hold that this concern is met well enough by the frequent approxi-
mate truth of claims that we know various propositions. The theory also has
21 Saying this about the lack of truth is applying the part of Braun and Sider’s view that
tells us about how vagueness results in a lack of truth-value. That part is supposed to be
correctly reported only when we are not ignoring vagueness. Another part of their view
implies that if we were ignoring vagueness when we were considering the application of
their theory just now, then it would have been acceptable to affirm “we can say and think
many true things” and it would have been unacceptable to reject that.
it, however, that when we attend to vagueness we can see that there is no true
proposition that is either believed or known by us (potential exceptions aside).
Braun and Sider could also hold that when a “knowledge” attribution is ap-
proximately true, our psychological relation to the true propositions in the rele-
vant cloud places us in a very knowledge-like relation to each of them, a relation
that we could think of as a “precisified” knowledge relation. Nevertheless, in
their view, when we are subjects of an approximately true “knowledge” attribu-
tion, we have a more direct cognitive relation to the cloud of true propositions
that are the legitimate disambiguations of the attributed “knowledge” content.
It is this more direct relation that induces the precise knowledge-like relation.
The latter relation is not knowledge. One difference is that in knowledge a sin-
gle proposition is related to the subject by the psychological relation of belief.
In contrast, each proposition to which we are related in the precisified way is
not psychologically distinguished from any other, by belief or otherwise, in the
cloud of true propositions that legitimately disambiguate the belief content.
We can salvage our having lots of knowledge. There is a different way to
take into account Braun and Sider’s reasons for finding pervasive vagueness
and their requirement for truth of the sentence or the thought of a unique
proposition meant or believed.22 The new account also accepts that we lack
semantic and cognitive access to virtually any single precise proposition. The
new thought is that we do often have psychological access to single proposi-
tions, but those propositions are indefinite in certain ways. The general idea
is that some properties and some propositions are partially indeterminate in
their extensions. There are borderline cases where the properties are neither
instantiated nor not instantiated. Reality is indeterminate in the matter. Where
a partially determinate property is neither definitely instantiated nor definite-
ly not instantiated, a proposition asserting it to be instantiated is indetermi-
nate – neither true nor untrue. The properties are also partially determinate.
The properties have definite instances in central cases, away from any relevant
borderline. Where a partially determinate property is definitely instantiated, a
proposition asserting it to be instantiated is true, just plain true.
For example, a typical ripe lemon, L, definitely has the partially determinate
property of being yellow. A proposition asserting L to have the property is just
plain true. Where a partially determinate property definitely fails to be had, a
22 Several existing theories of vagueness allow some ordinary sentences to be true. For a
helpful survey and bibliography, see the Stanford Encyclopedia of Philosophy entry,
“Vagueness” (https://plato.stanford.edu/entries/vagueness/). The present approach is in-
tended to allow ordinary truth while maximally accommodating Braun and Sider’s views
both about truth and about semantic vagueness and its mental counterpart.
158 Conee
proposition asserting that it is had is definitely untrue. For example, a typical

ripe blackberry, B, definitely does not have the partially determinate property
of being yellow. A proposition asserting B to have the property is just plain
untrue. Where a partially determinate property is neither definitely had nor
definitely not had, a proposition asserting that it is had is indeterminate in
truth-value, neither true nor untrue. A yellowish brownish onion, O, neither
definitely has the partially determinate property of being yellow nor definite-
ly fails to have it. A proposition asserting O to be yellow is neither true nor
untrue.23
Assuming that there are such properties and propositions, they are apt
candidates for the meanings of our vague predicates and sentences, and for
the contents of our vague concepts and thoughts. Assuming their existence,
it is plausible that our vague sentences, concepts, and thoughts single out the
properties and propositions that are as determinate in extension as are the
vague sentences, concepts, and thoughts respectively. The partially determi-
nate meanings allow us to concur with Braun and Sider that being true, as it
applies to sentences (and to thoughts), is having a unique meaning (or con-
tent) that is true. But we can disagree with their view that virtually no ordinary
sentences (or thoughts) have unique meanings that are true. Instead we often
do have the psychological relations of meaning and thinking to some partially
determinate proposition. Our meanings and contents are truths when they are
partially determinate true propositions.
The present approach to vagueness avoids a liability of Braun and Sider’s
theory. Their theory holds that sentences and thoughts that include some
vagueness are approximately true at best, not just plain true, even when the at-
tributions made by the vague sentences and thoughts are far away from any bor-
derline cases. Consider for example the sentence, “A man with no hairs on his
head is bald.” Vagueness makes even this sentence untrue, according to Braun
and Sider’s theory (2007: 5, 10).24 To all appearances though, typical t okens of
23 Truth is more complicated for propositions that are about mixtures of determinate and
indeterminate instances of properties. Continuing with the properties of being a lemon
and being yellow, does a definitely ripe lemon that is largely discolored by disease but par-
tially yellow refute the claim, “All ripe lemons are yellow”? If the extension of yellowness
is indeterminate in this case, then the whole sentence may be indeterminate. But since in
this situation not all ripe lemons would be determinately yellow, the case also gives some
credibility to a verdict of untrue. Working this out is part of a complete theory of truth
for sentences and thought contents. These details are not vital here. Such mixed cases
are not good candidates for justified beliefs or knowledge, which are our focus. Enough
determinate examples of truth remain.
24 According to Braun and Sider’s view, this claim that the sentence is untrue, and the
other theory-derived claims of lack of truth, are only approximately true. But although
the view’s proponents are committed to applying it to descriptions of its applications
the sentence are true. A typical “a man with no hairs on his head” token desig-
nates an extreme sort of definite baldness. The doubts about a pplication that
arise from the vagueness of “bald” (and “man”, “head,” and “hair”) are not raised
by typical uses. In contrast, those typical tokens are just plain true according to
the present view. The subject term “a man with no hairs on his head” means a
partially determinate property or a conjunction of such p roperties. In typical
uses the extension of that property or conjunction of properties is definitely
included in the extension of the partially determinate property that is predi-
cated by “is bald.” Consequently the sentence token is just plain true.
The present approach to vagueness bears out our evidence about truth and
knowledge. According to the present view, when the contents of our thoughts
are propositions that make attributions within the extensions of their proper-
ties or relations, the contents of the thoughts are true. Our thoughts very often
make such attributions, according to the appearances that constitute our basic
evidence in the matter. That is, our evidence supports that very often things
are the ways that we think they are. So, going by this evidence, very often we
have true beliefs. Also going by our evidence, we often meet the other stan-
dard conditions on knowing those truths. We have no contrary or defeating
evidence about this, assuming that the present view adequately accommo-
dates the challenge of vagueness. We can reasonably conclude that the present
view of how vagueness impacts belief and knowledge does not have skeptical
implications.
Are there such meanings and contents – properties and propositions with
only partially determinate extensions? Their existence will not be defended
here, apart from noting the support for them that is provided by their explana-
tory value. They would help us to agree with almost all that Braun and Sider
contend about vagueness and truth while affirming the range of true belief and
knowledge that our evidence supports.
6 Conclusion
The extraordinary arguments that we have reviewed for external world skep-
ticism do not succeed. The arguments have instructive flaws. First, we have
seen that the strength of justification that can give us knowledge of a proposi-
tion need not entail its truth for us to understand what that strength is and
to examples, the rest of us are not so committed. If we can tenably account for the truth
of claims that include vague terms, in the present way or in some other way, then we
can sensibly hold that it is just plain true that according to their theory these things are
untrue.
160 Conee
why it is strong enough. Second, we have seen that some of a proposition’s

own entailments can be evidence that we have against it. Third, we have seen
that knowledge can be gained without the knowledge being guaranteed by our
cognitive performance. And fourth, we have seen that we can know the truth
of vague thoughts if the contents of the knowledge are themselves somewhat
indeterminate.
Bibliography
BonJour, L. 2010. “The Myth of Knowledge,” Philosophical Perspectives 24: 91–104.

Braun, D. & T. Sider. 2007. “Vague, So Untrue,” Noûs 41: 133–156.
Chisholm, R. 1966. Theory of Knowledge. Englewood Cliffs, N.J.: Prentice-Hall.
Chisholm, R. 1977. Theory of Knowledge 2nd Edition. Englewood Cliffs, N.J.: Prentice-Hall.
Chisholm, R. 1989. Theory of Knowledge 3rd Edition. Englewood Cliffs, N.J.: Prentice-Hall.
Conee, E. 2015. “Debasing Skepticism Refuted,” Episteme 12: 1–11.
Conee, E. 2017. “Good to Know,” Philosophical Studies 174: 311–331.
Feldman, R. & E. Conee 2004. “Making Sense of Skepticism,” Evidentialism: Essays in
Epistemology, 277–306. London: Oxford University Press.
Reed, B. 2007. “The Long Road to Skepticism,” Journal of Philosophy 104: 236–262.
Reed, B. 2008. “A New Argument for Skepticism,” Philosophical Studies 142: 91–104.
Shaffer, J. 2010. “The Debasing Demon” Analysis 70: 228–237.
Smith, M. 2016. “Scepticism by a Thousand Cuts,” International Journal for the Study of
Skepticism 6: 44–52.
Sorensen, R. 2016. “Vagueness,” In E. Zalta (ed.), The Stanford Encyclopedia of Philosophy.
Vogel, J. 2014. “E & ~H.” In D. Dodd & E. Zardini (eds.), Skepticism and Perceptual Justi-
fication, 87–107. New York: Oxford University Press.
Wasserman, W. 1980. “Chisholm’s Definition of the Evident,” Analysis 40: 42–44.
Chapter 9
Internalism, Memory, and Skepticism

Brett Coppenger
1 Preliminaries
I know that Eli Whitney invented the Cotton Gin. I know that Newton’s theory
of gravitational force is the mass of one object multiplied by the mass of an-
other object divided by the square of their distance. I know that the street I
grew up on was in Long Beach, California. Each of these examples illustrates
something that I know on the basis of memory. And, I think that I know a lot
of other stuff like this, too. In fact, it seems as if most of my knowledge is in
some way tied to my memory. My ability to recall names, dates, and places as
well as my ability to recognize names, dates, and places all seem importantly
tied to memory. Clearly, memory plays an integral role in human knowledge.
However, I also know that my memory is fallible. I seem to remember my
family having a pet desert tortoise and I seem to remember falling off of a
toboggan when I was a baby. Yet, in both of these cases, I have come to realize
that what I am remembering is a photo (I remember seeing a picture of my
family with a tortoise (I was not even born yet!) and I remember seeing a photo
of myself face-down in the snow). It is not uncommon for me to misremem-
ber details of an event, and it is even more common for me to second-guess
my memories. Clearly, dependence on memory brings with it the possibility
of error.
While I think it must be admitted that memory plays an essential role in
the vast majority of what we commonsensically call knowledge, this role is far
from uncontroversial. The goal of this chapter is to explore some prob-
lems related to memorial justification and the bearing of the problems on
internalism.
1.1 Propositional Memory and Episodic Memory

A full analysis and evaluation of the epistemic status of memory would include
an examination of both propositional memories and episodic memories.
Propositional memories take propositions as objects: I remember that I have
four children. Episodic memories take experiences as objects: I remember
each of my children being born.

162 Coppenger
There is an informative tradition that attempts to characterize the clarity

and vivacity of memory experiences in order to makes sense of the beliefs
those experiences justify.1 However, despite this tradition and the importance
in dealing with episodic memory, the focus of this project will be propositional
memory.
1.2 Knowledge and Justification

While the focus of this paper will be to ask about the justificatory status of
beliefs in relation to memory, not everyone agrees that this is the a ppropriate
starting point. Some have argued that remembering is factive. That is, if
I am remembering that P, then necessarily P occurred (I cannot remem-
ber something that did not happen). Instead of pursuing the question of
whether m emory is factive, I think it is natural to draw a distinction between
remembering and seeming to remember. To crystalize this issue, and to put it
in its historical context, consider Russell’s Five-Minute Man Hypothesis:
There is no logical impossibility in the hypothesis that the world sprang

into being five minutes ago, exactly as it then was, with a population that
“remembered” a wholly unreal past.2
As Russell makes explicit, he does not think the hypothesis is plausible.

Instead, he simply hopes to illustrate that the situation is possible. I could have
been created five minutes ago complete with all of the apparent memories I
currently have. As a result, the question I will be pursuing is, when an agent S
seems to remember that P, what justification does S have for P?
1.3 Preservationism and Generativism

Contemporary answers to the question of memory justification often fall
into one of two categories: either the faculty of memory is seen as preserving
justification, or having memory experiences is seen as generating justification.
Preservationism focuses on the initial formation of the belief.3 The idea, ac-
cording to the Preservationist, is that memory records or holds the justification
of our beliefs. When I form a new belief, and that belief is justified in light of
some evidence, the function of memory is to retain or store that justification.
As a result, consider the following principle:
1 See for example Locke (1689), Russell (1921), and Audi (2011).
2 Russell (1921).
3 For defenses of Preservationism see Huemer (1999) and McGrath (2007).
Internalism, Memory, and Skepticism 163
Preservationism: If S is justified in believing P at T1, and P is retained in

memory, then when S seems to remember that P at T2,
S is prima facie justified in believing that P.
Thus, on this view, consider again my belief that Eli Whitney invented the Cot-
ton Gin. According to the Preservationist, if that belief was justified when I
formed it (i.e. on the basis of hearing an authority teach it, or reading about
it in an encyclopedia, etc.), then when I recall that belief from memory now
(whether the original evidence is available or not), it is still justified. My mem-
ory has preserved its justification.
Generativism attempts to locate the source of justification for memorial be-
liefs in present experience.4 The idea, according to the Generativist, is that the
beliefs we seem to remember should be seen as having a positive epistemic status
unless we have reason to doubt them. When I find myself remembering some-
thing, I have justification for accepting my memory as accurate. My memory
experience generates justification. As a result, consider the following principle:
Generativism: I f S seems to remember that P, then S has prima facie jus-

tification for P.
Thus, on this view, consider again my belief that Eli Whitney invented the Cot-
ton Gin. According to the Generativist, since I am having the experience of
seeming to remember that Eli Whitney invented the Cotton Gin, my belief is
prima facie justified on the basis of that experience. The experience of seem-
ing to remember generates justification.
2 Skeptical Problems Concerning Memory
I have already suggested that we commonsensically take ourselves to know a

lot, and that a lot of that knowledge depends on memory. So what, then, might
motivate a skeptical conclusion given the clear dictates of commonsense?
While it would be impossible in the present discussion to give a full survey of
the different problems that seem to threaten memorial justification, there are
three distinct kinds of problems I will explore:5
4 For defenses of Generativism see Audi (2011) and Lackey (2005).

5 Two additional problems come immediately to mind, even though space constraints pre-
vents their discussion here: first, the Problem of Forgotten Defeaters (see Huemer 1999), sec-
ond, the Problem of Stored Belief (see Frise 2017a).
164 Coppenger
2.1 The Problem of the Fleeting Present6

Any analysis of memory is made difficult by the realization that one is attempt-
ing to make sense of current mental states as well as past mental states. This
issue is made especially difficult by the realization that the present is con-
stantly slipping into the past. One kind of problem faced when dealing with
memory has to do with our cognitive limitations on occurrent processing. If
the justification of memory belief has something to do with my recognition of
current mental states, how much can I hold before my mind’s eye at one time?
Goldman (1999: 283–284), in presenting this kind of problem, contends that:
If justification is contingent on the agent’s ability to know what justifiers

obtain, the agent should not be permitted to believe a proposition p at
t unless she can know by t whether the relevant justifiers obtain. Since
it necessarily takes some time to compute logical or probabilistic rela-
tions the simultaneity model of justification needs to be revised so that
an agent’s mental states at t justify her in believing only at t + e, for some
suitable e. The value of e cannot be too large of course, lest the agent’s
mental states change so as to affect the justificational status of p.
The problem, here, is that if memorial justification requires any kind of pro-
cessing, and that processing cannot be done occurently, then the processing
is relying on memory. However, the worry is that this processing would then
delegitimize any conclusion reached. Consider an example: imagine being
asked about the Pythagorean Theorem. Further, imagine that a justified belief
in the Pythagorean Theorem requires being able to see the steps of a proof of
the Pythagorean Theorem. In order to avoid vicious circularity, under these
conditions, one would need to be able to hold the proof and the Pythagorean
Theorem before one’s mind. And, then, the question becomes how many of my
beliefs (mathematical or otherwise) include proofs (or evidence) that push me
beyond the capacities of what I can currently entertain?
2.2 The Problem of Forgotten Evidence7

It should come as no surprise that much of what we experience (or learn) is
not remembered. Even when I do seem to remember some belief, I rarely (if
6 Goldman in introducing this problem calls it the Problem of the Doxastic Interval, see Gold-
man (1999).
7 See McCain (2015), for a thorough discussion of the effectiveness of the Problem of Forgotten
Evidence on Evidentialism.
ever) remember the particular evidence I had when I first formed that belief.
As Goldman (1999: 280) points out,
Many justified beliefs are ones for which an agent once had adequate
evidence that she subsequently forgot. At the time of epistemic appraisal
she no longer possesses adequate evidence that is retrievable from
memory.
In order to see the importance that forgotten evidence can play, consider Gold-
man’s (1999: 280–281) case involving Sally and Mary:
Sally reads a story about the health benefits of broccoli (based on good
science) in some reputable publication. She then forms a justified be-
lief regarding the benefits of broccoli. Mary reads a story about the
health benefits of broccoli (based on bad science) in some illegitimate
publication. She then forms an unjustified belief regarding the benefits
of b roccoli. They both share the same belief (that broccoli is beneficial to
one’s health), but the acquisition seems important.
The problem, here, is that it seems intuitive to say that memorial justification
depends not only on the belief we remember, but also on the evidence we had
for that belief. Of course, if that evidence is forgotten (and the realization is
that we seem to forget a lot of our evidence), then so much the worse for our
belief. Consider, again, my belief that Eli Whitney invented the Cotton Gin. I
have absolutely no recollection of where I was, or when it was, that I learned
that Eli Whitney invented the Cotton Gin. In fact, at best, the proposition Eli
Whitney invented the Cotton Gin simply sounds right to me. How is this belief
justified, if all of my evidence has been forgotten?
2.3 The Problem of Epistemic Defeat8,9

In discussions of epistemic defeat it is helpful to distinguish different types of
defeaters. There are rebutting defeaters and undercutting defeaters. Rebutting
defeaters are reasons that count against the belief in question. If I believed that
it was likely to rain today, hearing a forecast for no rain would rebut my belief.
8 I first presented a version of this problem in a paper presented at the 2011 Annual meeting of
the Ohio Philosophical Association (“Huemer’s Problem of Memory Knowledge.”). The pres-
ent discussion benefits from helpful comments on that paper.
9 For a thorough presentation of this kind of problem, and especially its impact on Preserva-
tionism see Frise (2017b).
166 Coppenger
Alternatively, undercutting defeaters call into question the legitimacy of one’s

reasons for the belief in question. If I believed that I was sitting at my desk right
now, considering Descartes’ Dream Argument would undercut my belief.
The Problem of Epistemic Defeat that I wish to develop is properly un-
derstood as a kind of undercutting defeater. In the same way we can call into
question perception as a source of justification to undercut perceptual beliefs,
and call into question testimony as a source of justification to undercut testi-
monial beliefs, we can call into question memory as a source of justification to
undercut memorial beliefs.
Consider two different approaches to developing the problem. First, the
Problem of Epistemic Defeat from Personal Experience, and second, the Prob-
lem of Epistemic Defeat from Psychology.
In order to see the Problem from Personal Experience, one need only recog-
nize that as we get older our memory seems more and more prone to error.10 I
find myself double-guessing events of my childhood with increasing regularity.
In addition, when it comes to mundane details regarding the previous week’s
committee meetings, I seem not to retain much at all. This perceived increased
ineffectiveness in my memory leads to a troubling realization. When I seem
to remember that P, doesn’t my perceived ineffectiveness give me a reason to
doubt my belief that P?
In order to see the Problem from Psychology, one need only imagine that
current trends in psychology show the ineffectiveness of memory.11 Memory,
psychology tells us, is complicated at best. The mind (or brain) fills in details,
adds details, and subtracts details. If I am to trust current psychological theory
(in this context), and I seem to remember that P, don’t I have a reason to doubt
my belief that P?
3 Varieties of Internalism
Because the goal of this chapter is to explore how the internalist can an-
swer skeptical challenges involving memory, it will be helpful to distinguish
10 I take this to be a datum that is uncontroversial when viewed from the perspective of
personal experience. Of course, as will be developed later, this problem has clear ties to
an issue involving circularity (see Section 6).
11 While I think one could successfully defend the thesis that current research in psychology
supports a healthy skepticism about the accuracy of memory, an adequate defense of this
idea would take us too far afield. As a result, I have only attempted to develop the problem
as a conditional. If this is what psychology shows, then there is a problem. Of course, as
will be developed later, this problem has clear ties to an issue involving circularity (see
Section 6).
internalism (broadly) from externalism (broadly). Consider, again, Russell’s

Five-Minute Man Hypothesis.12 To see how the Hypothesis applies to our cur-
rent purposes, we need only compare ourselves with the five-minute versions
of ourselves.13 On this assumption, the five-minute version of me and I are
exact duplicates (both psychologically and physically).14
According to the internalist’s intuition in this case, if I (the real man) am jus-
tified in my beliefs, then so is my psychological (and physical) twin (the Five-
Minute Man), and vice versa. As a result, it is argued, the justificatory status of
a subject’s beliefs depends on the subject’s current mental states.15 External-
ists, on the other hand, deny that justification is limited in this regard.
This broad characterization of internalism (and externalism) leaves out
many of the intricacies of current prominent versions of internalism (and
externalism). Because each one of these formulations of internalism might
handle skeptical challenges involving memory differently, it is worth trying to
distinguish, in some detail, the different kinds of internalism.16
3.1 Traditional Internalism

What distinguishes traditional internalism from other kinds of internalism?
Richard Fumerton (2004: 75), one prominent and systematic defender of the
view, contends that,
[T]he fundamental internalist concern is that having knowledge or justi-

fication in the externalist sense doesn’t seem to satisfy philosophical curi-
osity. It doesn’t seem to provide any assurance of the sort the philosopher
seeks when wondering about the truth of various propositions.
Thus, according to Fumerton, the essential feature of internalism is not its

claim that justification depends on the internal states of the believer, nor
any interest in accommodating our commonsense assumptions about the
12 While Russell does not utilize the Five-Minute Man Hypothesis to distinguish internalism
from externalism, the thought experiment’s applicability is clear.
13 Of course, the situation being described here is more commonly presented and dealt with
in the literature as the New Evil Demon Problem.
14 When the five-minute version of me is created it is my exact duplicate. However, immedi-
ately following creation psychological and physical similarities could (will) diverge. How-
ever, for sake of simplicity I will ignore these subtle changes. As an alternative hypothesis,
imagine the Every-Instant-Man: a psychological duplicate that is created every instant
(i.e. each time I undergo any change, a new duplicate is created.).
15 ‘Current mental sates’ are not meant to be synonymous with occurrent mental states.
One’s current mental states could include dispositional states as well.
16 For a more complete analysis of the varieties of internalism along similar lines, see Cop-
penger (2016).
168 Coppenger
j ustification of our ordinary beliefs. Instead, what characterizes traditional in-

ternalism is the aim of emphasizing the need for philosophical assurance, on
the basis of evidence that can withstand the strongest skeptical challenges,
that our ordinary beliefs are true. And the key to obtaining this assurance is to
be directly acquainted or confronted with facts to be known.
According to the traditional internalist, the only way one could achieve this
kind of assurance, this kind of robust answer to the skeptic, would be by seeing
the connection of one’s belief to the truth. It is not enough that one has a true
belief that happens to be true; what one needs in order to be justified is some
way of directly seeing that one’s belief is true.
On this kind of reading, traditional internalism can be traced back to Des-
cartes and his search for indubitable belief from the first person perspective.
What Descartes wanted was a belief that he could be confident in, a belief that
rested on an indubitable foundation that left no room for error. Descartes’ proj-
ect was one of trying to secure philosophical assurance that skepticism was
wrong, and it was only in light of this assurance that Descartes could proceed
with peace of mind.
On this kind of view the intuition behind the new application of Russell’s
Five-Minute Man Hypothesis is that because the key to obtaining assurance
is to be directly acquainted or confronted with the facts to be known, the
real-man and the Five-Minute-Man will be in the same situation. Because
they have the same mental states, they are limited to the very same objects of
acquaintance.
3.2 Evidentialism
What distinguishes evidentialism from other kinds of internalism? The evi-
dentialist maintains that justification is a matter of believing in accordance
with one’s evidence. On this view, my belief that I am sitting at my desk is
justified only if that belief accords with the evidence that I currently possess.
Conee and Feldman (2004: 83), in defending evidentialism, argue that, “the
epistemic justification of a belief is determined by the quality of the believer’s
evidence for the belief.” They have famously defended an account of what they
call evidentialist justification (EJ):
EJ: Doxastic attitude D toward proposition p is epistemically justified for

S at t if and only if having D toward p fits the evidence S has at t. (2004: 83)
Proponents of evidentialism can go on to argue that what counts as evidence

for a subject is that subject’s mental states. When these ideas are combined, a
version of mentalism results, “If any two possible individuals are exactly alike
mentally, then they are alike justificationally, e.g., the same beliefs are justified
for them to the same extent” (Conee and Feldman 2004: 57).
This kind of internalism appears to capture our commonsense intuitions
about the similar justificatory status of two subjects who share all the same
evidential states. On this kind of view the intuition behind the new application
of Russell’s Five-Minute Man Hypothesis is that since justification is limited to
the evidential states of a believer, two individuals who have exactly the same
mental states will also have the same evidential states (as long as we are grant-
ing the thesis of mentalism).
3.3 Phenomenal Conservatism

What distinguishes phenomenal conservatism from other kinds of internal-
ism? This type of conservatist might maintain that there is an intimate con-
nection between how things seem to a subject and the justified beliefs of that
subject. On this view, I believe that I am sitting at my desk because that visually
and tactilely seems to me to be the case; and that belief is justified because it is
based on that visual and tactile seeming, with that same propositional content.
Huemer (2007: 39) has argued that, “when we form beliefs, our beliefs are
based on the way things seem to us.” Huemer contends that beliefs based on
seemings can be justified, and he defends phenomenal conservatism:
If it seems to S that p, then, in the absence of defeaters, S thereby has at

least some degree of justification for believing that p. (2007: 30)
Because seemings are a kind of internal state, conservatism, fits nicely into the
category of internalism. On this kind of view the intuition behind the new ap-
plication of Russell’s Five-Minute Man Hypothesis is that since justification is
the result of what seems to be the case for the believer, two individuals who
have exactly the same mental states will also have the same seeming states.17
4 Developing a Strategy of Attack
Depending on the kind of internalism that someone holds the efficacy of the
kinds of problems of memory outlined in Section 2 could very well differ. Un-
surprisingly, different views will have different problems. However, since each
of the varieties of internalism presented in Section 3 is a version of internalism
17 Of course, if someone takes a phenomenal seeming as part of one’s evidence, then phe-
nomenal conservatism could be interpreted as a variety of evidentialism.
170 Coppenger
it would also not be too surprising if there was a kind of common problem that
threatened each version of internalism. In what follows, I hope to motivate a
strategy of attack that seems to threaten each of those prominent versions of
internalism.
The kind of strategy I am interested in depends on two distinct steps: first
the skeptic calls into question the legitimacy of the justification a subject has
for a particular memorial belief. Then, second, the Problem of the Fleeting
Present is used to challenge any attempt to circumvent the challenge posed
in step one.
An analogy at this point will be instructive. Imagine being stuck in an inter-
rogation room with a stranger named Richard. After some discussion, Richard
raises the possibility that everything he has said up until this point might have
been a lie. Unfortunately, because you are stuck in this interrogation room,
the only way for you to determine the veracity of Richard’s claim is to depend
further on conversation with Richard. Of course, at this point, your epistemic
position seems dire.
5 Applying the Strategy to the Varieties of Internalism
While ultimately I am curious (worried) about the implications of the strategy

developed in Section 4, I also think (am hopeful) that the strategy might be
limited. In what follows I intend to sketch how the strategy could be employed
towards each of the prominent varieties of internalism already presented.
5.1 Traditional Internalism

The traditional internalist could adopt an account of either Preservationism or
Generativism with reference to memory justification. Because the traditional
internalist, as I understand the view, is ultimately concerned with the assur-
ance of the truth of her beliefs what will be of ultimate interest is how either
Preservationism or Generativism provides that kind of assurance.
If the traditional internalist were to adopt Preservationism then the Prob-
lem of Forgotten Evidence seems like it might be of primary concern. How can
my seeming to remember that P justify me in believing that P if I have forgot-
ten the evidence I had for P? Surely, it would be inappropriate to be assured of
truth of the belief that P simply on the basis of seeming to remember. At this
point a subject should be actively questioning the assurance of her belief.
If the traditional internalist were to adopt Generativism then the Problem
of Epistemic Defeat seems like it might be of primary concern. How can my
seeming to remember that P justify me in believing that P if I have reason to
q uestion the epistemic status of my seemings? Surely, it would be inappropri-

ate to be assured of the truth of the belief that P while questioning the le-
gitimacy of my seemings. At this point the traditional internalist Generativist
should actively question the assurance of her belief.
Regardless of how the traditional internalist came to question the assur-
ance of her belief, the issue now becomes how might she answer that question.
But, of course, at this point, the strategy of attack calls into question the legiti-
macy of trying to reason or infer any conclusion that depends on the use of
memory. Ultimately, my worry is that any such inference will run afoul of the
Problem of the Fleeting Present.
5.2 Evidentialism
Perhaps, the evidentialist will fare better.
If the evidentialist were to adopt Preservationism then the Problem of
Forgotten Evidence seems like it would threaten exactly the same issue that
caused trouble for the traditional internalist. If justification depends on evi-
dence, and the evidence is forgotten, how could my belief be justified?18
Of course, the typical move here is to introduce either the possibility of new
evidence, or the possibility of dispositional beliefs that are doing epistemic
work. However, of seminal importance is the idea that at this point the evi-
dentialist could be questioning the evidence that is available for her memorial
belief.19
If the evidentialist were to adopt Generativism then the Problem of Epis-
temic Defeat seems like it might be of primary concern. On this approach, my
seeming to remember that P is part of my evidence for P. However, just as was
the case with traditional internalism, it seems like the Problem of Epistemic
Defeat can be used to call into question the epistemic status of this particular
belief. At this point the evidentialist should be questioning the available evi-
dence for her memorial belief.
Regardless of how the evidentialist came to question the evidence available
for her belief, the issue now becomes how might she answer that question. But,
of course, at this point, the strategy of attack calls into question the legitimacy
of trying to reason or infer any conclusion that depends on the use of memory.
Ultimately, my worry is that any such inference will run afoul of the Problem
of the Fleeting Present.
18 As McCain (2015) helpfully points out, this problem presupposes the idea that the belief
in question is still justified (despite the loss of the original evidence).
19 To be clear, the problem, as I am trying to develop it, only applies to the evidentialist who
is concerned with the issue at hand.
172 Coppenger
5.3 Phenomenal Conservatism

Perhaps, the phenomenal conservatist will fare better. The natural move for
the conservatists to make at this point is to endorse Generativism.20 And by
now, my concern should be apparent. The Problem of Epistemic Defeat should
be of primary concern. It can be granted that I seem to remember that P, and
seeming to remember P gives me some reason to believe that P. However, given
the Problem of Epistemic Defeat, it might also seem plausible that memory
seems unreliable to me. At this point, the conservatist should be questioning
the legitimacy of her memorial seemings.21
Regardless of how the conservatist came to question the epistemic status
of her memorial seemings, the issue now becomes how she might answer that
question. But, of course, at this point, the strategy of attack calls into question
the legitimacy of trying to reason or infer any conclusion that depends on the
use of memory. Ultimately, my worry is that any such inference will run afoul
of the Problem of the Fleeting Present.
6 Conclusions
I have tried to show that once one is actively calling into question the evi-
dence they have for a memorial belief, it is especially hard to adequately de-
fend against a skeptical conclusion. If this line of argument is successful, and
memorial beliefs prove difficult for the internalist to justify, one might con-
clude at this point, all the worse for internalism (yay externalism!). However,
before rushing to this kind of conclusion, a few more considerations should be
addressed.
6.1 Memory and Circularity

It would be appropriate to point out that much of the forgoing discussion suf-
fers from an important kind of problem. In attempting to motivate the prob-
lems of memory for internalism, especially as it pertains to the Problem of
Epistemic Defeat, I have crucially depended on memory. Couldn’t the problem
at this point be rejected? The very idea that our memory might not work in
20 There is of course no reason in principal that an internalist conservatist could not en-
dorse Preservationism. It would not be the first time a philosopher attempted to occupy
logical space. However, for my present purposes, because Generativism seems so natu-
rally associated with conservatism, my discussion will be limited to that particular view.
21 The problem as I am developing it only exists for the subject who has the seeming that
memory is unreliable. And, again, it seems like this seeming could be the result of either
personal experience or current research in psychology (see Section 3).
this case or that case depends on remembering those cases! The very idea that
psychological research might show that memory is unreliable would depend
on remembering that research! How can this problem even make sense given
this context?
Importantly, I think it is worth focusing on the implications of circularity for
the present discussion. As Fumerton (1995: 50–51) has correctly pointed out:
Suppose that one day a skeptic gets the bright idea of asking [an] eight
ball whether or not conclusions reached via eight-ball reasoning are ra-
tional, and the eight ball answers “No.” To the chagrin of the supporters
of commonsense eight-ball reasoning, the results of the experiment are
duplicated again and again. In what position are eight-ball reasoners left?
I supposed die-hard proponents of eight-ball reasoning can argue that
our skeptic who concludes that it is irrational to believe the dictates of
the eight ball, is certainly in no position to use the eight-ball’s “answers”
to reach that conclusion. This seems right, of course. The skeptical argu-
ment that proceeds from observations about what the eight ball indicates
and reaches a conclusion about the illegitimacy of eight-ball r easoning
is epistemically self-refuting. But should the eight-ball reasoners be
celebrating?
Thus, following Fumerton, if we use memory to show that memory is prob-

lematic (through personal experience or current research in psychology), then
there is an issue of circularity. However, it should be asked, what is the appro-
priate reaction to noticing this circularity? Do we then ignore what personal
experience or current research in psychology might show? Certainly, the an-
swer is no.
6.2 Moving Forward

So what, at this point, is the internalist to do? There are two strategies of reply
that I will only mention in closing. The first brings the discussion back to Des-
cartes, while the second is bound to leave many internalists unsatisfied.
6.2.1 Cartesian Inferentialism

Despite the apparent circularity in attempting to vindicate memory as a source
of justification, it seems as if Descartes was at least consistent in his efforts.
Consider the method: first, I hold before my mind a proof of God’s existence
(which includes God’s character). Second, I see as a clear and distinct implica-
tion that God is not a deceiver. As a result, I have reason to trust my memorial
seemings; when it seems to me that P, I have reason to believe that P. And,
174 Coppenger
importantly, I have reason to reject any challenge to this inference: God would
not allow me to be deceived on such a massive scale.
My claim at this point is not that Descartes was right. A proof of God’s ex-
istence that avoids the Problem of the Fleeting Present is contentious, at best.
To add to this challenge, maintaining that God’s lack of deception necessarily
(or clearly and distinctly) follows is contentious at best, too. However, despite
these difficulties, the project is at least feasible. In other words, we can under-
stand what Descartes was up to (if this is what he was up to).
Of course, evaluating the success or failure of this project would take us too
far afield of the present discussion. But, ultimately, it is clear that one could
maintain internalism while still trying to avoid the strategy for attack outlined
in this paper.
6.2.2 Epistemic Descent

An entirely different approach to addressing the strategy for attack outlined in
this paper has to do with attempting to relax the constraints on justification
required by internalism without abandoning internalism. While the internal-
ism / externalism distinction is typically construed as a distinction between
two kinds of justification, one might try to develop an account of internalism
that makes sense of ideal justification and then by relaxing the necessary and
sufficient conditions on ideal justification one could develop different kinds
of related but non-ideal justification.22 If this kind of view is successful, then
even if memorial justification cannot meet the standards of ideal justification
it might still be possible to meet the requirements of an account of non-ideal
memorial justification.
6.3 Closing
The project of this chapter has been to motivate a kind of strategy for argu-
ing for skepticism with regard to memorial justification against some different
prominent varieties of internalism. Even if successful, the strategy for attack
outlined in Section 4 should not be viewed as a refutation of internalism. In-
stead, if successful, the strategy for attack should be viewed as an attempt to
outline what I take to be a central threat to a fundamental tenant of inter-
nalism: we commonsensically take ourselves to know a lot, and a lot of that
knowledge depends on memory. But, if internalism is true then either our
commonsensical claims about knowledge are false, or there is work to do in
answering the problems outlined in this chapter.23
22 For approaches of this variety see Fumerton (2004) and Coppenger (2012).
23 Thanks to Kevin McCain for substantive comments on an earlier draft of this paper.
Bibliography
Audi, R. 2011. Epistemology: A Contemporary Introduction to the Theory of Knowledge,

third edition. New York: Routledge.
Bertrand, R. 1921. The Analysis of Mind. London: Allen & Unwin.
Conee, E. & R. Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford
University Press.
Coppenger, B. 2012. “Achieving Epistemic Descent” (Doctoral Dissertation). Iowa City:
University of Iowa.
Coppenger, B. 2016. “Introduction.” In B. Coppenger and M. Bergmann (eds.), Intel-
lectual Assurance: Essays on Traditional Epistemic Internalism, 1–13. Oxford: Oxford
University Press.
Frise, M. 2017a. “Internalism and the Problem of Stored Beliefs,” Erkenntnis 82: 285–304.
Frise, M. 2017b. “Preservationism and the Epistemology of Memory,” The Philosophical
Quarterly 67: 486–507.
Fumerton, R. 1995. Metaepistemology and Skepticism. Lanham, MD: Rowman and
Littlefield.
Fumerton, R. 2004. “Achieving Epistemic Ascent.” In John Greco (ed.), Sosa and His
Critics, 72–85. Malden: Blackwell.
Goldman, A. 1999. “Internalism Exposed,” Journal of Philosophy 96: 271–293.
Huemer, M. 1999. “The Problem of Memory Knowledge,” Pacific Philosophical Quarterly
80: 346–357.
Huemer, M. 2007. “Compassionate Phenomenal Conservatism,” Philosophy and Phe-
nomenological Research 74: 30–55.
Lackey, J. 2005. “Memory as a Generative Epistemic Source,” Philosophy and Phenom-
enological Research 70: 636–658.
Locke, J. 1689. Essay Concerning Human Understanding.
McCain, K. 2015. “Is Forgotten Evidence a Problem for Evidentialism?” The Southern
Journal of Philosophy 53: 471–480.
McGrath, M. 2007. “Memory and Epistemic Conservatism,” Synthese 157: 1–24.
Chapter 10
Inferential Internalism and the Problem of

Unconscious Inference
Richard Fumerton
1 Introduction
Let me start with a truism (it never hurts to get off to a good start).1 The more
conditions one requires for justified belief, the harder it will be to avoid skepti-
cism. And I have often been accused of over-intellectualizing what is plausible
to demand for justified belief. In this paper I’ll try to explain how one might
accept very strong conditions for ideal justification while holding the door
open for a less demanding sort of justified belief, particularly when the belief
in question involves some sort of critical inference.
In a number of places (1995, 2004a, 2004b), I have defended what I have
called the principle of inferential justification (pij). When I was careful (and I
haven’t always been careful), I argued that with respect to ideal propositional
justification, there is epistemic justification for S to believe one proposition P
on the basis of another different proposition E only if 1) S is justified in believ-
ing E and 2) S is justified in believing that E makes probable P (where entail-
ment can be viewed as the upper limit of making probable). Recently (2015),
I have also expressed sympathy for the view that one might relax clause 2) of
the principle to require only awareness of the relevant relation between E and
P (where awareness is something that might not require the kind of concep-
tualization involved in belief). I have also argued that pij (or a close variant
of it) is analytic, and that its analyticity suggests that we should search for a
recursive analysis of justification, one that relies on a base clause employing
the concept of noninferential justification.
One can define inferential justification positively or negatively. We can say
that justification for believing P is inferential when it is at least partially con-
stituted by the justification we have for believing propositions other than P. Or
we can say that justification is inferential when it is not noninferential where
we give a positive account of noninferential justification. Both approaches
1 I would like to thank Ted Poston, Kevin McCain, Mark Walker, and Juan Comesana for helpful
comments on a draft of this paper.

Inferential Internalism & Problem of Unconscious Inference 177
fail to distinguish a number of distinctions one might want to make about the
nature of inference. In “Inferential Appearances,” Michael Huemer (2016) use-
fully distinguishes the following: fully explicit inference, inference against a
background, and noninferential dependence.
According to Huemer (2016: 144), when one engages in fully explicit infer-
ence one “consciously infers P from E alone, and no (non-logical) background
beliefs are required to explain the justification for the inference.” Put in terms
of premises and conclusion, the fully conscious inference involves consciously
entertaining all of the premises critical to reaching the relevant conclusion.
Someone S’s inference against a background also involves fully conscious in-
ference from one proposition, or set of propositions, say E, to another P, but it
is also true that, in some sense, S wouldn’t have made the inference—Huemer
says the inference wouldn’t have “made sense”—but for the fact that S had
other beliefs in the “background.” I’ll say more about this below.
The case of inference that involves noninferential dependence sounds like
a bit of an oxymoron, but the basic idea is that the case is just like inference
against a background but for the fact that no conscious inference at all takes
place. We can still distinguish this case from belief that enjoys noninferential
justification, for the belief in question still wouldn’t have occurred but for cer-
tain background beliefs, and wouldn’t have made epistemic sense, but for the
background beliefs.2
The last two concepts of inference presuppose an understanding of what a
background belief is. The most obvious suggestion is that a background belief
of yours at some time t is a belief of which you are not conscious at t. It is
tempting to try to analyze unconscious belief in terms of dispositions to be-
lieve consciously. As we ordinarily talk, we do ascribe indefinitely many beliefs
to people, the content of which they are usually not entertaining at the time
we make our assertion.
It is tempting, but dangerous, to try to employ conditionals to capture the
idea of unconscious (or as it sometimes called, dispositional) belief. There are
both intuitive and technical problems. On an intuitive level the view doesn’t
2 One might start worrying about what makes it appropriate to describe the movement from
background beliefs to the belief in question a case of inference at all. A full account of what
makes something appropriately described as inference would take us far afield. But if one
thinks that the basing relation is fundamentally causal, it wouldn’t be a stretch to describe
certain causal connections between background beliefs I have and my belief that P as a case
of inference provided that there is the right sort of causal connection between those back-
ground beliefs and my belief that P. Whenever one gives a causal account of anything, one
must deal with the thorny problem of “deviant” causal chains. I’ll say a bit more about what
makes a causal relation plausibly construed as a kind of inference below.
178 fumerton
allow us to capture the distinction between someone’s already being in the

relevant belief state, and someone’s being such that they would, for the first
time, acquire the belief were they to consider various issues or be prompted by
relevant stimuli. So when I describe my colleague Evan as believing that there
are universals, I’m not merely saying that if he were to consider the question
(perhaps for the first time) he would answer it in the affirmative. That might
be true of Evan, but we need to account for the fact that he now believes the
proposition in question. He also believed it yesterday, the day before yesterday,
and, in fact, has believed it for as many years as I have known him.3
The more technical problem was raised by Robert Shope (1979) in a discus-
sion of what he called the conditional fallacy. If we analyze the truth condi-
tions for the ascription of a belief state in terms of subjunctive conditionals,
we will be driven to absurd conclusions. Suppose that Jones is in a deep coma
with no mental life at all. It seems that we can still truly describe Jones as being
such that if he were to consider the question of whether he is in such a coma,
he would believe that he wasn’t. But it surely isn’t correct to suggest that while
in the coma Jones believes that he is not in the coma. The problem seems to be
that the conditions described in the antecedent of the subjunctive “disturb” in
a problematic way the very conditions under which we wanted to ascribe the
belief. The problem in some respects is not unlike the problem (at least on one
interpretation of it) of ascribing simultaneously both location and momentum
to an electron. Any attempt to measure the location disturbs the conditions
that make possible measurement of the momentum and vice versa. A more
mundane example is the problem anthropologists encounter when trying to
study the normal behavior of people in a community when the very presence
of the anthropologist is likely to disturb at least some patterns of behavior that
would otherwise have been exemplified.
One needs a fix for both of these problems. For the first, it might help to
require not only the truth of the relevant subjunctive conditional but also
that the ground of the relevant conditional be caused by the fact that the per-
son in question has once considered the proposition in question (or at least
considered some proposition that entails it in a way that the believer would
quickly grasp). I have also argued elsewhere (forthcoming) for the rather ex-
treme view that we might take a person’s occurrent unconscious belief states
to be just like that person’s conscious belief states but without the awareness
that is consciousness. That presupposes, of course, that one can be in a mental
state without being aware of that state. Freudians, as we know, have long talked
3 For further discussion of the distinction between dispositional beliefs and dispositions to
believe, see Audi (1994).
about unconscious desire, fear, jealousy, love, hatred, and so on, and if the list
makes sense, one should presumably include belief. I used to think that to be
intelligible such talk would need to be again analyzed in terms of subjunctives,
but where the antecedents of the subjunctives describe something other than
mere contemplation of relevant propositional content—whatever it is that
psychiatrists do to people in order to make them “aware” of their subconscious.
But if one understands “conscious mental state” in such a way that the term is
not redundant, the most natural understanding of a conscious mental state
is just a state of which one is conscious—a state to which one bears a certain
relation. One is conscious of a mental state when one is directly aware of the
state. And there is no reason, in principle, why the state couldn’t continue to
exist when one no longer stands in that relation of awareness to the state. It
would, of course, be an empirical question as to whether there are any such
states, and one obviously couldn’t settle the question through introspection!
One might, however, appeal to explanatory considerations to posit the exis-
tence of occurrent (but unconscious) states like love, fear, hatred, and belief.
If we do make a distinction between a disposition to believe some proposi-
tion and an occurrent but unconscious belief, we might add to Huemer’s dis-
tinctions a fourth category of belief that might involve a kind of inference, or
at least that is not clearly noninferential. Dispositions, we may suppose, have
grounds, and one can imagine a person who forms the belief that P, who engag-
es in no conscious inference, and who isn’t even caused to believe P by some
actual occurrent belief of which he was not conscious. That person might, how-
ever, still be caused to believe P by the fact that he has dispositions to believe
various propositions. It is not clear that subjunctive facts (the facts that make
subjunctives true) are the right kinds of facts to stand in causal connections,
but the grounds of such subjunctives certainly are. There is no reason to deny
that I might be so constituted that only if I would believe certain propositions
E were I to consider them (perhaps for the first time) would I also form some
other belief P when prompted by certain stimuli. Of course, the mere existence
of a causal connection might not be enough to suggest that the belief that P
involved inferential justification. It might well be true that if I were not the kind
of person who would believe that 57 is greater than 56, then I wouldn’t be the
kind of person who would believe that 1 + 1 = 2. But it is a stretch to suppose
that the former is in any way relevant to the justification I have for believing
the latter. Still the disposition to form a belief that E whose ground causally
contributes to my belief that P might be accompanied by another disposition
to believe the proposition that P wouldn’t be justified but for the existence of
justification to believe E. That’s the sort of case that might suggest an inferen-
tial justificatory dependence, one that would require that to have justification
180 fumerton
for believing P, one would need justification for believing E. We might call this
dependence on dispositions to believe.
It seems to me that all four categories of belief that involve other beliefs or
dispositions to believe are such that their justification might depend on the
epistemic status of those other beliefs. The case is probably clearest in the case
of fully explicit conscious inference. The first clause of the principle of infer-
ential justification states that one can be justified in believing P on the basis
of E only if one is justified in believing E; garbage in, garbage out. Virtually all
foundationalists, internalists and externalists alike, agree on this proposition.
But even in cases two and three, where there are background beliefs playing
a critical role in belief formation, it is plausible to suppose that those beliefs
had better be justified if the belief they help produce is to be itself justified. At
least that is so on the supposition that the background beliefs were essential to
the relevant belief formation. So to take a controversial example, if my belief
that the object before me is red relies on a background belief that the condi-
tions of perception are “normal,” then there better be good reason for me to
believe that the conditions of perception are, indeed, normal if there is to be
justification for me to believe that the object is red.
Less obviously, perhaps, it also seems to me that even in the case of disposi-
tions to believe that are epistemically and causally relevant to a belief that I
form, it is plausible to suppose that there had better be justification for me
to form the relevant belief described in the consequent of the subjunctive
(whether or not I actually form the belief). So, again, to take a controversial
case, let’s return one more time to that fake barn country so often discussed
in the context of Gettier counterexamples to the traditional account of knowl-
edge as justified true belief.4 I’m driving along in a car, look out the window,
seem to see a barn, and naturally enough form a belief that there is a barn
there. In the example, I happen to see the one real barn in an area where that
barn was surrounded by realistic barn facades (a movie company was filming
there), any one of which would have fooled me into thinking that it was a barn
when it wasn’t. Now I may or may not have a background belief as I look at the
barn that there is nothing unusual about the countryside here. It seems highly
unlikely that I would have an actual background belief to the effect that I am
not in an area where Hollywood set designers have erected a great number of
barn facades. On the other hand, I suspect that I would have the disposition
to form the belief that there aren’t a lot of barn facades around here, were I
to consider the proposition. It may also be true of me that I wouldn’t have
formed the belief that there was a barn there but for the fact that I am the
4 Originally suggested by Carl Ginet and discussed in more detail in Ginet (1988).
kind of person who has that doxastic disposition, or one like it. And it might
further be true of me that I would recognize that my belief about the barn
wouldn’t be justified but for the existence of justification that would support
the relevant false conclusion about the absence of barn facades. Put another
way, there is a ground for these dispositions to believe various propositions
describing the conditions of perception where if one takes that ground away,
one loses justification.
Reflection on these matters might point to a subtly different direction in
which one might search for a solution to Gettier counterexamples. Given
Gettier’s original two counterexamples to the analysis of knowledge as justi-
fied true belief, almost everyone soon considered the “no false lemma” solution
to the problem. The basic idea is that one add to the traditional account of
knowledge a fourth condition, one that states that in order to know that P one’s
belief that P must not only be justified and true, but any other beliefs essential
to one’s reaching the conclusion that P also be justified and true. The fourth
condition as stated doesn’t distinguish among the kinds of belief-dependent
justification we have discussed above, but most epistemologists attracted to
the solution would surely understand the condition as requiring that back-
ground beliefs essential to one’s forming the belief that P also be justified and
true. I’m now suggesting that one might also require that if one wouldn’t have
believed that P but for the fact that one had a grounded disposition to form
a given belief (where one also has a disposition to believe that the content
of that belief is epistemically relevant), then there had better be proposition-
al justification for that belief and the proposition in question had better be
true.5
I’m not foolish enough to claim a solution to the Gettier problem (the epis-
temological landscape is littered with the remains of such failed attempts),
but I do think that the expansion of the no-false-lemma view discussed above
would help deal with examples that were initially thought to pose a problem
for the view, including, of course, the hypothetical situation involving fake
barns. Many of the cases presented as problems for the no-false-lemma view
appeal to situations in which there is a quick move from some prompt to a
conclusion, a move that seemed to involve no conscious consideration of the
relevant falsehoods that made one’s arriving at the truth “lucky.” But phenom-
enology by itself won’t tell us all of the critical beliefs that might be a part of
5 After I wrote this, Ted Poston pointed out to me that there is a very strong connection be-
tween my suggestion here and one made by Bill Lycan (2006). Lycan employs the concept of
tacit assumption, but there may not be much difference between his idea of a tacit assump-
tion, and my idea of a disposition to believe (contrasted with a dispositional belief).
182 fumerton
both the background, and the dispositions to believe, that might be causally
critical to the formation of the justified belief that we don’t want to count as
knowledge.
2 Inferential Internalism and Kinds of Belief-Dependent Justification
Most epistemologists are determined to avoid skepticism, at least skepticism

with respect to the possibility of justifying belief in so-called commonsense
propositions—the beliefs we have about our immediate past, the physical
world, other minds, “uncontroversial” science, and the like. And as Goldman
argued rather persuasively in “Internalism Exposed,” most traditional versions
of internalist foundationalism are going to have a devil of a time avoiding such
skepticism. If we restrict ourselves to a foundation consisting of truths describ-
ing our current conscious states, our foundation will simply be too fragile to bear
the weight of the doxastic edifice it is supposed to support. Matters are made
worse if one insists that in order for someone S to possess inferential justifica-
tion, S must be aware of the relevant evidential connections. Children and most
adults have all sorts of justified beliefs even though they lack the conceptual so-
phistication even to entertain propositions describing evidential connections.
Externalists, by contrast, typically allow for a much broader foundation.
Nor do they insist that when one’s belief is formed as a result of one’s having
other beliefs, the believer must have some sort of access to the legitimacy of
the inference. So, for example, all that Goldman’s original reliabilism requires
for a noninferentially justified belief is that the belief be formed as a result of a
process that is unconditionally reliable and doesn’t take as its “input” other be-
liefs.6 I assume he would take belief-dependent processes to be not only those
that take as “input” other conscious beliefs, but also those that take as input
background beliefs. I don’t know whether he would take a process to be belief-
dependent if it takes as “input” that which grounds a disposition to believe.
Now on a view like Goldman’s it is an entirely empirical question as to which
of our beliefs result from belief-independent processes. As I have pointed out
many times (op. cit.), reliabilism allows for the possibility of any belief’s being
noninferentially justified. On the other hand, if the world is as we take it to
be, I also suspect that rather few beliefs would have that status. Certainly, if
we allow that when the ground of a disposition to believe is relevant to one’s
forming a belief that fact renders the process belief-dependent, then beliefs
that result from belief-independent processes might be few and far between.
6 The view became more complicated to allow for the relevance of potential defeaters.
As I implied earlier, it would surprise me if the beliefs we form about the colors
and shapes of objects immediately before us under ideal conditions of percep-
tion were formed by belief-independent processes. As the moderns were quick
to point out, it is entirely plausible to suppose that lurking as either causally
critical background beliefs or causally critical dispositions to believe are all
sorts of beliefs about internal and external conditions of perception. And as
we argued, these beliefs will need to be justified if the relevant output beliefs
to the belief-dependent processes are to be justified.
Nevertheless, even if the epistemic sailing isn’t completely clear for our re-
liabilist, at least the reliabilist (and this is true of most other externalists) isn’t
burdened by anything like the second clause of our principle of inferential jus-
tification. And that’s a really good thing, because unless an inference involves
conscious awareness of all of the premises one uses to reach a conclusion, how
could one be aware of the relevant connection between premises and con-
clusion? But there do seem to be all sorts of inferentially justified beliefs that
don’t involve that sort of explicit inference. Beliefs we form about the shapes of
objects viewed from all sorts of different perspectives, it seems to me, involve
a host of background assumptions about such conditions of perception as that
we stand in certain spatial relations to those objects.
So if we are inferential internalists, what story can we tell about how people
get justification for their beliefs about the shapes of objects based on their sub-
jective experiences? Well the first part of the story is to remind everyone that
pij’s two conditions on inferential justification are necessary conditions only
for ideal justification. The next step is to insist that the justification for many
of our beliefs falls short of the ideal. We can then allow for various forms of
degenerate justification.
But why engage in such masochistic epistemology? Once we admit that our
common ascriptions of rational or justified belief require far less than ideal
justification, why don’t we simply embrace the more realistic standards for
full-fledged justification embraced by the externalists? We’ll probably end up
characterizing the same set of beliefs as justified. The inferential internalist,
however, will do it begrudgingly insisting on a pejorative characterization of
the justification in question as degenerate.
The first answer to this question is that by starting with the ideal and allow-
ing degenerative justification as we strip away elements of ideal justification
we’ll avoid the new evil demon problem.7 The most common forms of exter-
nalism, reliabilism, for example, struggle mightily to accommodate the very
strong intuition that the victim of demonic machination has justified beliefs
7 I discuss this in more detail in Fumerton (2016).

184 fumerton
(or more modestly, beliefs that have the same epistemic status as do the beliefs
of inhabitants of the actual world—should the actual world be as we take it to
be). I won’t argue the case here, but I think that none of the attempts to save
reliabilism, for example, from the problem work. With respect to inferential in-
ternalism, the inhabitants of the demon world enjoy just as much justification
as do we for believing what they believe.8 Those who have ideal justification
will have that justification. Those who fall short of ideal justification, perhaps
in virtue of failing to bring before consciousness all of the premises critical to
their reasoning, will still have precisely the same justification as their counter-
parts do in the actual world. At least all this is so if one embraces the view that
probabilistic connections, like entailment, are internal relations holding be-
tween propositions. The inferential internalist who embraces an acquaintance
theory of noninferential justification must hold such a view for only by doing
so can one hope to avoid a potentially infinite regress involved in justifying
one’s belief that the relevant probabilistic connections hold.9
There is another reason the inferential internalist insists on awareness of
evidential connections as part of what is required for at least ideal justifica-
tion. It is the desire, as old as philosophy, to link knowledge and justification to
epistemic assurance. Only after one discovers the background beliefs or dispo-
sitions to believe that would justify one’s conclusions and only after one sees
the relevant connections between those truths and one’s conclusion, is one in
a position to understand, to resolve epistemic puzzles to one’s satisfaction. One
can convince people empirically what the right solution is to the Monte Hall
puzzle,10 but only when one “sees” why “switching” is more likely to work does
one’s puzzlement disappear.
But what about degenerate inferential justification (and I would include
among kinds of degenerate inferential justification all but fully explicit con-
scious inference from all relevant premises justifiably believed, where the
believer is also aware of the relevant connection between premises and con-
clusion)? When one reaches a conclusion only in virtue of beliefs that stay
in the background, or perhaps worse, in virtue of whatever grounds certain
8 Of course, if the skeptic is right, neither of us have very much in the way of justification.
9 For an extensive discussion of this point see Fumerton (2004a).
10 The puzzle (supposedly raised by Monte Hall himself) is this. In the game show con-
testants who won prior contests were allowed to choose one of three doors (1, 2, and 3).
Behind one of the doors was a great prize. Behind the others were “gag” prizes worth next
to nothing. After a contestant chose a door, Hall would always open one of the unchosen
doors showing them a gag prize. He would then give the contestant the opportunity to
change the original choice. Most people initially think “sticking” or “switching” are in-
different with respect to the probability of winning the prize. In fact you double your
chances of winning if you switch.
ispositions to believe, one is hardly in a position to become aware of the rel-

d
evant connection between the content of the background beliefs or beliefs
you have a disposition to form, and your “conclusion.” As we admitted, it is
even a bit strained to think of the situation as one in which there was any sort
of “inference” to any sort of “conclusion.”
All that seems right to me. I suppose it is not altogether uncontroversial to
suppose that one cannot become aware of a relation without being aware of
its relata. If a car speeds by me at 100 miles an hour, there is probably a sense
in which I might notice that the car appeared to exemplify a color darker than
light blue. And I might be able to notice that relation even if I couldn’t for my
life tell you which of many different dark colors was the one exemplified by the
car. But it is not clear what one should say about a case like this. The relata of
the darker than relation are colors, and perhaps one should recognize the ex-
istence of determinable as well as determinate colors. If so, one might still dis-
cover a relation holding between some determinable of which one was aware
and the determinate shade of light blue of which one was thinking. One can
probably be aware of one note on the piano being lower than another without
have a very clear awareness of either note, but again the determinate/deter-
minable distinction might be invoked to explain the phenomenon.
In the case of background beliefs of which one is not conscious, however, it is
hard to see how one has any sort of conscious grasp of the relata of critical rela-
tions such as entailment or making probable. One might suppose one is better
off in Huemer’s second case—the case where one does consciously entertain
premises which with the content of background beliefs entail or make prob-
able one’s conclusion, but I’m not sure that the difference makes a difference.
P together with X might entail Q, but if I’m trying to get myself aware of the
entailment and all I do is focus on the content of P and Q, how would I find it?11
3 Conclusion
Degenerate justification, if it exists, doesn’t bring with it the kind of assur-

ance that comes with awareness of evidential connections—the kind of as-
surance that the skeptic seeks. That’s precisely why it is degenerate.12 So why
give the belief that lacks ideal justification any sort of epistemic honorific
11 See Rhoda (2008) for a discussion of how one might try to answer this question.
12 If there is this concept of degenerate justification, there is good reason to believe that we
can develop analogous concepts defined in terms of degenerate justification. So we might
introduce a concept of degenerate knowledge, degenerate practical reasons for actions,
and so on.
186 fumerton
label? It is for the simple reason that where the critical beliefs are in the
background, or even when they are merely dispositional, it might not be that
difficult for the believer to bring them out of the shadows and into conscious
light.13 It might then not be that difficult for the believer to become aware
of the critical connection between the totality of crucial premises and the
“conclusion” that was reached more instinctively without laborious conscious
inference. Being close to having ideal justification isn’t the same as having it.
But it is better than being far away.
Bibliography
Audi, R. 1994. “Dispositional Beliefs and Dispositions to Believe,” Nous 28: 419–434.
Fumerton, R. 1995. Metaepistemology and Skepticism. Boston: Rowman and Littlefield.
Fumerton, R. 2004a. “Epistemic Probability,” Philosophical Issues 14: 149–164.
Fumerton, R. 2004b. In R. Schantz (ed.), “Inferential Internalism and the Presupposi-
tions of Skeptical Argument,” The Externalist Challenge, 157–168. Berlin: De Gruyter.
Fumerton, R. 2005. “Speckled Hens and Objects of Acquaintance,” Philosophical Per-
spectives 19: 121–139.
Fumerton, R. 2015. “What the Internalist Should Say to the Tortoise,” Episteme 12:
209–217.
Fumerton, R. 2016. “Rising Above the Animals: The Search for Intellectual Assurance.”
In M. Fernandez (ed.), Performance Epistemology, 151–166. Oxford University Press.
Fumerton, R. Forthcoming. “Conscious and Unconscious Mental States?” In D. Jacquette
(ed.), Philosophy of Mind: From Antiquity to the Present, Vol 2. London: Bloomsbury.
Ginet, C. 1988. “The Fourth Condition.” In D. Austin (ed.), Philosophical Analysis,
105–117. Dordrecht: Kluwer Academic Publishers.
Goldman, A. 1999. “Internalism Exposed,” Journal of Philosophy 96: 271–293.
Huemer, M. 2016. “Inferential Appearances.” In M. Bergmann & B. Coppenger (eds.),
Intellectual Assurance, 144–164. Oxford University Press.
Lycan, W. 2006. “On the Gettier Problem.” In S. Hetherington (ed.), Epistemology Fu-
tures, 148–168. Oxford University Press.
Rhoda, A. 2008. “Fumerton’s Principle of Inferential Justification, Skepticism, and the
Nature of Inference,” Journal of Philosophical Research 33: 215–234.
Shope, R. 1978. “The Conditional Fallacy in Contemporary Philosophy,” Journal of Phi-
losophy 75: 397–413.
13 Though it certainly might require a philosopher to serve as a mid-wife to give birth to the
relevant ideal justification.
Chapter 11
The Philosopher’s Doom: Unreliable at Truth or

Unreliable at Logic
Bryan Frances
1 Introduction
Vaguely put, I will argue that it is highly probable that we philosophers are
either highly unreliable at finding truth or highly unreliable at logic.1 I don’t
mean to say that we are merely (highly) unreliable at finding deep philosophi-
cal truths, or that we aren’t good at spotting validity in complicated cases. Most
of us who aren’t fools already know that. I mean to argue for something more
damning: that either in our acceptance of commonsensical claims we philoso-
phers are very often mistaken, or we are highly unreliable in identifying valid-
ity in even the simplest cases—the ones we cover in teaching elementary logic
applied to natural language. Either way, most (not all; flatter yourself that you
are among the exceptions) philosophers are doomed.
2 Attempted Philosophical Reductios of Common Sense
By commonsensical claims I mean claims such as “There are trees on Earth,”

“Some people are rich,” “My car has existed for years,” “Modus ponens is truth-
preserving,” “Contradictions aren’t true,” and “There are no two entirely mate-
rial objects occupying the same place at the same time made of the very same
matter.” Roughly put, I use “commonsensical claim” to pick out claims the (a)
clear majority of (b) philosophers think are (c) obviously true. In the final sec-
tion I’ll briefly explore the intuitive idea that some claims are more common-
sensical than others.
An attempted philosophical reductio of common sense: an argument of the
following form in which each C is a commonsensical claim, →← is a contra-
diction, and it seems obvious that the argument is deductively valid: C1; C2;
1 This research is supported by the programme Mobilitas Pluss project mobtt45 and the Cen-
tre of Excellence in Estonian Studies (European Regional Development Fund) and is related
to research project iut20-5 (Estonian Ministry of Education and Research).

188 frances
C3; thus, →←. Different reductios have different numbers of C claims. This is a
mere subset of philosophical arguments against common sense: the ones with
nothing but commonsensical premises. Roughly put, an attempted reductio is
successful when at least one commonsensical claim involved in the reductio is
untrue. I will define “success” more rigorously later in this section.
The sorites paradox is a good example reductio. Here are the commonsensi-
cal C claims that appear to entail a contradiction:
1. Anyone with $0 is not rich.

2. If anyone with $0 isn’t rich, then anyone with $1 isn’t rich.
3. If anyone with $1 isn’t rich, then anyone with $2 isn’t rich.
…
109 + 1. If anyone with $999,999,999 isn’t rich, then anyone with $1,000,000,000
isn’t rich.
109 + 2. There is someone rich with less than or equal to $1,000,000,000.
There are exactly two responses to any given attempted reductio. To see this,
note first that either at least one of the Cs is untrue or all of the Cs are true. If
at least one of the Cs is untrue, then the reductio is successful, since each C is
commonsensical. If the Cs are true and derivation of →← is valid, then →← is
true; so in that case the reductio is successful as well since it is commonsensi-
cal that →← isn’t true. If the Cs are true and derivation is invalid, then there are
just two possibilities:
– Derivation is formed correctly according to the rules of inference, but those

rules are not truth-preserving. Thus, it is a successful reductio, since it’s
commonsensical that elementary inference rules are truth-preserving.
– Derivation is formed incorrectly according to the rules of inference, which
means at least one of the Cs has some hidden semantic structure so that
the elementary inference rules don’t apply in the obvious way to derive →←;
thus, no successful reductio.
When the latter possibility applies we say the attempted reductio is logical-
ly opaque. Thus, with respect to any attempted reductio, there are exactly 2
options:
– Successful reductio: some commonsensical claim is untrue: one of the Cs,

the claim that the derived contradiction is untrue, or the claim that such-
and-such elementary logical principle is truth-preserving.
– Logically opaque reductio: at least one of the Cs has hidden semantic struc-
ture so that the elementary inference rules don’t apply in the obvious way.
The Philosopher’s Doom 189
This enables one to say the Cs are true, the contradiction is untrue, and the
elementary rules of inference are truth-preserving: the Cs don’t entail the
contradiction, contrary to appearances.
In other words: reject some part of common sense or accept logical opacity. We
figured this out without taking any position on the reductio; we used logic
alone. There are no other possibilities.2
There are many philosophical problems that generate the same disjunctive
choice, ones that find their origins in the Statue-Clay problem, the Problem
of the Many, Curry’s paradox, the Grelling-Nelson paradox, the Paradox of the
Knower, the Liar paradox, and the Tibbles-Tib problem (the last two appear in
Section 4 below). The problems aren’t always tied to so-called philosophical
paradoxes; for instance, attempted reductios can be generated from the nor-
mative ethics of resource distribution. All we require is that we can use the
philosophical issue in question to formulate an apparently straightforwardly
valid derivation of a contradiction from just commonsensical claims and using
commonsensical rules of inference. Again, some of the claims will be more com-
monsensical than others.
The logical opacity response says there is some subtle context dependen-
cy or other semantic complexity present in the commonsensical C claims in
question, a semantic complexity that retains the truth of the commonsensical
claims plus the elementary logical principles used in the derivation of the con-
tradiction. For instance, when it comes to the sorites there are various contex-
tualist approaches (Åkerman 2012 is an overview). For those not familiar with
sorites issues: paradoxical sorites arguments need not employ terms with obvi-
ous context dependency such as predicates like “is tall” or “is rich” (e.g., what
counts as tall for a 17-year-old girl isn’t what counts as tall for a 17-year-old girl
on a high school basketball team). It’s well known that soritical arguments
work just as well when employing predicates such as “is a pumpkin,” definite
descriptions such as “the person in the corner,” and even proper names such
as “Bertrand Russell!” In order to embrace the opacity response for the sorites
paradoxes the context dependency has to be subtle and extremely pervasive,
something very far beyond the obvious contrast-class feature of “is rich” and
applying to virtually all natural language sentences. Anyone familiar with the
sorites knows that it is patently inadequate to support the opacity response
with the bland observation that “is rich” is context dependent in the ordinary
ways.
2 Even if there are several kinds of truth, the members of the apparently inconsistent sets are
usually similar enough in topic that they will have the same kind of truth. So I don’t see how
truth pluralism would ruin my argument.
190 frances
3 Why It is So Likely that Several Mutually Independent

Philosophical Reductios are Successful
Thus, we are faced with many attempted philosophical reductios of common

sense. For each one: either the reductio is successful or opaque. Think of each
reductio as a marble on a large bowl. How many marbles are there? And how
diverse are they?
Recall that the main goal of this essay is to establish a disjunctive thesis:
either (1) there is a high likelihood that there are (i) several (ii) mutually in-
dependent successful philosophical reductios, or (2) there is enough hidden
semantic structure in even simple sentences of natural language to make us
highly unreliable at spotting validity in our philosophical work. My argumen-
tative strategy will be the following. I will formulate an assumption P; next,
I will argue that if P is true, then (1) is true; and then I will argue that if P is false,
then (2) is true. The disjunction of (1) and (2) follows.
For fun, and to warm up to the argument, suppose the odds are small that
any given attempted reductio is successful: for each one there’s only a 25%
chance the success option is right. As for the talk about “chance,” I have in
mind our responses to questions such as “What do you think the odds are that
the sample contains magnesium?” (said in the context of a chemistry lab in
college) or “What do you think the odds are that quarks really exist?” (asked
of a physicist). Suppose further that there are twenty mutually independent
attempted reductios. Does that seem reasonable to you?
If so, then a simple calculation shows that there is a 99.7% chance that at
least one of the reductios is successful, which means that there is a 99.7%
chance that there is a successful philosophical reductio of common sense. Of
course, if you think the 25% figure is too low, then the chance of a successful
reductio goes much higher than 99.7%.
But are there twenty mutually independent attempted philosophical
reductios?
On the one hand, philosophers are profoundly territorial. “Here is my ex-
planatory gap argument against physicalism—and it relies on principles that
are crucially different from those found in the works of others!” To a certain
extent, they are right: the offered arguments use distinct premises, ones that
are arguably not metaphysically equivalent (occasionally they aren’t even ma-
terially equivalent). For instance, it’s arguable that there are at least two impor-
tantly different brain-in-a-vat arguments for radical scepticism. And it would
be a mistake to think that the biv arguments aren’t importantly different from
the dreaming arguments or the evil demon ones. Construed this way, there
probably are not twenty but hundreds of attempted philosophical refutations
of common sense. To get a sense of the numbers, if there were 100 of them, and
they were mutually independent of one another, then even if the odds that any
given one of them is successful was a measly 5%, then the odds that at least
one is successful would be about 99%.
On the other hand, it’s a tempting thought that those arguments are not
really independent of one another. For instance, it’s surely intuitive that the
various arguments for radical scepticism, even if employing metaphysically
inequivalent premises, are at some deep level “the same argument.” Similar
thoughts apply to the pot of semantic paradoxes or at least some of the para-
doxes of material composition (e.g., Statue-Clay and Tibbles-Tib). For the sake
of making life difficult for my argument, I will accept this tempting intuition.
On top of that, it’s hardly clear what “independent” means for arguments.
At this point, make assumption P: we can accurately model the situation (of
the multiple attempted reductios) with the assumption M that there are just a
dozen mutually independent philosophical reductios in all of philosophy (not
just metaphysics and the philosophy of logic) and each is only about 25% likely
to be successful—both figures being conservative in my opinion (only twelve
in all of philosophy?). So, there are two assumptions here: P is an assumption
about the accuracy of modeling assumption M. If you think the 25% figure
is far too high, that’s fine! Section 4 of this essay is for you. It simply does not
matter what number we throw in for a probability, at least for the sake of es-
tablishing my disjunctive thesis. So even if you think it’s impossible to offer an
epistemically reasonable number, or that it’s “subjective” or “relative” in some
ways, that’s fine too since it still won’t hurt my argument.
With the 12 and 25% figures the results are as follows:
Chance of at least 1 success: 0.97

Chance of at least 2 successes: 0.84
The argument thus far:
– Claim P: we can accurately model the situation with the assumption M that
there are at least a dozen mutually independent attempted philosophical
reductios of common sense, and for each the odds of success are at least
25%.
3 The numbers can be calculated here: http://stattrek.com/online-calculator/binomial.aspx.

Thanks to Amirhossein Kiani for advice and direction.
192 frances
– Disjunct (1): hence, there is a high likelihood that there are (i) several (ii)
mutually independent successful philosophical reductios.
Again, I don’t care if P is true. If P is false, then we get disjunct (2), as we are
about to see.
We don’t know, with certainty, that any of the reductios are successful.
All I am claiming, at this point in the argument for my first disjunct, this: it’s
highly probable that several mutually independent philosophical reductios of
common sense are successful—on assumption P.
In order to understand the significance of the argument thus far we have
to respond to two objections. The first objection says that even if there is a
successful reductio, it targets a small portion of common sense, so the signifi-
cance is minute. This is mistaken. For instance, if the compositional nihilist’s
argument refutes common sense it doesn’t just show that “There are trees and
tables” is false: virtually all positive, first-order claims about macroscopic ob-
jects end up false. A similar point holds for the sorites and semantic paradoxes.
The attempted reductios hardly have anything special to do with trees, rich
people, or modus ponens. Some of those reductios target an infinity of com-
monsensical claims. In addition, the semantic paradoxes are not limited to bi-
zarre sentences such as “This very sentence is false.” As is well known in the
philosophy of logic, the paradoxes also come from ordinary sentence tokens
(or uses of such tokens) such as “What Harry said about the baseball team just
isn’t true.” Hence, admitting the existence of a successful reductio often means
admitting that an enormous class of commonsensical claims are false. More-
over, the reductios are mutually independent, so most of the time they target
different classes of commonsensical claims.
The second objection comes from the philosopher who is happy to admit
that such-and-such attempted reductio R1 is successful but none of the others,
R2-R12 are. However, as we saw above the odds are reasonably high, 0.61, that
there are at least three successes—and they are mutually independent. Hence,
it’s probable that this philosopher has at least two sets of false commonsensi-
cal beliefs (those corresponding to the successful reductios she missed). One
can happily accept my probabilistic argument only if one is happy to concede
that a “good many”—yes, this has to be vague, given what we said earlier about
argument independence and modeling—attempted reductios are successful.
4 What if Assumption P and (1) are False?
Suppose the “25%” figure in assumption P of Section 3 is wildly inflated, and P is

false. So we are free to embrace common sense, at least apparently. What follows?
If the “25%” figure is wildly inflated, then opacity applies virtually every
time (since those are the only two possibilities: successful or opaque reductio).
So in virtually all those reductios there were semantic complexities pres-
ent that meant that the apparently valid derivation of a contradiction was
merely apparent: in virtually all cases the commonsensical claims don’t have
the semantics they appear to have—even when they appear to be very simple
sentences.
But if that’s right, then the semantic complexities are very well hidden. For
instance, there certainly doesn’t appear to be any semantic hocus-pocus going
on in the sorites claims (recall that such claims can involve terms such as “is a
pumpkin,” “the person in the corner” or “Bertrand Russell,” thereby losing the
obvious contextual dependence of “tall” or “rich”) or the claims that generate
the Liar paradox, the Problem of the Many, the Statue-Clay problem, or the
Tibbles-Tib problem, for just four examples. To see this fact in the flesh, here
is the latter problem (Tib = all of the cat Tibbles except her tail; Tail = her tail;
between 1pm and 2pm Tail is cut off and destroyed).
1. At 1pm Tibbles has Tail as a part.

2. At 1pm Tib does not have Tail as a part.
3. If at 1pm Tibbles and Tib don’t have the very same parts, then Tibbles ≠
Tib at 1pm.
4. If Tibbles ≠ Tib at 1pm, then Tibbles ≠ Tib at 2pm.
5. At 2pm Tibbles and Tib are made up of the exact same matter in the exact
same place.
6. Tib and Tibbles are entirely material objects.
7. There are no two entirely material objects occupying the same place at
the same time made of the very same matter.
We definitely know this with regard to the set of (1)–(7): either opacity applies
or one of these is untrue:
– At least one of (1)–(7)

– Elementary logic is truth-preserving
– Law of noncontradiction
However, there certainly doesn’t appear to be any context dependency here

that saves the consistency of (1)–(7). Keep in mind that context dependency
doesn’t matter here unless it saves the consistency! I’m not saying that the ap-
pearances are accurate. All I’m doing is making the obvious point that if they
aren’t accurate, then the truth-conditional semantics of (1)–(7) is very subtly
context dependent. And notice that it’s not merely that we are wrong about
194 frances
details of logical form: we are so wrong that the “obvious” inconsistency—the

one that led us to focus on the problem in the first place—is no inconsistency at all.
A similar point holds for the liar paradox, with the set of (1)–(4) appearing
inconsistent:
1. S is identical to “S isn’t true.”

2. If “S isn’t true” is true, then S isn’t true. (Cf. if “Dogs bark” is true, then dogs
bark.)
3. If S isn’t true, then “S isn’t true” is true. (Cf. if dogs bark, then “Dogs bark”
is true.)
4. It's not the case that S is true and not true.
For this attempted reductio we know that either opacity applies or one of these
is untrue:
– (1)
– Elementary logic is truth-preserving
– All instances of T-schema are true (that’s (2) and (3))
– Law of noncontradiction
But then how do we know our philosophical arguments are valid? Suppose you
work for years and come up with a beautiful argument for some interesting
thesis. But you now have insufficient reason to think your argument is valid.
Your clever opponent can always say in response to your beautiful argument
“Well, I admit that your argument against my position looks awfully good, as
your premises are highly plausible, but given what we have discovered about
the massive pervasiveness of hidden semantic complexity, I’ll bet the prem-
ises are true and the conclusion is false despite the fact that it certainly looks
as though your premises entail your conclusion.” What now? Keep in mind
that you are unreliable at spotting validity even in simple sentences such as the
sorites claims or the claims involved in Tibbles-Tib or the Liar—not difficult
cases such as those involving belief sentences, intensional sentences, etc.
Notice that under the opacity option we are very probably highly unreliable
about semantics. For instance, if we are wrong about the semantics of the Cs
in the sorites, Tibbles-Tib, or Liar sentences, then we will be wrong about the
semantics of a great many sentences we use in philosophical arguments, since
so many are structurally similar to those.
And how do you propose to teach logic from now on, knowing that you are
unreliable at figuring out the logical forms of simple sentences?
One response is the highly confident one: “I know what the semantic com-
plexities are that ruin all the attempted philosophical refutations of common
sense, and they don’t ruin my arguments since I know how to sidestep them.”
It must be pleasant to live with such supreme confidence in one’s ability to
discover what has eluded virtually everyone else.
But suppose you really are the exception. That’s nice for you, but it is im-
plausible in the extreme to hold that the philosophical community as a whole
has figured out what you discovered. After all, most philosophers think the
apparently inconsistent sets really are inconsistent.
5 The Consequences of the Disjunctive Conclusion
There are two options: either (1) is true, so the odds are high that there are
several mutually independent successful philosophical reductios of utterly
commonsensical claims such as “There are trees,” or (2) is true, and we are bad
at spotting validity in even very simple cases of natural language—including
the ones in our own philosophical arguments. I won’t pause to consider self-
application (e.g., self-defeat) issues brought up by my argument.
One might think that if disjunct (2) is true, then disjunct (1) is true. After all,
if an argument shows that we are bad at spotting validity and other logical no-
tions in simple cases, as disjunct (2) has it, then that argument also shows that
some of our commonsensical ideas about semantics are false. Hence, I have
really argued that disjunct (1) is true: it’s highly probable that there are several
mutually independent successful philosophical reductios of common sense,
whether or not assumption P is true.4
The only reason I hesitate to endorse this argument is that I have doubts
about how “commonsensical” judgments amongst philosophers are about
truth-conditional semantics (in the sense articulated at the beginning of Sec-
tion 2 above). I won’t explore that empirical matter here.
The truth of the disjunctive thesis has interesting epistemic implications for
philosophical practice. A great many philosophers are so confident in the truth
of certain claims that they will dismiss arguments against them even without
any serious investigation of those arguments. If someone offers an argument
that I don’t have a brain or that there are no prime numbers, well, so much
for her argument: I can reasonably conclude it’s unsound—either invalid or
including false premises—even before I evaluate it.
Some of these philosophers have this attitude regardless of the kind of anti-
commonsensical argument involved; others restrict their dismissal to philosoph-
ical arguments. The latter philosophers will not dismiss out of hand obviously
4 Thanks to Ted Poston, Henrik Sova, Amirhossein Kiani, and others for comments along this
line.
196 frances
empirical arguments coming from our best scientific investigations. But when
it comes to philosophy, they make what we will call the Moorean move:
When you encounter what you know to be a philosophical argument that

goes radically against common sense, an epistemically rational thing to
do is retain your commonsensical belief and conclude that the argument
is unsound—and this can be done independently of finding specific fault
in the argument (which is not to say that you take yourself to have failed
to find such fault).
G.E. Moore (1925), David Lewis (1973), Kit Fine (2001), Anil Gupta (2006),
William Lycan (2001), Jonathan Schaffer (2009), and Thomas Kelly (2008)
are representative philosophers who make the move and defend or assume
its epistemic rationality; one also sees the move made a great many times in
philosophical conversation.
The Moorean move often gets made with respect to arguments for exter-
nal world scepticism. However, one cannot reasonably restrict the Moorean
move to those arguments—and in practice it certainly isn’t so restricted; con-
temporary metaphysics provides examples. It would be bizarre to be confident
enough in “He knows he has hands” to make the Moorean move but not con-
fident enough to make the move for other commonsensical claims—such as
“He has hands.”
When it comes to the class of attempted philosophical reductios I’m focus-
ing on, in order to save common sense opacity must apply every time. Since each
member of {C1, C2, C3, ~→←} is commonsensical for our class of reductios, by
definition, if one accepts the logical inconsistency—as Mooreans often do—
then as we saw earlier one is forced to either give up the truth of one of those
members, accept the truth of some contradictions, or reject the truth-preserving
nature of some elementary logical principles (the ones used in the derivation of
a contradiction from the set). Hence, if common sense is true, as the Mooreans
think it is, then opacity has to apply to each of those attempted reductios.
Hence, if you are the type of philosopher who thinks that philosophy rarely,
if ever, refutes common sense, then you are forced to posit pervasive logical
opacity in our philosophical arguments.
There is an additional problem with the Moorean move, one that applies
to anyone who admits that at least some anti-commonsensical philosophical
arguments are sound, whether or not they use at least one premise that doesn’t
rise to the level of being commonsensical. If there are sound philosophical re-
ductios of common sense, then the Moorean move is flawed, as it recommends
rejecting the soundness of all those arguments. But how bad would the unreli-
ability of the Moorean move be?
All we have thus far is the idea that several of the attempted reductios
are successful. Suppose for the moment that you thought that there were
only around a dozen of them in all of philosophy and only two or three were
successful—clearly optimistic assumptions. Would making the Moorean move
to retain common sense still be reasonable for you?
It may seem so: just because you think the move isn’t guaranteed to avoid
the dismissal of a successful reductio of common sense—because you think
it fails twice, say—hardly means one should give up the move. After all, if an
action type works 70–80% of the time, say, then for most action types that is
a pretty good percentage and one should stick with it even knowing it is less
than perfectly reliable.
However, things are otherwise for the action type under consideration, the
Moorean move. To see this, suppose you came to believe that at least one of
these utterly ordinary and familiar claims is untrue and the corresponding
philosophical reductio against it is successful:
There are trees on Earth

There are people who are rich
If an ordinary sentence is true, then it’s not false too
Modus ponens is truth-preserving
Under that supposition would you be at all comfortable continuing to make

the Moorean move—even when applied to philosophical arguments that rely
on premises that aren’t commonsensical? I would think that a successful refu-
tation of even one of those claims, or other comparable claims, would be in-
credibly damaging to the status of common sense. As pointed out earlier, many
of the attempted reductios target a large class of commonsensical claims; the
reductios hardly have anything special to do with trees, rich people, or modus
ponens. If I can be wrong about the existence of trees, or rich people, or the
truth-preservingness of modus ponens, then how can I be confident in just
about anything from common sense? If common sense is wrong about any one
of those (and the millions of others just like them), it is no minor flaw.
That’s why I think that once one has admitted that several philosophical
reductios are successful—even a small number—we cannot reasonably con-
tinue making the Moorean move. However, this thesis is not as important as
the disjunctive thesis or the thesis that the Moorean move requires opacity
across the board and embraces (2).
As a point of clarification, I’m not saying that if one continues to make the
Moorean move, then one’s retained belief in the commonsensical claim the
argument targets is unreasonable. Instead, I’m saying that a certain cognitive
move, the Moorean one, will no longer be reasonable. It may turn out that the
198 frances
retained belief in the commonsensical claim is unreasonable as well; in fact,

that seems to me to be likely. But I am not focusing on that separate issue.
In the remainder of this essay I will briefly consider the separate conse-
quences of each disjunct of my thesis.
If (1) is true, then two things are probably true. First, it means that common
sense is not a reliable guide to truth. Hence, we would have to find another
guide to truth or substitute something else for truth as a goal. Second, much of
the hand-wringing over semantics that one sees in metaontology and related
areas of philosophy looks like a waste of time. The main reason behind the
hand-wringing has been the idea that the debates over outlandish theses such
as compositional nihilism can’t really be debates over whether there are trees.
When the nihilists say “There are no trees” it is commonly thought we must
slap on some non-straightforward interpretation—because, it is thought, no
one reasonable could really be suggesting there are no trees. Nihilists and near-
nihilists frequently encourage this move. Similarly, when anti-realists about
numbers say “There are no numbers,” we should not interpret them straight-
forwardly, since no competent philosopher would reject either “There are
more than ten prime numbers between 1 and 100” or “If there are more than ten
prime numbers between 1 and 100, then there are numbers.” Another example:
color eliminativists aren’t really saying that tomatoes aren’t red or any other
color. But now that we know that several philosophical reductios, that are mu-
tually independent, refute common sense, as disjunct (1) says, why bother with
such semantic complexities—especially when there has never been much in
the way of decent independent linguistic evidence for the semantic shenani-
gans posited by metaontologists?
On the other hand, if (2) is true, as the Mooreans as well as others would
have us believe, then we have something akin to a full-scale emergency on our
hands: we now know that most of us are highly unreliable in spotting validity
and related logical notions in ordinary, perfectly straightforward, and simple
discourse. If that’s true, then as mentioned earlier it’s difficult to see how we
are to continue arguing with one another when our conversational partners
are often if not always free to accept all our premises while denying our conclu-
sions even when the argument seems as straightforwardly valid as they come.
The right response to this predicament seems to be that the philosophical
community should make a colossal shift to the philosophy of language so we
can learn more about truth-conditionally relevant semantic complexities that
are so well hidden.
As I foreshadowed earlier, I see just one way around the significance of this
disjunctive conclusion, assuming it’s true.5 Notice that claims such as “There
5 Thanks to Alex Davis for making this point as well.

are trees in my backyard” are more commonsensical than “There are no two
entirely material objects occupying the same place at the same time made
of the very same matter” or “If the sentence ‘The sentence in room 101 isn’t
true’ is true, then the sentence in room 101 isn’t true” (the latter is a typical
premise in liar-type arguments for contradictions, coming from one half of the
T-schema). They are more commonsensical in two respects: a higher percent-
age of philosophers who understand them will accept them, and many more
philosophers understand them. (I am not sure of the first respect, but never
mind; let’s give the objection a fighting chance.) Now suppose we discover
that the false commonsensical claims are the ones that are, by a large margin,
the least commonsensical; hence, the damage to common sense is only slight.
With this d iscovery we would be in a position to conclude that although some
philosophical reductios of common sense are sound, the false commonsen-
sical claims are not terribly commonsensical. Furthermore, perhaps opacity
is true only for sentences that we already had a pretty good idea were much
more semantically complicated than we once thought (e.g., those of the form
“S knows that P”).
I suppose this damage-control outcome is possible, but I know of no reason
to be so optimistic.
Bibliography
Åkerman, J. 2012. “Contextualist Theories of Vagueness,” Philosophy Compass 7:

470–480.
Gupta, A. 2006. Empiricism and Experience. Oxford: Oxford University Press.
Fine, K. 2001. “The Question of Realism,” The Philosophers’ Imprint 1 (1).
Kelly, T. 2008. “Common Sense as Evidence: Against Revisionary Ontology and Skepti-
cism.” In P. French and H. Wettstein (eds.), Midwest Studies in Philosophy: Truth and
Its Deformities: 53–78.
Lewis, D. 1973. Counterfactuals. Cambridge, MA: Harvard University Press.
Lycan, W. 2001. “Moore against the New Skeptics,” Philosophical Studies 103: 35–53.
Moore, G.E. 1925. “A Defence of Common Sense.” In J. Muirhead (ed.), Contemporary
British Philosophy (2nd series), 192–233. London: George Allen & Unwin.
Schaffer, J. 2009. “On What Grounds What.” In D.J. Chalmers, D. Manley, and
R. Wasserman (eds.), Metametaphysics: New Essays on the Foundations of Ontology,
347–383. Oxford University Press.
Part iii
New Responses to Skepticism
∵
Chapter 12
The Skill Model: a Response to Skepticism about

the Intellectual Virtues
Sarah Wright
1 Introduction
My aim in this chapter is to motivate the use of the skill model of virtue within
responsibilist virtue epistemology. The analogy between skills and virtues has
been a part of virtue ethics since the ancient Greeks. I will explore the ways
that that skill analogy has been used within virtue reliabilism; it is a tool that
has provided the reliabilist with a response to two different skeptical worries.
I then turn to virtue responsibilism and ask why its intellectual virtues have
not been modeled on skills. I look to the arguments collected and put forward
by Linda Zagzebski as the primary impediments to the use of the skill model
in responsibilism. I then explore Julia Annas’ responses to these arguments
and explain her recently developed positive account of the moral virtues as a
special kind of skill. Annas’ skill model is aimed at the moral virtues; since the
responsibilists’ intellectual virtues are analogous to the moral virtues, the skill
model of virtues is well suited to be applied to the intellectual virtues. I then
demonstrate the strengths of this view, showing how it can avoid the particular
kinds of skepticism relevant to responsibilist virtue epistemology, including
the challenge of situationism.
2 Virtue Reliabilism
Virtue reliabilism is an approach to virtue epistemology which bears a clear re-

semblance to reliabilism in general. Reliabilist intellectual virtues are disposi-
tions or traits of the agent that reliably produce true beliefs. Virtue reliabilism
differs from reliabilism in general by requiring that the relevant true-belief-
producing dispositions be integrated into the agent; fleeting dispositions are
not considered virtues.1 However the range of virtues recognized by this view
1 Sosa (1991) and Greco (2000).

204 Wright
are still broad; both in-born abilities like eyesight and hard-won character
traits like open-mindedness are candidates to be intellectual virtues.
Reliabilist intellectual virtues are modeled very closely on skills. This is made
apparent by the fact that John Greco now classifies them as a special case of
ability; Ernest Sosa now simply talks about beliefs that are “adroit,” where be-
ing adroit come from being grounded in a skill. And skills are a natural model
for reliabilist virtues. Skills are marked by the ability to achieve some desired
end and to do so reliably. The close analogy with skills is also revealed when
we look at the examples that are used. Sosa repeatedly compares intellectual
virtues with the skill of archery; he also says that epistemic normativity should
be modeled on performance normativity, where we evaluate an action on the
basis of its successful display of skill.2 Greco often uses comparisons with skill
in sport, particularly in baseball. He also says that knowledge is a special case
of the general category of success through ability. The analogy with skill allows
both authors to explain how we can possess and develop our intellectual abili-
ties; it also provides a way to talk about the value of the intellectual virtues. Just
as skills are valuable in light of what they produce when exercised correctly,
intellectual virtues can be evaluated for the success they achieve in producing
true beliefs.
2.1 Virtue Reliabilism and Knowledge Skepticism

Virtue reliabilism has been developed to address skepticism about knowledge
from its very inception. In his landmark article “The Raft and the Pyramid.”
Sosa drew his motivation for introducing virtue epistemology from the seem-
ingly intractable disagreement between foundationalism and coherentism.
Each account of justification was motivated, in part, by the concern that fail-
ures of the opposing view might lead to skepticism. Sosa’s diagnosis was that
both approaches to justification have a shared assumption – only states with
propositional content can serve as the justifier for our propositional beliefs.
This assumption artificially limits the range of possible justifiers, courting
skepticism. Sosa proposed a rejection of this assumption, allowing instead that
justification may be grounded in features of the believers, particularly their
(reliabilist) intellectual virtues. Permitting the virtues of the believer to serve
as a justifier widens the sources of justification, providing a stronger basis from
which to address skepticism.
Greco’s development of his “agent reliabilism” was also motivated specifi-
cally as a response to skepticism. He diagnosed a source of persistent skeptical
worries when we put together Cartesian skepticism, recognizing a gap b etween
2 Sosa (2007) and (2011).

THE SKILL MODEL 205
our evidence and the external world, with Humean skepticism, calling into
doubt any possibility of bridging this gap through reason. Greco’s solution
was to recognize that while reason cannot close this gap, it can be bridged by,
“stable and reliable dispositions that make up one’s cognitive character” (2000:
177). Thus Greco’s introduction of reliabilist virtues was also a response to
skepticism, in this case skepticism about our knowledge of the external world.
2.2 Virtue Reliabilism and Virtue Skepticism

Thinking about virtues as skills is also used by virtue reliabilists to avoid skepti-
cism about the intellectual virtues themselves. If virtues must reliably lead to
true beliefs, one might wonder if there are any such virtues. Won’t any disposi-
tion, however reliable in ordinary situations, become unreliable in in adverse
circumstances, like that of deception, drunkenness, or just bad lighting? If our
best dispositions fail in these circumstances, perhaps we don’t have any dis-
positions worthy of being called virtues. This is particularly problematic for
virtue reliabilism, as skepticism about the virtues will lead to skepticism about
knowledge on this view. For virtue reliabilists characterize knowledge as true
belief reached through the exercise of the intellectual virtues.3 Anyone who
lacks virtues, lacks knowledge on this view.
But seeing virtues as skills allows us to recognize the limits of what we
should ask from them. Sosa is careful to remind us that we should not expect
too much of any virtue. Though they can be expected to produce true beliefs
in a reliable manner, this production is always relative to a particular normal
environment. Even the best eyesight can be deceived through colored lights or
special effects; the specific conditions of perception matter. An early definition
of intellectual virtue proposed by Sosa takes this into account, restricting the
range across which a virtue will be expected to produce true beliefs:
A subject S’s intellectual virtue V relative to an “environment” E may be

defined as S’s disposition to believe correctly propositions in the field F
relative to which S stands in conditions C, in “environment” E.4
This proposal embeds the central feature of reliability in reaching the truth. But
it does not require that one can succeed in any conditions or any environment.
This restriction is similar to one that we would expect in a characterization
of skill in general: skills help us to reach our aims in normal situations. But no
skill is robust enough to be successful in all environments. Sosa notes that we
3 Greco (2010).
4 Sosa (1991: 140).
206 Wright
expect a skilled archer to reliably hit the target in ordinary competition condi-
tions. But we do not expect an archer to be reliable in very windy conditions or
when she has been slipped a drug.5 Her unreliability in these aberrant condi-
tions does not undermine her skill of archery. The skill analogy makes it plau-
sible that all virtues, like skills, must be relative to a set of normal conditions.
Such a restriction makes it far more likely that there are dispositions that meet
the reliabilists’ requirement for intellectual virtues.
3 Virtue Reponsibilism
In contrast to virtue reliabilism, virtue responsibilism is an approach to vir-

tue epistemology which focuses on analogies between the moral and the in-
tellectual virtues. While virtue responsibilism has been developed in different
ways by different authors, these accounts share a focus in developed traits of
intellectual character that in some way reflect on the evaluation of the person
possessing them. While virtue reliabilists search broadly for all the cognitively
integrated dispositions of the individual which reliably produce true beliefs,
responsibilists focus more narrowly on developed traits of intellectual charac-
ter. This narrow focus fits with the responsibilists’ idea that intellectual virtues
will (like moral virtues) reflect positively on the person who has them. While
good vision might be quite useful, and while it might reliably lead to true be-
liefs, we do not think that the person with good eyesight is a better person
than someone with poor eyesight. As a result, virtue responsibilists will reject
eyesight as a virtue. Open-mindedness on the other hand is a disposition in an
individual that plausibly has to be developed and maintained over time. It also
reflects on the evaluation of the person possessing it; an open-minded person
is a better person than someone who is closed-minded. Open-mindedness,6
intellectual courage,7 and intellectual humility8 are among the character traits
that have been considered as intellectual virtues by responsibilists.
3.1 Why Have Responsibilists Rejected the Skill Analogy?

Since the skill analogy is widely used in reliabilist virtue epistemology, we might
wonder why responsibilism doesn’t take the same tack. This rejection of the
skill analogy seems even more puzzling when we consider that r esponsibilists
5 Sosa (2007).
6 Baehr (2011) and Riggs (2010).
7 Zagzebski (1996) and Baehr (2011).
8 Roberts and Wood (2007) and Whitcomb et al. (2015).
THE SKILL MODEL 207
tend to model themselves on the kinds of virtue ethics promoted by the An-
cient Greeks. Within ancient Greek discussions of virtue, it was widely held
that virtue was a skill or techne.9 This is why we see Socrates in Plato’s dia-
logues using analogies between virtue and shipbuilding, housebuilding, and
horse husbandry. The Stoics frequently use an analogy between the virtues and
the skill of archery. However our contemporary view of ancient virtue ethics
is biased by our tendency to focus on Aristotle. He argued against the claim
that virtues are skills; but that position made him a unique voice in ancient
debates about virtue.10 Aristotle’s rejection of virtues as skills has recently been
endorsed and further developed by Linda Zagzebski.
3.2 Zagzebski’s Arguments against the Skill Model of Virtue

The first argument Zagzebski considers comes from Sarah Broadie as she in-
terprets and expands on Aristotle’s argument for a difference between skills
and virtues. Broadie notes that we evaluate the outputs of skill and virtues dif-
ferently. When we evaluate a skill, we focus on the thing that is produced (Ar-
istotle’s example is of a grammatical sentence); so long as the product is good
(the sentence is grammatical) we are satisfied. But this focus on outcomes is
not reflected in our evaluation of virtuous actions; we also want the right ac-
tion to be produced by a virtue in the person acting. Right action that comes
from virtuous motivation and from a virtuous character is more valuable than
accidental right action. This difference in the way we evaluate them appears to
be a systemic difference between skills and virtues.
Zagzebski next looks at Philippa Foot’s claim that skills, but not virtues,
can be unexercised capacities. A skilled archer might choose not to engage in
archery, even when the circumstances provide the opportunity to do so. But
virtues are not like this. Truly possessing a virtue will lead us to exercise that
virtue whenever the circumstances provide an opportunity to do so. When a
generous person finds themselves in a situation when they can help someone
in need, we expect them to act. If they do not, we take this as evidence that
someone we thought to be generous doesn’t really have the relevant virtue.
The fact the virtue will not be an unexercised capacity marks it off from other
skills which can be.
9 Paul Bloomfield (2000) states that “There was once a philosophical consensus that virtues
are skills.”
10 Paul Bloomfiled (2000) and Julias Annas (1995b, 2003) have highlighted that Aristotle is
unique among ancient Greek philosophers in rejecting the claim that virtues are a kind
of skill (technê).
208 Wright
Zagzebski’s third argument depends on contrasting the value of the exercise

of skill and the value of the exercise of virtues.11 We value virtues intrinsically;
whenever they are exercised they are good. Skills, on the other hand, can be
used for either good or bad purposes. The skill of medicine can be used to heal
someone who is sick; but it can also be used to subtly poison a rival in an un-
detectable way. In order to evaluate the goodness or badness of the particular
exercise of a skill we must look at its results. Thus the value of a skill is instru-
mental, not intrinsic as the value of the virtues is.
Finally, Zagzebski cites an argument from James Wallace, who points out
that skills are closely associated with particular techniques.12 While these tech-
niques may be difficult to master, they are not flexible in the way that virtues
should be. Virtues, on the other hand are not technically difficult, even if they
are hard to exercise; they do not require that one learn specific techniques.
The technique-dependent nature of skills further marks them off from virtues.
Zagzebski concludes from these arguments that we must distinguish between
virtues and skills. Her foundational role in responsibilist virtue epistemology
has led to a widespread acceptance of the skill/virtue distinction and a rejec-
tion of the skill model of virtue within responsibilism.
3.3 Annas’ Arguments for the Skill Model of Virtue

This rejection is unfortunate, particularly since Julia Annas has argued that
virtues are best understood as a kind of skill, and she has addressed the argu-
ments that Zagzebski gives for making a distinction between skills and virtues,
finding them wanting. She notes that we should be careful about the structure
of any argument purporting to show that virtue is not a skill. Annas (2003b: 17)
notes that:
[T]he thesis that virtue is a skill is a claim that virtue is one kind of skill,
and thus that the idea of skill is central in helping us to understand what
virtue is. Against this claim, pointing out obvious differences between
virtues and skills is ineffective.
Rather than looking only for some differences between skills and virtues, we
need to look at similarities that can show virtues to be a specific kind of skill.
A search for similarities between virtues and a kind of skill offers a response
to all of the arguments presented by Zagzebski, but particularly to Broadie’s.
Perhaps the way that we assess and evaluate the outcome of some skills does
focus on the outcomes alone; there are times we simply want the thing that
11 Zagzebski (1996: 116).

12 Zagzebski (1996: 108).
THE SKILL MODEL 209
the skill produces.13 If I need bread, I might not be concerned with the general
bread making skills of the baker; I might only care that I get some well-baked
bread. But there are other instances when the outcome of a skill is valued both
for its intrinsic properties and for the way it was made. I might value a loaf of
bread made at a favorite local bakery more highly simply because of where
and how it was baked. Of course I might choose a local bakery because of the
quality of its products, but once I value that bakery, learn about its history,
and meet the bakers, I may value it for more than the quality of the bread
it produces. I might prefer the locally made bread to one made at a bakery I
don’t know – and I might do so even if I don’t have an expert enough palate to
distinguish between the bread made at my bakery and that made at another.
If my valuing in this way is reasonable, then the extra value that I assign to the
products of my local bakery cannot come from the value of the output of their
skill alone; rather I value the bread as a product of my local baker’s skill. This
re-establishes the parallel between virtue and a kind of skill. While some skills
mays be valued for their products alone, in others we think that the value of
the skill adds to the value of the product.
In responding to the other arguments, Annas makes use of the Stoic claim
that virtues are the “skills of living.” This means that virtues are a kind of “glob-
al expertise in your life” (2003b: 19). In contrast, we might note that most skills
are “local;” they are only needed in a particular time and place. The local skill
of violin playing may be called for at a concert, but not while waiting in line at
the bank. Still, we want to say that the skill remains, even in times that it is not
called for. But the skill of living is always called for. As a skill with global scope,
we expect it to never be unexercised. So, contra Foot, there are skills that are
always exercised. This is a real difference between local and global skills, but
it does not demonstrate that virtues cannot be a kind of skill; we have only
learned that virtues cannot be a local skill.
Thinking of virtues as the skill of living provides a response to the argument
based on the claim that virtues have intrinsic value, while skills only have in-
strumental value, being usable for good or bad purposes. If virtue is the skill of
living, then any exercise of the virtues will be a move in the direction of a life
well-lived or eudaimonia. Most other skills will not come with this guarantee,
simply because their ends are not similarly determined; medicine can be used
for either the good end of healing or the bad end of harming. In contrast, living
well is our final end (telos) and it is never bad. So the skill of living is intrinsi-
cally good, but this feature doesn’t follow from a difference between virtues
13 In Aristotle’s terms I might simply desire the ergon of the craft. But, contra Aristotle, I am
arguing here that we value a craft for more than the ergon that it produces. (Thanks to Ted
Poston for pointing out this connection.)
210 Wright
and skills, only from the fact that the skill of living has a pre-determined end
of eudaimonia.
3.4 Annas’ Development of the Skill Model of Virtue

Recognizing virtues as a kind of skill, and in particular as the skill of living,
provides Annas with the grounds to reject a principled distinction between
skills and virtues. In responding to Zagzebski, Annas overturns negative argu-
ments against the view of virtues as skills. In her most recent book, Annas goes
far beyond this critique, painting a positive picture of the moral virtues as a
particular kind of skilled expertise.14 The arguments in this book support the
idea that the skill analogy is a fruitful way to explore and understand the moral
virtues. Since the responsibilist virtue epistemologist takes there to be at least
a strong analogy between the moral and intellectual virtues, Annas’ arguments
give us a reason to consider the skill analogy in responsibilist virtue epistemol-
ogy as well.
What sort of skills are the virtues on Annas’ view? They are characteristic
dispositions; they are persistent, active, developed and developing. These are
all traditional features of the virtues. Annas also insists that they are not mere
habits. A habit can be mindless and routine; virtues are active and flexible.
This flexibility is similar to that displayed by those who possess expert practi-
cal skills (or technê). Annas’ recurring example is that of an expert pianist. Her
playing is not a mere habit, but expresses her thoughts about the piece she
is playing; she may play the same piece differently at different times as she
focuses on one or the other aspect of it. In addition, her playing is sensitive to
the situation in which she finds herself. She may play differently in a large or
small concert hall, in response to an awkward piano, or when she senses her
audience is getting bored.
3.5 Two Aspects of Virtue: the Need to Learn and the Drive to Aspire
Annas notes that the sort of expert practical skills which are the basis for her
model of the virtues also have two features in common. The first is what she
calls the “need to learn.” The beginner who is learning a skill might begin by
mimicking the expert blindly or by following some rules of thumb. However in
order to acquire a skill, the learner must come to an understanding of why the
expert is making her choices. This understanding is required in order to devel-
op the skill in oneself; no amount of repetition or copying will take the learner
to the level of expert skill without understanding. In the ordinary course of
learning, this understanding is conveyed from teacher to student by the giving
14 Annas (2011).
THE SKILL MODEL 211
of reasons. Training in a skill aims to make the learner sensitive to the same
kinds of reasons that the expert is using. The expert reacts to a variety of rea-
sons in a particular instance. Those observing the expert can often only see her
behavior; they may miss the real reason behind the expert action. But other
experts will recognize the same reasons and better understand the observed
expert action. So the “need to learn” requirement given by Annas ensures that
the virtuous person will be sensitive to the correct reasons and will be able to
articulate those reasons in teaching others.
This requirement of being able to articulate reasons is perhaps the most
controversial element of Annas’s account; one might think that it is possible
to be skilled without being able to articulate the reasons that one is sensitive
to in the exercise of that skill.15 However, Annas rejects the claim that such
a person is a real expert. In doing this she is following Plato. In the Gorgias,
Plato argues that virtues, like expert skills, require understanding of the under-
lying structure of the craft; Plato contrasts this with simply having a “knack”
for bringing about a result. Flattery, he says, is not an expertise because it lacks
any underlying principles; rather than understanding any deep pattern, the
flatterer simply figures out how to please different people.16 The lack of articu-
lable underlying principles also explains why flattery is not teachable. But ex-
pert skills and virtues must be teachable. Annas notes in her earlier work that
teachability was required for both skills and virtues in ancient Greek thought.17
The second feature of the expert practical skills that inform Annas’ model
of the virtues is that they imbed a “drive to aspire” (2011: 16). When someone is
committed to developing these skills, they do not rest on their laurels. Instead,
if someone values piano expertise, they are driven to maintain and develop
that expertise. Of course, not all learners are like this. Annas admits that the
desire to aspire is a demanding feature of expertise, and so will not be pos-
sessed by all who are trying to learn. Those who do not have the appropriate
drive will “lapse into simple repetition and routine” (Annas 2011: 18). Routine
may be good enough to get the job done in some cases, but it is not associ-
ated with skilled expertise. The expert, in desiring to be an expert, desires to
improve her skills as she can. This may be very difficult, but it is required for
true expertise.
It is here that we finally have the required elements to reply to the earlier
argument from Wallace that virtues cannot be skills because skills are tied to
15 Thanks to Kevin McCain for raising this objection.

16 Paul Bloomfield (2000) further explores why a knack is not sufficient for virtue or skill.
17 Annas (1995) and (2003). For a development of the point that Plato required teachability
of both skills and virtues see Carter and Poston (2017).
212 Wright
particular techniques. While skills may employ techniques, they should not be
understood as being limited by them. A dedicated pianist will always be mo-
tivated to learn a new playing technique; if she really has the drive to aspire,
she won’t ignore a new technique that may improve her playing. And this is
parallel to the motivations that drive a virtuous person. She will also be moti-
vated to learn new techniques, in so far as they can help her to better respond
to the needs of others and the situations around her. While the techniques and
methods relevant to virtue may seem less technical, they are still relevant. For
example, the generous person who finds herself in a new culture will be mo-
tivated to learn what the local practices of gift-giving are. Failing to learn this
technique will limit her practice of generosity in her new culture.
3.6 Annas’s Skill Model Applied to Intellectual Virtues

Annas holds that these two aspects of expert skills apply to moral virtues, but
they can also be applied to responsibilist intellectual virtues. Responsibilists
think of virtues as developed traits of character, fitting with Annas’ basic re-
quirements. As we develop our intellectual virtues, we may begin by simply
modeling our behavior after those we think more intellectually virtuous than
ourselves. We might mimic those who are open-minded, intellectually coura-
geous, or intellectually careful. But to truly develop these virtues in ourselves,
we need to understand the reasons behind the virtuous person’s actions. One
could behave in ways that look open-minded, but only to be winning to oth-
ers or to avoid picking fights with them. The intellectually virtuous person will
be intrinsically motivated by a love of truth and this will motivate her open-
mindedness; her reason for considering the views of others will have to do with
the ways that it will help her to come to discover that truth and help her avoid
her own biases leading to falsehood. So intellectual virtues can also fit with
Annas’ idea that the virtues are expert skills that involve sensitivity to and the
ability to give reasons.
Responsibilist intellectual virtues can also easily be modeled to meet the
requirement of a drive to aspire. If the intellectually virtuous person values
truth for itself, she will always be looking for ways to improve her virtue and
so to better reach that truth. This can be nicely demonstrated by the virtue of
intellectual humility. An intellectually humble person will recognize their own
limitations; being driven to find the truth, the intellectually virtuous person
will look for ways around those limitations, perhaps even depending on others
in a naturally humble way.
These applications show that it is possible to develop a responsibilist vir-
tue epistemology that uses Annas’ skill model of the virtues. What advantages
does such a model display? To answer this question we must first turn to look
at the specific kinds of skepticism that pose a problem for the responsibilist.
THE SKILL MODEL 213
As we will see, some of the skeptical worries for the virtue reliabilist do not
arise on a responsibilist account of virtue epistemology; however there are also
special skeptical challenges for the responsibilist.
3.7 Responsibilist Virtue Epistemology Does Not Generate

Knowledge Skepticism
Above we have seen how the skill analogy can help the virtue reliabilist answer
skeptical concerns. Skepticism about the virtues is tied to skepticism about
knowledge in virtue reliabilism because reliabilists require possession and ex-
ercise of virtues for knowledge. This particular version of skepticism does not
affect virtue responsibilists. In general, responsibilists tend to avoid the project
of developing a definition of knowledge. Lorraine Code (1987) gives a princi-
pled reason for avoiding the definitional project; she worries that any analysis
of “S knows that P” broad enough to be relevant to all instances of knowledge
will have to ignore important differences between knowers, their epistemic lo-
cations, and their epistemic communities. Roberts and Wood (2007) also reject
the goal of giving a definition of knowledge; instead they see a role for virtues
in a regulative epistemology, an approach which focuses on guiding our epis-
temic practices. On these responsibilist accounts of virtue, skepticism about
virtue possession does not entail skepticism about knowledge.
Zagzebski stands out among virtue responsibilists by taking the definitional
project as important and by offering a definition of knowledge using respon-
sibilist virtue terms. As part of that definition she first provides an account of
acts of intellectual virtue:
An act of intellectual virtue A is one that arises from the motivational

component of an intellectual virtue A, is an act that persons with vir-
tue A would characteristically do in those circumstances, and is suc-
cessful in reaching the truth because of these other features of the act.
(1996: 108)
Zagzebski (1996: 207) then defines knowledge as “a state of true belief aris-
ing out of acts of intellectual virtue.” While this definition makes essential
reference to intellectual virtues, it does not require the possession of intel-
lectual virtues in order for someone to have knowledge. Instead knowing only
requires that the believer is motivated in the way that the intellectually virtu-
ous person would be and that they act as the virtuous person would. These
requirements rule out knowledge in some cases. The person who is motivated
by a desire to stay self-deceived won’t be motivated as the intellectually virtu-
ous person would. A person who refuses to listen to objections won’t be acting
as the intellectually virtuous person would. Both errors stop the believer from
214 Wright
coming to know. But it is perfectly possible to have knowledge on Zagzeb-

ski’s account even if one does not have the intellectual virtues; one must only
be motivated and act in the same way that an intellectually virtuous person
would.
Since Zagzebski’s own definition of knowledge does not require the pos-
session or exercise of the intellectual virtues for knowledge, it is compatible
with the possibility that very few people have the intellectual virtues. A lack
of intellectual virtue does not entail a lack of knowledge, and so widespread
virtue possession is not required for widespread knowledge. Since Zagzebski
is the lone virtue responsibilist to use the intellectual virtues in a definition of
knowledge, virtue responsibilism can avoid the virtue reliabilist concern that a
lack of intellectual virtues will lead to knowledge skepticism.
4 How the Skill Model Can Help the Responsibilist Address

Virtue Skepticism
While skepticism about the possession and exercise of the virtues need not
lead the responsibilist to skepticism about knowledge, it may still prove to be a
problem in its own right. As noted above, virtue responsibilists often focus on
the ways that intellectual virtues can guide our reasoning, playing a regulative
role in guiding inquiry. If we do not possess virtues, then our exercise of them
cannot play a guiding role. If virtues are to play a role in guiding our behavior,
they need to be the kind of character trait we can actually develop.
One alternative here is to think of virtues as an aspirational goal, rather than
as a character trait that we can fully develop. An aspirational model of the vir-
tues is compatible with there being very few, or even no, people with the fully
developed virtues. This is one natural interpretation of the Stoics’ account of
virtue. They hold up the image of the sage as the model of a person who has
developed all the virtues. But they decline to identify themselves as sages, or
even to identify exemplars of sagacity. We can infer that they believe the range
of people to have actually developed the virtues to be quite small. Still, they
think that we should aspire to be virtuous, aiming at a very high ideal.
While the virtue-as-ideal model may work for the Stoic theory, it is hard to
see how this model may be applied to learning an intellectual virtue. We might
put ourselves into the shoes of the non-virtuous person, longing to become vir-
tuous and looking for a way to do so. We might begin with the particular virtues
that have clear labels which are used in ordinary discourse. As we explore and
come to a better understanding of intellectual virtues like intellectual courage
and open-mindedness, we may find ourselves puzzled about how to handle
THE SKILL MODEL 215
d ifficult cases of each virtue and particularly instances where the virtues seem
to conflict. At this point, progress is best made by finding an epistemic exem-
plar who can provide us with guidance. Such a person can tell us how to bal-
ance the demands of our epistemic virtues; phronesis, or the practical wisdom
that brings together our conceptions of the virtues into harmony, seems not to
be a virtue that we have a clear account of in ordinary language. This establish-
es a need for an exemplar to follow in our learning of the virtues. The exemplar
may even serve as the meta-ethical grounding of the virtues. Zagzebski (2010)
has argued for this position, over the competing view that grounds virtues in
their role in a good life. If our development of the intellectual (and moral)
virtues follows this plausible model, it will be impeded by the absence of epis-
temic (or moral) exemplars. If virtues are only an aspirational ideal, there may
be no exemplars for us to identify and follow. So skepticism about intellectual
virtue might undermine our ability to understand virtue and to make progress
towards it, even when virtue is viewed as an ideal.
Why might one be skeptical about the existence of responsibilist intellec-
tual virtues? We have already seen one reason for questioning whether any
dispositions are reliabilist intellectual virtues, and we have seen how the virtue
as skills approach helps to address this skepticism. If our responsibilist intel-
lectual virtues have a reliabilist component, as Zagzebski’s do, then the relia-
bilist arguments will be relevant here. However many responsibilists do not
require that the virtues are reliable.18 These responsibilists do not have to face
skeptical arguments based on a lack of reliability. But both types of responsi-
bilist face a new skeptical worry; if their virtues are developed traits of char-
acter, they will require correct motivation. Motivation is an important part of
traditional moral virtues, and this element is included in modern intellectual
virtues. The role of motivation can be seen in Zagzebski’s definition of an act
of virtue, which requires that the act “arises from the motivational component
of an intellectual virtue.” Even if the individual performing an act of virtue
does not need to have the virtue in question, the virtue picks out a distinc-
tive appropriate motivational profile, which must be the motive of the per-
son performing the act. Because Zagzebski subsumes the intellectual virtues
under the moral ones, her intellectual virtues require a motivational compo-
nent. While the moral virtues may be motivated in a range of ways, Zagzebski
holds that the intellectual virtues are unified by sharing a single underlying
motivation – to obtain truth and avoid falsehood. The particular intellectual
18 Montmarquet (1993) and Baehr (2006 and 2011) both argue that the intellectual virtue
need not be reliable. Other authors like Code (1987) simply don’t build in a reliabilist
component.
216 Wright
virtues are d istinguished from each other by their each having a characteristic
motivation to achieve this end in a distinctive way.
But how can we come to have a desire for the truth for its own sake? One
argument for skepticism might be found in this requirement. While we often
value truth instrumentally, it might seem to be a mystery how we could come
to value truth intrinsically. If we must intrinsically value truth in order to de-
velop the intellectual virtues, few people might be capable of developing these
virtues. Note that this kind of skepticism does not arise in virtue reliabilism
since it requires the possession of dispositions that reliably get to the truth, but
it does not require any motivation instrumental or intrinsic behind this dispo-
sition. So a unique question arises for the responsibilist: how can we come to
desire truth for its own sake?
Annas’ skill model of the virtues has an answer. She addresses the question
of how the moral virtues can be motivating by regarding the question as con-
fused. She argues that we should not first generate a theory of when we should
judge an action to be virtuous and then ask how that judgment motivates us;
this would be to split a conjoined psychological state. Our judgements of an act
as courageous or as generous already contain a motivational component; that
is what makes virtue terms thick terms. Annas claims that by the time we are
ready to act as adults, our natural motivations have already been developed and
educated to connect with the virtues we have been trained to accept. As adults,
we may change our views about the nature of particular virtues; we may come
to think of pacifism as more courageous than going to war. But the motivational
evaluation behind our concept of courage does not need to be reconstituted.
In the case of the moral virtues it seems clear that the virtue terms we use are
already terms of praise, commending the action at the same time we identify it.
Is the same true for intellectual virtues? Zagzebski (2012: 33) has recently ar-
gued that a natural desire for truth is “part of every prereflective self.” If we do
have a natural non-instrumental desire for the truth, then this desire is a can-
didate to be developed through cultivation on Annas’ skill model. That cultiva-
tion will involve the identification and valuing of the intellectual virtues. Just
like the moral virtue terms, intellectual virtue terms can also be thick. We don’t
have to ask whether intellectual courage or open-mindedness (understood as
virtues) are valuable or how they can motivate those who understand them to
be virtues. Annas’ developmental picture shows how the motivation behind a
virtue (moral or intellectual) can be grounded in perfectly natural motivations;
this undermines the concern that an end as abstract as the truth or the good
can provide motivation. Instead it explains how it is possible for individuals
to develop the motivational component of the moral and intellectual virtues,
avoiding skepticism about both.
THE SKILL MODEL 217
5 The Situationist Skeptical Challenge
Another source of skepticism about the intellectual and moral virtues might
be found in the literature bridging psychology and philosophy which has come
to be known as situationism.
The situationist skeptical challenge aims to show that there are no robust
traits of character of the sort needed for either virtue ethics or responsibilist
virtue epistemology.19 Situationism takes its inspiration from psychological ex-
periments which show that our behavior is sometimes better predicted by the
details of the situations we find ourselves in rather than by any of our character
traits. When these empirical results are combined with the idea that virtues
ought to be able to predict behavior, philosophical situationists come to the
conclusion that there are no virtues.20
5.1 Situationism as a Source of Skepticism about the Moral Virtues

One of the many experiments cited to support situationism about the moral
virtues was carried out on Princeton Theological Seminary students.21 The
subjects were put in a situation where they encountered a person in apparent
need, a conspirator slumped in a doorway, coughing and groaning. The experi-
mental set-up was intended to check for correlation between helping behavior
and other variables that we might expect to predict helping. The first variable
was a measure intended to capture the subject’s general pattern of motivation,
a stand-in for character traits. Subjects were asked to characterize the basis of
their religious belief as either a means (e.g. to salvation) or as an end (e.g. as
a quest for meaning in life). The second variable was situational, intended to
make salient the religious basis for helping. Some of the subjects were asked
to read the parable of the Good Samaritan while the others read a text about
vocational alternatives to ministry. The third variable was also situational, hav-
ing to do with the amount of time students thought they had. All students
were asked to prepare a talk and sent to the next building to give it; some were
told they were late, others that they were right on time, and others that they
had time to spare. Of these three variables, only the third showed a signifi-
cant correlation with helping behavior. 63% of the subjects who thought they
had time to spare offered assistance, compared to 45% of those who thought
they were just on time, and 10% of those who thought that they were late. The
19 Alfano has developed a separate situationist critique of virtue reliabilism in his (2013).
20 Harman (1999) and (2000) holds this strong position, while Doris (2002, 6) allows that
“situationism does not preclude the existence of a few saints.”
21 Darley and Batson (1973).
218 Wright
e xperimenters’ conclusion was that only situation, not character predicted a

pattern of helping behavior.
This experiment (along with many others like it) is taken to support the
psychological situationist conclusion; behavior is better predicted by the situ-
ations that people find themselves in than by their character traits. Bringing
this conclusion into contact with the philosophical literature on character
traits and virtues, Gilbert Harman thinks we should conclude that we do not
have the sort of robust character traits that we would need to predict behavior.
But these robust character traits, when positive, are the basis for virtues; so
Harman (1999: 327–329; 2000: 224) concludes that we should move away from
virtue ethics. John Doris takes the empirical studies to support a less radical
conclusion; while they do not show that we lack character traits at all, they do
provide evidence against the kind of very general character traits we appeal to
when we talk about virtues like courage, generosity, and honesty. Instead Do-
ris suggests that the evidence only suggests that we have local character traits
like “compassion-when-not-rushed” (1998: 508). While we could use these local
traits to predict some behavior, Doris’ view still eliminates any character traits
broad-based enough to be called virtues.
5.2 Situationism as a Source of Skepticism about Intellectual Virtues

What does this argument have to do with intellectual virtues? Recently propo-
nents of situationism have argued that it applies to virtue epistemology as well
as to virtue ethics. One experiment that has been used by Mark Alfano in argu-
ing against responsibilism is Solomon Asch’s (1951, 1956) conformity experi-
ments. Subjects in this experiment are asked to perform a simple visual task
of picking which of three comparison lines is the same length as an original.
This is an easy task and subjects could ordinarily perform this task correctly
more than 99% of the time. However the “critical subjects” in these experi-
ments faced a situation where all the other seeming co-subjects in the room
were actually conspirators; these conspirators were unanimous in identifying
the same incorrect line as the matching one. Facing unanimous dissent from
the other apparent subjects, the critical subjects went along with the majority
(and against their own perception) in their reports in about one third of the
trails, and 75% of the subjects went along with the majority at least once. The
original studies had either 7 or 9 conspirators, but later studies showed that a
unanimous dissent from as few as three others can generate a similar persua-
sion in critical subjects.22 However a unanimous dissent of two produced only
a weak effect and the presence of only one disagreeing report had a negligible
22 Asch (1956).
THE SKILL MODEL 219
effect. Alfano (2013: 135) concludes from this experiment that most people
do not have the virtue of intellectual courage; if they did, we would expect
to see them disagreeing more often, particularly on such simple perceptual
judgments. P erhaps we could ascribe to people the local trait of intellectual-
courage-unless-faced-with-unanimous-dissent-of three-or-more, but Alfano
argues that this character trait is not robust enough to count as an intellectual
virtue.
6 Responding to Situationism
How should the virtue theorist respond to both of these experiments? Julia
Annas recommends that we consider carefully exactly what virtues are before
we bring them into conflict with experimental results. She worries that situ-
ationists are thinking of virtues too much like blind habits. When we are first
learning to be virtuous we might habitually follow the dictates of simple ethi-
cal rules like “never tell a lie.” These rules may be necessary at a certain stage
of our development, but simple rule-following isn’t how virtuous adults make
their moral choices. Instead those who are virtuous are sensitive to particular
reasons. This follows from Anna’s requirement that virtues involve a need to
learn and that a virtuous person will be able to articulate her reasons for act-
ing in a particular way. In responding to situationism, Annas (2003a: 28) says a
“virtue is not a habit of reliably producing behaviour of a type independently
fixed; it is a disposition to act on reasons of a certain kind.” The honest person
may, in extreme circumstances, tell a lie. This is not in any conflict with their
honesty; rather it is a natural result of the honest person recognizing both rea-
sons to tell the truth and reasons to lie. Annas (2003a, 2011) argues that virtue,
like all practical expertise, “is highly situation-sensitive.” We should expect
virtues to explain the action of a virtuous person, but we should not look to
virtues to give us a simple prediction of their behavior in complex situations.
And the experiments cited in favor of situationism embed just such complex
situations – they are novel or atypical, and they generate conflicting reasons.
6.1 Response One: Novel or Atypical Situations

Alfano’s focus in his interpretation of the Asch experiments is on the virtue of
intellectual courage to speak up in the face of contrary opinions. He argues that
this is an important type of intellectual courage to develop because it helps
us to avoid “pluralistic ignorance.” This term describes an attitude that people
take when they act in accordance with a social norm that they do not per-
sonally believe in. While they recognize that their own actions do not reflect
220 Wright
their beliefs, they also tend to over-attribute belief in the social norm to oth-
ers, based on their actions in accordance with it norm. Alfano notes that the
phenomenon of pluralistic ignorance has been used to explain phenomena as
diverse as the longevity of Soviet communism and binge drinking by college
students. He suggests that the intellectual courage to speak out (and not just
the courage to believe differently) is the virtue needed to combat pluralistic
ignorance.
Note that ordinary examples of pluralistic ignorance highlight something
odd about the Asch experiment. We are prepared to encounter a difference of
opinion from others when we are discussing politics or drinking habits; these
are areas in which speaking honestly about our beliefs may help to change the
general belief. They are central cases for intellectual courage in speaking out.
But we do not expect to find disagreement concerning perception of nearby
objects in good viewing conditions. These are not central cases for intellectual
courage in speaking out.
It is a surprise to find ourselves disagreeing with others about simple per-
ceptual facts. This surprise can be illustrated by the public reaction to the im-
age of a dress posted on line in 2015 that appeared blue and black to some
people while appearing white and gold to others.23 The debate about the color
of this dress drew attention and became an internet phenomenon because the
experience was so unexpected. We do not expect disagreement on perceptual
matters that should be clear-cut. In the Asch experiments, subject’s surprise at
the situation of perceptual disagreement can be inferred from their dramatic
change in demeanor through the course of the experiment; some became
withdrawn and quiet, while another started laughing when giving his answers.
These reactions of surprise seem to indicate that the subjects were unsure of
how to act as they found themselves in an atypical situation. After debriefing,
one subject reported that, “If it was an issue more emotionally charged or more
important, involving a more emotional result, I’d be much less likely to accept
their judgments” (Asch 1956: 39). While this prediction might simply be wish-
ful thinking on the part of the subject, his reaction also reflects the oddity of
his situation and its difference from contexts in which we expect to have to
stand up to others.
It should be obvious that exercising a virtue in a novel or atypical situa-
tion can be difficult. As we have seen above, exercising virtue is a matter of
responding to the right kinds of reasons; in a novel situation the right kinds of
reasons may be difficult to discern. As we have also noted, the virtuous person
23 See Corum (2015) for the images and discussion of the optical phenomenon that pro-
duced these two impressions.
THE SKILL MODEL 221
may well need to learn techniques specific to a particular context in order to

exercise virtue.
Remember the case of someone who has the virtue of generosity but who
faces their first instance of gift-giving in a new culture. On Annas’ model the
virtues require a drive to aspire, so if this person is truly generous she will be
motivated to become a gracious gift-giver in her new circumstance. This moti-
vation gives her a reason to seek out information and learn techniques about
the relevant gift-giving norms in the new culture; her virtue generates in her a
need to learn those things relevant to the exercise of her virtue. We do not, and
should not, expect the virtuous person to instantaneously adapt her behavior
to an unanticipated circumstance. She must have relevant information about
her new circumstance to do so.
The requirement that one have relevant information to exercise virtue in
a novel situation is backed up by the skill analogy. An expert pianist may be
able to coax a passible performance out of a damaged and weathered piano;
but we don’t expect her to be able to do so instantaneously. Her preliminary
inspection and playing of the damaged piano will give her information about
how she might adapt her ordinary playing on a well-maintained instrument to
fit with the limitations of a damaged one. But she may well have to practice on
the damaged piano to develop new techniques for how to deal with its limita-
tions. The pianist has tools that help her deal with a novel instrument, but they
cannot help her adapt to it without a search for relevant information and adap-
tation of her ordinary techniques. Similarly, virtues on the skill model adapt us
to handle novel situations. The generous person is in a better position to adapt
to new giving situations than others, but she still needs information before she
can act in a novel situation.
The subjects in the Asch experiments certainly find themselves in an atypi-
cal and novel situation for exercising the virtue of intellectual courage. They
receive unexpected feedback when they hear multiple others giving an incor-
rect identification of the matching line. That feedback makes it difficult for
them to trust the situation they thought they were in. Is it a trick? Are the oth-
ers really reporting what they believe? These are questions that in intellectu-
ally virtuous person would ask himself in their situation. But the experimental
set-up does not allow for actions, like directly questioning the other apparent
subjects, that would provide the answers to these questions. It is no surprise
then that the subjects responded in different ways; some reporting their ex-
periences, others echoing the group. They may have come to very different
conclusions about the nature of their situation based on the limited informa-
tion provided. This gives us reason to doubt that the Asch experiments give us
evidence of a lack of the virtue of intellectual courage. They highlight novel
222 Wright
situations where we least expect to find clarity about what the intellectually
virtuous person should do.
6.2 Response Two: Conflicting Reasons

Annas’ account of virtues as skills focuses on the way that one reasons about
and then responds to a situation. So the important question to ask about these
experimental results is if the people in these experiments are considering the
right reasons. This is, of course, harder to observe than their behavior. But in
both the Asch and Good Samaritan experiments we have some evidence that
subjects did consider the relevant reasons; they were simply confounded by
the fact that the experimental set-up introduced conflicting reasons.
As we have noted, Alfano’s focuses on the virtue of intellectual courage to
speak up against the views of others. But this is not the only kind of intellectual
courage he considers; he also talks about the intellectual courage to believe
and to doubt, where this can apply to oneself or to others. Highlighting the
courage to believe others, he notes that our contemporary lives are full of trust
in others; we trust others for information, and we trust them for food safety.
Intellectually virtuous people will trust others, not blindly, but in the right way
and at the right time. Heidi Grasswick (2017) has also made this argument in
defending feminist virtue epistemology from situationism; rather than trust-
ing all the time, we should trust others to the degree that they have a track
record of working to support our interests or the interests of our group. Still
well-regulated trust is needed. And not just practically. Those who simply re-
fuse to listen to the testimony of others are not intellectually virtuous. Thus we
can see that the virtue of intellectual courage could be realized in speaking up
against the views of others, but also through trusting the testimony of others.
The possibility of conflict between the demands of virtues also explains why
we should not expect virtues to perfectly predict behaviors. The honest person
recognizes reasons to tell the truth, but can also recognize conflicting reasons.
In the case of the murderer at the door, she recognizes the well-being of the
person pursued. But there can also be less clear-cut conflicts. What should the
honest person say about the friend’s haircut that she dislikes but knows is cur-
rently popular? The virtuous person must use her own developed judgment;
hence the role of phronesis in an overall life of virtue. Practical wisdom helps
us not only to exercise the individual virtues skillfully, it also helps us to bal-
ance apparently competing demands of different virtues.
Just such an instance of conflicting demands can be seen in the Asch ex-
periment. The subjects, upon hearing a number of other people report that
line 1 was the matching length, had a new reason to doubt their former belief
that, say, line 2 was the match. Even someone with the virtue of intellectual
THE SKILL MODEL 223
c ourage in speaking out should recognize this as a reason; listening to the re-
ports of others is part of having the intellectual courage to trust. Some subjects
in this experiment demonstrated reasoning that focused on trusting others
over speaking out. After being debriefed at the end of the experiment, some
who had been swayed by the group indicated that they had come to doubt their
own perceptions on the basis of accepting the reports of others as counter-
evidence. “There was such an overwhelming weight of opposition against me
that I wondered whether my eyesight or judgment were at fault—or an o ptical
illusion which either they or I did not perceive” (Asch 1956: 41). Even those
subjects who did not conform to the group reported that they had taken into
consideration the reports of the others. In fact, their reports sometimes sound-
ed as if they thought they ought to have trusted others more; one said, “In the
light of the opinions of the others I’d say I was wrong, but I answered as I saw”
(Asch 1956: 40). While an intellectually courageous participant should take his
own perception seriously, he should also take the perceptual reports of many
others seriously as well. Thus subjects in this experiment faced a conflict be-
tween legitimate epistemic concerns, and their being swayed to the majority
need not indicate that they did not take into consideration the reasons that are
characteristic of intellectual courage.
Similarly in the Good Samaritan experiments, subjects faced a conflict be-
tween the demands associate with two different virtues. They were confronted
by someone who appeared to need help, but they also were in the very act of
meeting an obligation to give a talk. Darley and Batson observed that those
subjects in the original experiment who did not help appeared anxious after
their encounter. The experimental design did not allow for them to ask the
subjects why they acted, but Darley and Batson (1973: 108) hypothesized that
the subjects who were told that they must hurry faced a conflict of reasons:
Why were the seminarians hurrying? Because the experimenter, whom

the subject was helping, was depending on him to get to a particular
place quickly. In other words, he was in conflict between stopping to help
the victim and continuing on his way to help the experimenter…Conflict,
rather than callousness, can explain their failure to stop.
This hypothesis was empirically tested in a follow-up experiment by Batson

et al. (1978), which aimed to disentangle the effect of wanting to help the ex-
perimenter from the effect of being in a hurry. In addition to the situational
variable of hurry, the subjects in this experiment were carrying data that they
were told either were/were not “vital for successful completion of a research
project.” The subjects who were told that their data were not vital stopped
224 Wright
to help at rates higher than in the original experiment; those told their data
were vital and that they must hurry stopped to help only 10% of the time.
Their results suggest that the subjects were taking into account two conflict-
ing reasons. When they thought that their obligation to provide information
was more important, they followed this obligation, even over an obligation
to assist a stranger. Thus, we should not take a failure to stop (when hurrying
to help in an experiment) as direct evidence that subjects were ignoring or
failing to recognize the need of others as a reason; they just faced a conflict of
reasons.
We are now in a position to re-evaluate the import of the experiments
that situationists cite to motivate the rejection of moral and intellectual vir-
tues. When we are careful in thinking about the nature of virtues, we will
see that we should not expect them to provide predictions of behavior in
situations of conflict, even if we expect those same virtues to undergird pat-
terns of behavior in more clear-cut cases. This does not mean accepting that
our virtues flee in cases of conflict, but rather recognizing that there can
be conflicting reasons provided by different virtues. Navigating that conflict
is an ability that we expect virtuous people to do. But even this navigation
may not be smooth in novel or unexpected situations. When we recognize
this, we should see that even the surprising results of the empirical work
situationists cite do not provide us with a solid argument for rejecting the
existence of the virtues.
7 Conclusion
The response to the situationism just outlined demonstrates how useful the
skill model of virtue can be for the virtue responsibilist. We have seen how the
skill model already helps the virtue reliabilist avoid skepticism about knowl-
edge and skepticism about intellectual virtues. While virtue responsibilism
doesn’t generate skepticism about knowledge, it does face skepticism about
the intellectual virtues themselves. But these virtues become far less mysteri-
ous when we think of them as being developed and exercised just like expert
practical skills. These skills and virtues both depend on our becoming sensi-
tive to the right kinds of reasons, and both will generate motivations in us to
advance and develop our skills and virtues. Virtues on this model are nuanced,
just as skills are. When we deploy our virtues in complex and novel situations
they display a situation-sensitivity that is desirable, not the grounds for a cri-
tique. It is my hope that these arguments will motivate a further exploration of
THE SKILL MODEL 225
the skill model of virtues in responsibilism, and that responsibilism will ben-
efit from those developments.24
Bibliography
Alfano, M. 2013. Character as Moral Fiction. Cambridge, UK: Cambridge University

Press.
Annas, J. 1995. “Virtue as a Skill,” International Journal of Philosophical Studies 3: 227–43.
Annas, J. 2003a. “Virtue Ethics and Social Psychology,” A Priori 2: 20–34.
Annas, J. 2003b. “The Structure of Virtue.” In M. DePaul (ed.), Intellectual Virtue: Per-
spectives from Ethics and Epistemology, 15–33. Oxford, UK: Clarendon Press.
Annas, J. 2011. Intelligent Virtue. Oxford: Oxford University Press.
Asch, S. 1951. “Effects of Group Pressures upon the Modification and Distortion of
Judgment.” In H. Guetzkow (ed.), Groups, Leadership, and Men, 222–235, Pittsburgh:
Carnegie Press.
Asch, S. 1956. “Studies of Independence and Conformity: A Minority of One against a
Unanimous Majority,” Psychological Monographs 70: 1–70.
Baehr, J. 2006. “Character, Reliability, and Virtue Epistemology,” Philosophical Quarterly
56: 193–212.
Baehr, J. 2011. The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology. Ox-
ford: Oxford.
Batson, C.D., P.J. Cochran, M.F. Biederman, J.L. Blosser, M.J. Ryan, & B. Vogt. 1978. “Fail-
ure to Help When in a Hurry: Callousness or Conflict?” Personality and Social Psy-
chology Bulletin 4: 97–101.
Bloomfield, P. 2000. “Virtue Epistemology and the Epistemology of Virtue,” Philosophy
and Phenomenological Research 60: 23–43.
Carter, J.A. & T. Poston. 2018. A Critical Introduction to Knowledge-How. London:
Bloomsbury.
Code, L. 1987. Epistemic Responsibility. Hanover: University Press of New England and
Brown University Press.
Corum, J. 2015. “Is That Dress White and Gold or Blue and Black?” New York Times re-
trieved from https://www.nytimes.com/interactive/2015/02/28/science/white-or-
blue-dress.html.
24 My thanks to audiences at the Northwestern Pre-APA Epistemology Conference, the Or-

ange Beach Epistemology Workshop, and the 2017 Bled Philosophy Conference for com-
ments on this paper. Special thanks for the thoughtful and detailed feedback on earlier
drafts given me by Kevin McCain and Ted Poston.
226 Wright
Darley, J. & C.D. Batson. 1973. ‘“From Jerusalem to Jerico”: A Study of Situational and
Dispositional Variables in Helping Behavior,’ Journal of Personality and Social Psy-
chology 27: 100–108.
Doris, J. 1998. “Persons, Situations, and Virtue Ethics,” Noûs 32: 504–530.
Doris, J. 2002. Lack of Character: Personality and Moral Behavior . Cambridge UK: Cam-
bridge University Press.
Grasswick, H. 2017. “Feminist Responsibilism, Situationism, and the Complexities of
the Virtue of Trustworthiness.” In A. Fairweather & M. Alfano (eds.), Epistemic Situ-
ationism, 216–234. Oxford: Oxford University Press.
Greco, J. 2000. Putting Skeptics in Their Place: The Nature of Skeptical Arguments and
Their Role in Philosophical Inquiry. Cambridge: Cambridge University Press.
Greco, J. 2010. Achieving Knowledge. Cambridge: Cambridge University Press.
Harman, G. 1999. “Moral Psychology Meets Social Psychology: Virtue Ethics and the
Fundamental Attribution Error,” Proceedings of the Aristotelian Society 99: 315–331.
Harman, G. 2000. “The Non-Existence of Character Traits,” Proceedings of the Aristote-
lian Society 100: 223–226.
Long, A.A. & D.N. Sedley 1987. The Hellenistic Philosophers: Vol. 1. Cambridge: Cam-
bridge University Press.
Montmarquet, J. 1993. Epistemic Virtue and Doxastic Responsibility. Lanham: Rowman
and Littlefield.
Roberts, R.C. & W.J. Wood. 2007. Intellectual Virtues: An Essay in Regulative Epistemol-
ogy. Oxford: Oxford University Press.
Sosa, E. 1980. “The Raft and the Pyramid: Coherence versus Foundations in the Theory
of Knowledge,” Midwest Studies in Philosophy 5: 3–26. Reprinted in Knowledge in
Perspective.
Sosa, E. 1991. Knowledge in Perspective. New York: Cambridge University Press.
Sosa, E. 2007. A Virtue Epistemology: Apt Belief and Reflective Knowledge, Volume 1. Ox-
ford: Oxford University Press.
Sosa, E. 2011. Knowing Full Well. Princeton: Princeton University Press.
Zagzebski, L. 1996. Virtues of the Mind. Cambridge, UK: Cambridge University Press.
Zagzebski, L. 2010. “Exemplarist Virtue Epistemology,” Metaphilosophy 41: 41–57.
Zagzebski, L. 2012. Epistemic Authority: A Theory of Trust, Authority, and Autonomy in
Belief. Oxford: Oxford University Press.
Chapter 13
Two Strategies for Explaining Away Skepticism

1 Introduction1
Why not skepticism? After all, some philosophers suggest that we simply can-
not adequately respond to skepticism, or at the very least that the prospects
for doing so are particularly dismal.2 In light of this, perhaps we should simply
accept skepticism as a way of life.3 So, again, why not skepticism? The short
answer is that skepticism fails as an explanatory hypothesis.
The core skeptical thought is that our cognitive position with respect to
many ordinary beliefs is too weak to secure justification and knowledge. The
skeptic reasons that our cognitive position is constituted by appearances and
that we lack sufficient evidence, reason, or justification for taking those ap-
pearances to be accurate representations of the world. One common response
to skepticism is G.E. Moore’s response. According to the Moorean response,
start with an ordinary belief, claim it is knowledge, and then deduce the falsity
of any hypothesis incompatible with the truth of the mundane belief. Explana-
tionist responses to skepticism differ from the now popular Moorean respons-
es.4 What is often unnoticed about these explanationist responses is that they
come in at least two varieties. Our goal in this paper is to clearly lay out these
responses and how they work as well as their strengths and the challenges they
face. Although our primary goal is to consider these responses in their own
right, as is often helpful for truly appreciating a particular theory, we will con-
sider how each compares with the Moorean response.
1 Thanks to audiences at Auburn University and the 2016 Alabama Philosophical Society meet-
ing for helpful comments on earlier drafts.
2 See Lehrer (1971), Stroud (1984), and Fumerton (1995).
3 Reed (Chapter 5 of this volume)
4 There are hints of the explanationist response to skepticism as early as Locke (1690/1975).
More recent versions of this response have been proposed by Russell (1912), Jackson (1977),
Cornman (1980), BonJour (1985), (1999), Goldman (1988), Lycan (1988), Moser (1989), Vogel
(1990), (2005), and McCain (2014a), (2016). See Moore (1939), Pryor (2000), (2004), Pritchard
(2002), and Willenken (2011) for discussion and defense of the Moorean response to
skepticism.

228 McCain and Poston
2 Explanationism – A Brief Primer
Before digging into the specifics of the two kinds of explanationist responses to
skepticism, the relation between these responses and explanationism needs to
be made clear. Some may mistakenly assume that accepting an explanationist
response to skepticism commits one to accepting strong forms of explanation-
ism. Of course, explanationist responses to skepticism fit particularly well within
a broader explanationist epistemology, however, an explanationist response to
skepticism does not imply accepting more than a weak form of explanationism.
William Lycan (2002: 417) offers a helpful taxonomy of explanationist views:
Weak Explanationism:
explanatory inferences (inferences to the best explanation) from a
given set of premises can epistemically justify a conclusion.
Sturdy Explanationism:
Weak Explanationism + explanatory inferences can justify conclu-
sions without being derived from some other more basic form of
ampliative inference.
Ferocious Explanationism:
Sturdy Explanationism + explanatory inference is the only basic
form of ampliative inference.
Holocaust Explanationism:
Ferocious Explanationism + all inferences and reasoning, including
deductive, is derived from explanatory inference.
There is considerable controversy concerning the truth of the various kinds

of explanationism. For present purposes it will not be necessary to go into the
details of this debate.5 Instead, all that is needed at this point is recognition
that explanationist responses to skepticism, while consistent with all four
kinds of explanationism, are only committed to the truth of Weak Explana-
tionism. All that is required for either of the primary kinds of explanationist
response that we will consider is that explanatory inferences can provide justi-
fication for their conclusions. So, even if one finds the other, stronger forms of
explanationism implausible, she should not simply dismiss the explanationist
response on those grounds.6
5 For a sampling of the debate the reader is encouraged to look at Harman (1965), (1986), Leh-
rer (1974), van Fraassen (1980), (1989), Fumerton (1980), (1992), (1995), Lycan (1988), (2002),
Roche & Sober (2013), McCain (2014a), McCain & Poston (2014), and Poston (2014).
6 It is worth pointing out that there is some reason to think that one cannot at the same time
take skepticism seriously and deny explanationism (at least of the weak variety). According
to James Beebe (2017), skepticism presupposes explanationism.
Two Strategies for Explaining Away Skepticism 229
3 Getting Clear on the Skeptic’s Argument
With the relationship between explanationist responses to skepticism and ex-

planationism in hand it is time to turn to examining skepticism. To sharpen the
skeptical puzzle let us put on the table a particular skeptical argument, one that
starts with a popular skeptical hypothesis: “the Boltzmann Brain” hypothesis:
Given what we know about the nature of the universe, it is a priori quite
unlikely that we would have a state of low entropy like this, a state in
which there are galaxies, stars, planets, persons, animals, and so on. A
much more likely scenario is that a chance fluctuation of matter results
in an isolated state of low entropy in which there is an isolated brain
floating in space. This brain realizes conscious states of experiences like
ours with apparent memories of a long past. The conscious subject of
this Boltzmann brain—Brainy—is in a state phenomenologically indis-
tinguishable from a normal embodied state of a subject like us. Right now
it appears to Brainy that she is reading a philosophy book and thinking
about the implications of explanationism and skepticism.
The Boltzmann Brain can be used to generate a skeptical argument (BRAIN):
(1) Necessarily, if S knows that (e.g.) S has hands, then S knows that S is not a
handless Boltzmann brain.
(2) S does not know that S is not a handless Boltzmann brain.
So,
(3) S does not know that (e.g.,) S has hands.7
Of course, there are numerous other hypotheses that can be used to construct
analogous skeptical arguments. However, the points we make about BRAIN
are equally applicable to these other skeptical arguments.
4 Two Explanationist Responses to Skepticism
Although the two explanationist responses to skepticism that we will discuss

are importantly different, they are also similar in various ways. It is worth
7 There are important questions about how to best understand arguments for skepticism. For
example, there is an ongoing debate (see Cohen (1998), Brueckner (1994), Pritchard (2005),
and McCain (2013)) concerning the relationship between skeptical arguments that employ
closure principles (such as BRAIN) and those that construe skepticism as an underdetermi-
nation problem. We can set aside this question and similar questions for our present pur-
poses because the answers to them will not affect our arguments.
oting two of these similarities before describing the responses and their dif-
n
ferences in detail. First, an obvious similarity—both are committed to the truth
of explanationism (at least of the weak variety). Second, both explanationist
responses deny premise (2) of BRAIN on explanatory grounds. Very roughly,
both responses involve arguing that we do know that we are not handless
Boltzmann brains because a rival hypothesis best explains our experiences—
our ordinary real world hypothesis (rwh).8
The most helpful way to get a handle on the two explanationist responses
to skepticism is to situate them in broader approaches to epistemology. On
the one hand, we have what we might term a “first philosophy” approach to
epistemology, which begins with skepticism and uses the response to it as the
basis for a full theory of knowledge. On the other hand, we have a “second
philosophy” approach, which does not consider far-reaching skepticism to be
a primary concern. Let us take a look at both of these explanationist responses.
4.1 Explanationism & First Philosophy

First Philosophy Explanationism (Ex-1st) is inspired by the epistemological
project of Descartes. This variety of explanationism places a strong emphasis
on responding to the threat of skepticism. Accordingly, Ex-1st involves deal-
ing with the challenge of skepticism, and then, using the insights gained by
meeting this challenge to develop a full epistemology. So, skepticism, and
responding to it, is a central (if not the central) issue around which Ex-1st de-
velops as a theory. Importantly, Ex-1st does not guarantee an anti-skeptical end
result. It is compatible with this way of responding to skepticism that at the
end of the day skepticism is correct. After all, it could turn out on this picture
that skepticism of one variety or another is the best explanatory hypothesis on
offer. It could turn out this way, but supporters of this explanationist response
argue that things do not in fact turn out this way.
Ex-1st in large part revolves around meeting the skeptic’s challenge. But, it is
more than this. The Ex-1st explanationist not only attempts to meet the skepti-
cal challenge by responding to the skeptic’s argument (as much as possible)
and developing a full epistemological theory from the results, she attempts to
do this by meeting the skeptic’s challenge head on. Ex-1st begins from a neu-
tral starting point—this is why those making this sort of response to skepti-
cism acknowledge that skepticism is not guaranteed to be defeated; there is a
lot of work required to overcome the challenge of skepticism. Explanationists
who offer this sort of response to skepticism are careful to only make use of
evidence and explanatory virtues that do not privilege our rwh over skeptical
8 We borrow the term “real world hypothesis” from Vogel (1990).

rivals from the outset.9 Ex-1st insists on responding to the skeptic in a way that
does not beg the question against the skeptic by taking for granted anything
that cannot also be true given various skeptical hypotheses. For instance, an
Ex-1st explanationist will not assume that we have sense organs when arguing
against the skeptic because such an assumption conflicts with the Boltzmann
Brain hypothesis. Instead, this explanationist response will begin with data
that is present whether or not we are Boltzmann brains—namely, our own
non-factive mental states. Paul Moser (1990: 131–132) offers a nice brief encap-
sulation of the Ex-1st response to skepticism:
If (a) I seem to see a white piece of paper, (b) the contents of this per-
ceptual experience are best explained for me by the physical object
proposition that there is a white piece of paper here, and (c) nothing in
my overall perceptual and sensory experience indicates that there is not
a white piece of paper here or that the explanatory relation of (b) does
not hold, then the proposition that there is a white piece of paper here is
empirically epistemically justifiable for me.
Like all explanationist responses to skepticism Ex-1st involves a commit-

ment to the truth of weak explanationism. Ex-1st also seeks to deny premise
(2) of BRAIN on explanatory grounds. Typically, supporters of this sort of re-
sponse to skepticism will argue that the rwh is superior to the Boltzmann
Brain hypothesis on the grounds that the former has more explanatory power,
is simpler, and so on than the latter. One thing that those opting for an Ex-
1st explanationist response to skepticism might point out is that whereas the
rwh can make accurate predictions about how our sensory experiences will
be, the Boltzmann Brain hypothesis can only accommodate the data about our
sensory experiences after the fact.10 Vogel, for example, argues that aspects of
our sensory visual experience of apparent three-dimensional objects follow
from the geometry of such objects together with a common-sense hypothesis
about the way vision works. The Boltzmann Brain hypothesis has to add these
features of sensory experience into the content of the hypothesis. Another
thing would be to argue that necessary truths can play an explanatory role in
rwh explanations that can only be mimicked by contingent regularities in
Boltzmann Brain explanations. As a result of this difference, one can argue
9 For contemporary examples of this sort of explanationist response to skepticism see Bon-
Jour (1985), (1999), BonJour and Sosa (2003), Moser (1989), (1990), Vogel (1990), (2005), and
McCain (2014a), (2016).
10 See McCain (2012) for more on this point.
that the rwh is simpler than the Boltzmann Brain hypothesis in a very impor-
tant sense.11
There are other explanatory virtues that can be appealed to in making an
Ex-1st response to skepticism besides those mentioned above. Rather than list
the myriad ways in which one can make an Ex-1st explanationist response to
skepticism we will simply note one way that is not an option. Ex-1st does not
countenance appeals to the explanatory virtue of conservatism (the idea that
fit with background information or leading to fewer revisions to one’s overall
set of beliefs is a good-making feature of an explanation). Appealing to conser-
vatism can make it a fairly straightforward matter to respond to the skeptic’s
argument. Nonetheless, conservatism gives us a reason to prefer the rwh to
the Boltzmann Brain hypothesis only if we already have beliefs that presup-
pose the existence of mind-independent external world objects of the sort that
we typically take ourselves to interact with on a regular basis. In other words,
the virtue of conservatism will not favor either hypothesis given a completely
neutral standpoint in the skeptical debate. It is from just such a standpoint
that Ex-1st proposes to engage the skeptic, so conservatism is off the table for
this response to skepticism.
4.2 Explanationism & Second Philosophy

Second Philosophy Explanationism (Ex-2nd) begins with our natural concep-
tion of the world. We are human creatures who live in a physical world. We do
have knowledge and much of our knowledge comes by way of the senses. But
this natural conception of ourselves is riddled with philosophical puzzles. If
we genuinely know that we have hands on the basis of sense perception, then
why does it seem wrong to infer from that knowledge alone that we are not in
a skeptical scenario?
Aristotle’s epistemology provides direction.12 The goal (telos) of human
belief is scientia (roughly, a complete and coherent view of knowledge and
the causal basis for this knowledge). We begin in a state that may be called
pre-understanding in which we have viewpoints that need to be brought to
the maturity of science. If scientia is achievable, then the puzzles that our
11 For more on this see Vogel (1990) and (2005). Of course, the Ex-2nd explanationist can
also make these sorts of arguments against the skeptic.
12 It is worth noting that although there are definitely epistemic naturalist themes in Aris-
totle’s writings, and he does say a number of things that are congenial to explanationism,
it is not clear that Aristotle himself should be taken to be a representative of Ex-2nd.
However, since our present goal isn’t Aristotle scholarship, we will simply make use of the
components of his epistemology that help illustrate important points about Ex-2nd.
re-understanding gives rise to can be resolved into a comprehensive and co-

p
herent view of the world. (see Posterior Analytics 1.1)
Aristotle’s stress on the importance of coherence and the distinction be-
tween pre-understanding and understanding is a similar motif to what is
found in modern epistemic naturalism.13 Aristotle sees the necessity of having
starting points that are, in some respect, epistemic successes. This is due to his
reflection on the Meno paradox. Plato argues that no learning is possible unless
one already has knowledge. Aristotle refines the puzzle by arguing that there
are lower epistemic successes besides episteme. Contemporary epistemic natu-
ralists likewise see that a view from nowhere is impossible. We must begin with
a privileged, but defeasible, viewpoint. As W.V.O. Quine (1960: 3) puts the point:
The philosopher and the scientist are in the same boat. If we improve our
understanding of ordinary talk of physical things, it will not be by reduc-
ing that talk to a more familiar idiom; there is none. It will be by clarifying
the connections, causal or otherwise, between our ordinary talk of physi-
cal things and various further matters which in turn we grasp with the
help of ordinary talk of physical things.
Let us clear up a confusion about the relationship between epistemological

naturalism and metaphysical naturalism. This latter form of naturalism is
roughly the view that there are no non-physical aspects to reality. In episte-
mology, naturalism is aligned with the view that epistemology is an aspect of
empirical psychology. Epistemic naturalism is often aligned with the idea that
epistemology should use the methods of natural science and thus inference to
the best explanation has a key role to play. In our view, the idea that epistemol-
ogy should use inference to the best explanation is entirely separable from the
view that epistemology is a dimension of empirical psychology. Our emphasis
on Aristotle makes it clear that while the broadly Quinean view runs these two
elements together, they are separable. Poston (2014) provides a broadly Aristo-
telian conception of epistemology. The aim of epistemology is knowledge and
the methods are the explanatory virtues of simplicity, explanatory power, and
13 See Neurath (1983), Quine (1960), (1990), Quine and Ullian (1970), Goodman (1965), (1978),
Harman (1973), (1980), (1986), Kuhn (1977), Laudan (1984), Lycan (1988), Maddy (2007),
and Poston (2012a), (2013a), (2013b), (2014). See also Sosa’s (2007), (2009) virtue epistemol-
ogy. It is worth noting that in some cases these philosophers are concerned primarily with
more narrow issues in philosophy of science rather than responding to external world
skepticism or developing a full epistemological theory. Despite this fact, they each fall
squarely within the epistemological naturalist camp, and they each argue for claims that
lend support to Ex-2nd.
c onservativeness. Poston’s view is like Quine’s in the respect of claiming that

we can begin epistemology with our ordinary beliefs about the world and look
to the methods of science (and others if relevant) to increase the explanatory
power of our view. But it is unlike Quine’s naturalistic view in that it is not com-
mitted to the claim that epistemology is empirical psychology.
How does an Ex-2nd response work? Poston (2014) provides a contemporary
account. The key is to figure out how to make sense of a privileged starting
point that does not license an objectionable dogmatism.14 Poston interprets
this in terms of defeasible epistemic conservatism. Briefly, the idea is that a
person has epistemic justification for her existing beliefs provided they do not
conflict with her evidence. Epistemic conservatism provides some key evi-
dence for assessing other beliefs.
In addition to epistemically conservative belief, the explanatory virtues of
simplicity and power provide the rest of the Ex-2nd account. All other things
being equal, a theory that is simpler than a competing theory has more epis-
temic merit for believing. All other things being equal, a theory that is more
powerful than another also has more epistemic merit. An Ex-2nd view is thus
one that holds that a proposition p is epistemically justified for a subject if and
only if it is part of a virtuous explanatory system that beats competitor systems
in terms of explanatory virtues.
An Ex-2nd view applied to skepticism is straightforward. The conservative
aspect of Ex-2nd implies that one may begin assessing the skeptical hypothesis
with beliefs that privilege a non-skeptical view. The crucial question then is
how virtuous one’s initial view is compared with the skeptical view. This in-
volves several sub-questions: (1) How simple is the non-skeptical view and how
simple is the skeptical view? (2) How powerful is the non-skeptical view and
how powerful is the skeptical view? (3) Which view fits best with our back-
ground beliefs?
One challenge that the Ex-2nd response faces that the Ex-1st response does
not is dogmatism. Ex-2nd countenances privileging a particular starting point
over others. In response, compare Ex-2nd’s dogmatism to Moore’s dogma-
tism. Plausibly, Ex-2nd is better because it takes skepticism as a genuine hy-
pothesis that one may come to accept if its virtues are good enough. After all,
14 It is not implausible to think that any approach to epistemology that opts for a particular-
ist approach to the problem of the criterion is committed to giving some starting points
a privileged status. At a minimum particularism must allow that some of our beliefs have
some presumption in their favor when we start theorizing. See Chisholm (1973), (1982)
for discussion of particularism and the idea that some beliefs begin with presumption in
their favor. See Poston (2011) and McCain (2014b) for discussion of various explanationist
responses to the problem of the criterion.
“One could even end up … by finding that the smoothest and most adequate
overall account of the world does not after all accord existence to ordinary
physical things” (Quine 1960: 4). The simplicity and power of a skeptical hy-
pothesis may be enough to overturn its revolutionariness. So, skepticism is not
simply ruled out from the start by Ex-2nd as it is given Moorean dogmatism.
Skepticism is a live possibility for the Ex-2nd explanationist, nonetheless, the
skeptical hypothesis lacks the simplicity and power necessary to overcome its
revolutionariness.
5 Prospects for these Responses
Now that we have briefly explored both varieties of explanationist responses to

skepticism it is time to compare their strengths as well as the challenges that
they face. As we have already noted, both of these responses can be used to
respond to skepticism. Plausibly, they can both be successful in showing that
premise (2) of BRAIN is false. So, why might we opt for one explanationism
over another?
The primary difference between these explanationisms comes down to
whether the fact that a belief doesn’t conflict with any other beliefs and experi-
ences counts as some evidence for the belief. Ex-2nd allows that it does; Ex-1st
doesn’t. In a sense, epistemic conservatism is the source of both the strengths
and the challenges for these explanationisms. It is a strength of Ex-2nd be-
cause accepting epistemic conservatism makes responding to skepticism
much easier (though still not a foregone conclusion). The Ex-1st explanationist
has a harder time showing that skepticism is false because she does not privi-
lege any of her beliefs as the Ex-2nd explanationist does. Hence, a challenge for
Ex-1st lies in being able to show that skepticism really is an inferior hypothesis
when one starts from a truly neutral position. So, epistemic conservatism is a
strength of the Ex-2nd explanationist response to skepticism, and Ex-1st has a
more difficult challenge when it comes to responding to skepticism because it
does not embrace epistemic conservatism.
It seems that accepting epistemic conservatism strengthens Ex-2nd when
it comes to responding to skepticism, and denying epistemic conservatism
makes responding to skepticism more challenging for the Ex-1st explanation-
ist. One might think that this means Ex-2nd is the clear choice to make when
it comes to explanationist responses to skepticism. Concluding this may be a
bit too hasty though. After all, Ex-2nd faces a challenge that Ex-1st does not—
defending epistemic conservatism. Many epistemologists are suspicious of
epistemic conservatism; some even think it is a clear instance of “theft over
honest toil.”15 Hence, epistemic conservatism brings with it a challenge to Ex-

2nd, a challenge that Ex-1st does not share.
It is not our purpose to adjudicate between these two explanationist
responses here. Rather, our goal is simply to clarify the two distinct explanation-
ist responses to skepticism and point out where their differences lie. That be-
ing said, we think that it is plausible that both of these forms of e xplanationism
can overcome the challenges they face. Ex-2nd’s challenge can be met be-
cause epistemic conservatism can be defended from its detractors.16 Ex-1st’s
challenge of responding to skepticism from a truly neutral point can also be
overcome.17 So, what is the upshot? Both forms of explanationism are viable;
the choice of explanationist responses depends on one’s overall epistemology.
Before concluding it is worth pausing to briefly consider how these various
forms of explanationism and their chief internalist rival, Moorean dogmatism,
respond to a few key questions. While Moorean dogmatism agrees with Ex-
2nd on the existence of a privileged starting point, it remains silent on the
key explanationist themes of the importance of explanatory virtues. Addition-
ally, Moorean dogmatism might provide a response to skepticism, but it fails
to explain where the skeptic goes wrong, i.e. what’s wrong with accepting the
skeptical hypothesis. Without such an explanation Moorean dogmatism is apt
to strike many as overly dogmatic. Explanationism is a superior response to
skepticism. Which explanationist response should be accepted? Our hypoth-
esis is that it depends on whether the project of first philosophy is successful.
If it is, then Ex-1st is the better option. If first philosophy is not successful, then
neither dogmatism or skepticism follow. Ex-2nd offers a plausible account of
human knowledge.
Bibliography
Beebe, J. 2017. “Does Skepticism Presuppose Explanationism?” In K. McCain & T.

Poston (eds.), Best Explanations: New Essays on Inference to the Best Explanation,
173–187. Oxford: Oxford University Press.
15 Fumerton (2007). Also, see Foley (1983), Christensen (1994), and Feldman (2003).
16 See McCain (2008), (forthcoming a), (forthcoming b) and Poston (2012b), (2014).
17 Admittedly, there may be additional difficulties for Ex-1st when it comes to other kinds of
skepticism—particularly skepticism about memory. However, we will not delve into this
issue as it is beyond the scope of external world skepticism which is our concern here. See
Poston (2016) for discussion of the challenges that memory skepticism poses for any first
philosophy approach to epistemology.
BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard Uni-
versity Press.
BonJour, L. 1999. “Foundationalism and the External World,” Philosophical Perspectives
13: 229–249.
BonJour, L, & E. Sosa. 2003. Epistemic Justification: Internalism vs. Externalism, Founda-
tions vs. Virtues. Malden, MA: Blackwell.
Brueckner, A. 1994. “The Structure of the Skeptical Argument,” Philosophy and Phenom-
enological Research 54: 827–835.
Chisholm, R. 1973. The Problem of the Criterion. Milwaukee, WI: Marquette University
Press.
Chisholm, R. 1982. The Foundations of Knowing. Minneapolis, MN: University of Min-
nesota Press.
Christensen, D. 1994. “Conservatism in Epistemology,” Nous 28: 69–89.
Cohen, S. 1998. “Two Kinds of Skeptical Argument,” Philosophy and Phenomenological
Research 58: 143–159.
Cornman, J. 1980. Skepticism, Justification, and Explanation. Dordrecht: Reidel.
Feldman, R. 2003. Epistemology. New Jersey: Pearson Education, Inc.
Foley, R. 1983. “Epistemic Conservatism,” Philosophical Studies 43: 165–182.
Fumerton, R. 1980. “Induction and Reasoning to the Best Explanation,” Philosophy of
Science 47: 589–600.
Fumerton, R. 1992. “Skepticism and Reasoning to the Best Explanation,” Philosophical
Issues 2: 149–169.
Fumerton, R. 1995. Metaepistemology and Skepticism. Lanham: Rowman and Littlefield.
Fumerton, R. 2007. “Epistemic Conservatism: Theft or Honest Toil?” Oxford Studies in
Epistemology 2: 64–87.
Goodman, N. 1965. Fact, Fiction, and Forecast. New York: Bobbs-Merrill.
Goodman, N. 1978. Ways of Worldmaking. Cambridge, MA: Harvard University Press.
Goldman, A. 1988. Empirical Knowledge. Berkeley: University of California Press.
Harman, G. 1965. “The Inference to the Best Explanation,” Philosophical Review 74:
88–95.
Harman, G. 1973. Thought. Princeton: Princeton University Press.
Harman, G. 1980. “Reasoning and Explanatory Coherence,” American Philosophical
Harman, G. 1986. Change in View. Cambridge, MA: MIT Press.
Jackson, F. 1977. Perception: A Representative Theory. Cambridge: Cambridge University
Press.
Kuhn, T. 1977. The Essential Tension. Chicago: University of Chicago Press.
Laudan, L. 1984. Science and Values. Berkeley, CA: University of California Press.
Lehrer, K. 1971. “Why Not Skepticism?” The Philosophical Forum 2: 289–298.
Lehrer, K. 1974. Knowledge. Oxford: Oxford University Press.
Locke, J. 1690 1975. An Essay Concerning Human Understanding. Oxford: Clarendon

Press.
Lycan, W. 1988. Judgement and Justification. Cambridge: Cambridge University Press.
Lycan, W. 2002. “Explanation and Epistemology.” In P. Moser (ed.), The Oxford Hand-
book of Epistemology, 408–433. Oxford: Oxford University Press.
Maddy, P. 2007. Second Philosophy: A Naturalistic Methodology. Oxford: Oxford Univer-
sity Press.
McCain, K. 2008. “The Virtues of Epistemic Conservatism,” Synthese 164: 185–200.
McCain, K. 2012. “A Predictivist Argument against Scepticism,” Analysis 72: 660–665.
McCain, K. 2013. “Two Skeptical Arguments or Only One?” Philosophical Studies 164:
289–300.
McCain, K. 2014a. Evidentialism and Epistemic Justification. New York: Routledge.
McCain, K. 2014b. “The Problem of the Criterion,” Internet Encyclopedia of Philosophy.
McCain, K. 2016. The Nature of Scientific Knowledge: An Explanatory Approach. Swit-
zerland: Springer.
McCain, K. Forthcoming a. “Epistemic Conservatism and the Basing Relation.” In P.
Bondy & J. Adam Carter (eds.), Well-Founded Belief: New Essays on the Epistemic Bas-
ing Relation. New York: Routledge.
McCain, K. Forthcoming b. “Epistemic Conservatism: A Non-Evidentialist Epistemolo-
gy?” In L. Moretti & N. Pedersen (eds.), Non-Evidentialist Epistemology. Boston: Brill.
McCain, K. & T. Poston. 2014. “Why Explanatoriness Is Evidentially Relevant,” Thought
3: 145–153.
Moore, G.E. 1939. “Proof of an External World,” Proceedings of the British Academy 25:
273–300.
Moser, P.K. 1989. Knowledge and Evidence. Cambridge: Cambridge University Press.
Moser, P.K. 1990. “Two Roads to Skepticism.” In M. Roth & G. Ross (eds.), Doubting:
Contemporary Perspectives on Skepticism, 127–140. Dordrecht: Kluwer.
Neurath, O. 1983. Philosophical Papers 1913–1946. In R.S. Cohen & M. Neurath (eds.),
Dordrecht: Reidel.
Poston, T. 2011. “Explanationist Plasticity and the Problem of the Criterion,” Philosophi-
cal Papers 40: 395–410.
Poston, T. 2012a. “Basic Reasons and First Philosophy,” The Southern Journal of Philoso-
phy 50: 75–93.
Poston, T. 2012b. “Is There an ‘I’ in Epistemology?” Dialectica 66: 517–541.
Poston, T. 2013a. “BonJour and the Myth of the Given,” Res Philosophica 90: 185–201.
Poston, T. 2013b. “Is A Priori Justification Indispensible?” Episteme 10: 317–331.
Poston, T. 2014. Reason & Explanation: A Defense of Explanatory Coherentism. London:
Palgrave Macmillan.
Poston, T. 2016. “Acquaintance and Skepticism about the Past.” In B. Coppenger & M.
Bergmann (eds.), Intellectual Assurance: Essays on Traditional Epistemic Internal-
ism, 183–204. Oxford: Oxford University Press.
Roche, W. & E. Sober. 2013. “Explanatoriness is evidentially irrelevant, or inference to

the best explanation meets Bayesian confirmation theory,” Analysis 73: 659–668.
Quine, W.V.O. 1960. Word and Object. Cambridge, MA: MIT Press.
Quine, W.V.O. 1990. In Pursuit of Truth. Cambridge, MA: Harvard University Press.
Quine, W.V.O. & J.S. Ullian. 1970. The Web of Belief. New York: Random House.
Pritchard, D. 2002. “Resurrecting the Moorean Response to the Sceptic,” International
Journal of Philosophical Studies 10: 283–307.
Pritchard, D. 2005. “The Structure of Sceptical Arguments,” Philosophical Quarterly 55:
37–52.
Pryor, J. 2000. “The Skeptic and the Dogmatist,” Nous 34: 517–549.
Pryor, J. 2004. “What’s Wrong with Moore’s Argument?” Philosophical Issues 14: 349–378.
Reed, B. Forthcoming. “Skepticism as a Way of Life.” In K. McCain and T. Poston (eds.),
The Mystery of Skepticism: New Explorations. London: Brill.
Russell, B. 1912. The Problems of Philosophy. Oxford: Oxford University Press.
Sosa, E. 2007. A Virtue Epistemology. Oxford: Oxford University Press.
Sosa, E. 2009. Reflective Knowledge. Oxford: Oxford University Press.
Stroud, B. 1984. The Significance of Philosophical Sceptism. Oxford: Oxford University
Press.
van Fraassen, B. 1980. The Scientific Image. Oxford: Oxford University Press.
van Fraassen, B. 1989. Laws and Symmetry. Oxford: Clarendon Press.
Vogel, J. 1990. “Cartesian Skepticism and Inference to the Best Explanation,” Journal of
Vogel, J. 2005. “The Refutation of Skepticism.” In M. Steup & E. Sosa (eds.), Contempo-
rary Debates in Epistemology, 72–84. Malden, MA: Blackwell Publishing.
Willenken, T. 2011. “Moorean Responses to Skepticism: A Defense,” Philosophical Stud-
ies 154: 1–25.
Chapter 14
Reasoning One’s Way Out of Skepticism

Susanna Rinard
1 Introduction
Once someone has accepted the argument for external world skepticism, could
any line of reasoning persuade them that knowledge of the external world is
possible after all?
Many contemporary epistemologists think not. Here, for example, is Timothy
Williamson (2000: 27):
Nothing said here should convince someone who has given up ordinary
beliefs that they constitute knowledge…This is the usual case with philo-
sophical treatments of skepticism: they are better at prevention than at
cure. If a refutation of skepticism is supposed to reason one out of the hole,
then skepticism is irrefutable. (emphasis mine)
And James Pryor (2000: 517–520):
The ambitious anti-skeptical project is to refute the skeptic on his own

terms, that is, to establish that we can justifiably believe and know such
things as that there is a hand, using only premises that the skeptic allows
us to use. The prospects for this ambitious anti-skeptical project seem
somewhat dim…Most fallibilists concede that we can’t demonstrate to
the skeptic, using only premises he’ll accept, that we have any perceptual
knowledge….the ambitious anti-skeptical project cannot succeed. (empha-
sis mine)
I aim to show that this widely held view is mistaken. I think it is possible to
rationally persuade an external world skeptic that we have knowledge of the
external world. This paper presents an argument—which appeals only to prem-
ises that an external world skeptic could accept—for the claim that rationality
requires us to believe that skepticism is false.
The strategy is to argue that accepting the argument for external world
skepticism ultimately commits one to more extreme forms of skepticism in a
way that is self-undermining. Section 2 presents the argument for skepticism

Reasoning One’s Way Out of Skepticism 241
about the external world and shows that there is a parallel argument for skepti-
cism about the past. Section 3 argues that skepticism about the past leads to
skepticism about complex reasoning. Section 4 argues that it would be self-
undermining to accept skepticism about complex reasoning on the basis of the
argument from skepticism about the past, since that argument is complex. In
particular, one would end up believing a proposition P while believing that one
should not believe P. This combination of beliefs is not rational. So, Section 5
concludes that it is not rational to accept the argument for external world skep-
ticism, because doing so ultimately commits one to an irrational combination
of beliefs. Section 6 replies to objections. Section 7 argues that suspending
judgment on skepticism is also irrational. Section 8 argues that doxastic di-
lemmas are impossible—rationality cannot prohibit every possible doxastic
attitude—and so rationality requires us to disbelieve skepticism. Section 9 ar-
gues that the resulting anti-skeptical position is not subject to instability or
loop effects.1
In the course of giving this line of reasoning I take a stand on several issues
about which there is ongoing debate, such as the role of memory in complex
reasoning; which cognitive achievements are possible in a single moment;
which level-bridging principles hold; and whether doxastic dilemmas are pos-
sible. Space constraints preclude entering fully into the details of all these de-
bates here. So I cannot claim that my argument should convince every possible
external world skeptic. What I do claim, however, is that in each case, the view
I endorse is plausible, and—importantly—accepting external world skepti-
cism would not give one any special reason to deny it. If so, then my argument
should be rationally persuasive to a skeptic who agrees with the stances taken
here on several orthogonal issues in epistemology. If so, then I will have shown,
contra Williamson, Pryor, and others, that reasoning one’s way out of skepti-
cism is not impossible.
One consequence of this result is that we should encourage those who are
inclined to pursue the project of convincing skeptics other than those ad-
dressed here. Since, as I argue here, it is not impossible to convince a skeptic,
we have no reason, in advance, to expect such projects to fail; so, attempting
them is worthwhile, and those who do so should be taken seriously.
1 This may remind some readers of Wright (1991). While there are many similarities, the line of
reasoning presented here is importantly different from Wright’s. Moreover, Wright’s project
faces difficulties. Wright sets up the argument for external world skepticism in a distinctive
way. There is a different, more common version of the argument to which Wright’s criticisms
do not apply. Wright’s project is criticized in Brueckner (1992), Pritchard (2001), and Tymoc-
zko and Vogel (1992). Jessica Wilson (2012) also argues that skepticism is self-undermining,
but her strategy is quite different from mine.
242 Rinard
Concerning my own project in this paper, it is important to note that I

a ddress only those whose skepticism is based on traditional philosophical ar-
guments for external world skepticism. I am not aware of any other plausible
arguments for external world skepticism, but nothing said here establishes
that there aren’t any, or that, if there were, it would be irrational to accept skep-
ticism on their basis.
I’ll make a few final big picture remarks before moving on to the substance
of the argument. As noted earlier, most contemporary anti-skeptical projects
do not aim to convince an external world skeptic. In this respect, my project is
more ambitious than theirs. But there is another important difference—a re-
spect in which my project is less ambitious: I don’t try to diagnose the flaw in
the skeptical argument. I don’t isolate a particular premise as false, and explain
why, despite its falsity, we found it compelling. In this respect my project is simi-
lar to G.E. Moore’s (1962); he also aimed to establish that we should reject the
skeptic’s conclusion, but did not in the process diagnose the flaw in the skeptic’s
argument.
Moore held that the skeptic’s conclusion can be rejected because it is in-
compatible with common sense. Many contemporary epistemologists have a
broadly Moorean outlook, on which we can be sure, prior to substantive philo-
sophical inquiry, that the skeptical conclusion is false. On this view, the pri-
mary task for the epistemologist who seeks to respond to skepticism is simply
to identify which premise is false, and explain why, despite its falsity, we found
it so compelling. From this perspective, it may seem that my project addresses
one of the less important questions concerning skeptics and skepticism.
However, I have argued elsewhere (Rinard 2013) that this Moorean approach
is deeply flawed. Philosophical argument is, I have claimed, capable of ratio-
nally overturning our common sense convictions. This means that we cannot
simply dismiss the skeptic; we must take seriously the possibility that they may
be right. Rejecting the Moorean approach lends greater urgency to projects
like the one pursued in this paper: our right to hold even our most basic beliefs
about the world stands or falls with their success.
2 External World Skepticism Leads to Skepticism about the Past
This section argues that if it is rational to accept external world skepticism, then
it is rational to accept skepticism about the past. This is because there is an argu-
ment for the latter that is perfectly analogous to the argument for the former.2
2 There are many different formulations of the argument for external world skepticism. All
share the same basic strategy, originating in Descartes (1996). It’s plausible that all extend
It’s possible that the external world is largely as you believe it to be. Call this
scenario “Normal.” But there is also a possibility in which the way things ap-
pear to you is exactly the same, but these appearances are radically deceptive;
you are merely a bodiless brain in a vat (biv). The skeptic’s argument begins:
(1) One’s basic evidence about the external world is restricted to proposi-
tions about the way the external world appears.
The skeptic goes on to claim that this evidence is neutral between Normal
and biv; it doesn’t favor one over the other. After all, both hypotheses entail
that one has the perceptual evidence that one does, e.g. that one seems to see
hands, tables, chairs, etc. Since the hypotheses predict the evidence equally
well, they are equally well supported by the evidence.3,4 Thus the skeptic’s sec-
ond premise (followed by the third):
(2) Propositions about the way the external world appears are evidentially
neutral between Normal and biv.
(3) Neither Normal nor biv is intrinsically more worthy of belief, indepen-
dently of one’s evidence.5
to skepticism about the past, but space constraints allow detailed discussion of only one—
which I take to be one of the strongest. Here I comment briefly on two others. Some common
versions of the closure argument begin with the premise that for all one knows, one is a brain
in a vat. But the skeptic’s argument is stronger if it provides some further justification for
this claim, rather than taking it as an unargued premise. The underdetermination argument
focuses on the existence of a gap between sensory evidence and external world beliefs. But
what is the nature of this gap? If it is merely logical—that sensory evidence doesn’t entail
external world propositions—then inductive skepticism is required for the argument to even
get off the ground. Alternatively, if the gap is epistemic, then some further motivation is
needed.
3 Here, evidential support is incremental support, not overall support. The overall worthiness
of belief of a hypothesis depends both on (1) its worthiness of belief, independently of (prior
to) one’s evidence; and (2) the incremental support from one’s empirical evidence.
4 Kevin McCain (2012) argues that while common sense hypotheses genuinely predict the evi-
dence, skeptical hypotheses merely accommodate it. Here, I use “prediction” as equivalent
to “probabilification.” On this usage, a hypothesis predicts the evidence to the extent that it
probabilifies it. Since Normal and biv both entail the evidence, they both predict it, in this
sense, equally well.
5 Features such as simplicity, coherence, etc. are sometimes thought to make for greater in-
trinsic worthiness of belief. With premise (3), the skeptic is denying either that Normal has
greater simplicity/unification/etc. than biv, or that differences of this kind make for greater
worthiness of belief.
244 Rinard
From (1)–(3) it follows that one neither knows, nor is justified in believing, that
biv is false. Just one more premise—the closure principle—is needed for full-
on external world skepticism:6
(4) If one neither knows nor is justified in believing Q, and one knows that P
entails Q, then one neither knows nor is justified in believing P.
(1)–(4) yield the skeptic’s conclusion:
(5) For many external world propositions P, one neither knows nor is justi-
fied in believing P.7
To construct an analogous argument for skepticism about the past, first, con-
sider a more detailed version of biv. Suppose your creators want to deceive you
about your past as well as your external surroundings. Due to budgetary con-
straints, they can afford to keep your brain in existence for only one minute;
but, since they want to simulate a typical human experience, they implanted
your brain with false apparent memories such that what it’s like to have these
apparent memories is exactly the same as what it’s like for you in Normal to re-
ally remember what happened. Call this scenario biv(NoPast).8
We can now construct an argument for skepticism about the past simply by
taking our argument for external world skepticism and replacing “the external
world” with “the past,” and “biv” with “biv(NoPast)”:
(1*) One’s basic evidence about the past is restricted to propositions about
the way the past appears (i.e. the way one seems to remember things hav-
ing been).
(2*) Propositions about the way the past appears are evidentially neutral be-
tween Normal and biv(NoPast).
(3*) Neither Normal nor biv(NoPast) is intrinsically more worthy of belief,
independently of one’s evidence.
6 Although most accept it, Dretske (1970) and Nozick (1981) are two prominent deniers of
the closure principle for knowledge. Note, however, that it is far less common, and far more
implausible, to deny closure for justification. The argument for skepticism about justification
remains intact even if closure for knowledge is rejected.
7 Why many, and not all? Because this particular argument leaves a few beliefs untouched (e.g.
a brain exists). But it undermines the bulk of our substantive external world beliefs (e.g. I
have hands, tables exist, etc.).
8 Perhaps the most famous skeptical hypothesis concerning the past is Russell’s (1921, 159), in
which the world sprang into existence five minutes ago, complete with a group of people
who seem to remember what we actually remember.
(4*) If one neither knows nor is justified in believing Q, and one knows that P
entails Q, then one neither knows nor is justified in believing P.
Therefore,
(5*) For many propositions P about the past, one neither knows nor is justi-
fied in believing P.
One who accepts the premises of the argument for external world skepticism
should also accept the premises of the argument for skepticism about the
past. It would be unacceptably arbitrary to accept (1) while rejecting (1*).
The justification given for (2) carries over to (2*). Any reason for rejecting
(3*) would constitute an equally good reason for rejecting (3). (4) and (4*) are
identical.
This section defended the following claim:
Claim i: If it is rational to accept external world skepticism, then it is

rational to accept skepticism about the past.
3 Skepticism about the Past Leads to Skepticism about

Complex Reasoning
The rough idea behind the argument in this section is as follows: in complex
reasoning one relies on one’s memory. But if skepticism about the past is true,
one is not justified in relying on one’s memory, and so not justified in believing
the conclusions of complex reasoning.9
What do I mean by “complex reasoning,” and “skepticism about complex
reasoning”? Reasoning counts as complex when it involves multiple steps, not
all of which can be held in one’s head at once—that is, one cannot, all in one
moment, consciously grasp each step and how they all come together to yield
the conclusion. For example, suppose one begins with some premise A, and
then infers (either deductively or inductively) B from A, C from B, and so on,
finally concluding that G. Suppose that, by the time one infers G from F, one
no longer has in one’s head the details of the argument by which one reasoned
from A to G; one simply seems to remember having done so. Then the reason-
ing from A to G counts as complex. Most proofs in math and logic are complex;
so are non-deductive arguments for, say, the occurrence of climate change,
or the claim that the stock market will have an average annual return of at
9 Richard Fumerton (1995: 49) argues, in a similar vein, that skepticism about the past leads to
skepticism about reasoning that takes place over time.
246 Rinard
least 8% over the next century. Most interesting philosophical arguments are
complex.10
Skepticism about complex reasoning is the view that one could not come to
know, or be justified in believing, any proposition on the basis of complex rea-
soning. I will now sketch an argument for skepticism about complex reasoning
which has skepticism about the past as a premise. (I will then examine one of the
steps of this argument in more detail.) Let G be the conclusion of an arbitrary
complex argument. Consider an agent who is initially not justified in believing
G. They then carefully and correctly go through the argument for G. Since the
argument is complex, at the moment they conclude that G, they don’t have in
their head the earlier steps of the argument. They merely seem to remember
that they went through an argument for G. But if skepticism about the past is
true, they are not justified in trusting their apparent memory, because they are
not justified in believing any proposition about the past. For all they knows, they
haven’t even been in existence long enough to have gone through an argument
for G. So, by the time they conclude that G, they are not justified in believing it.
(Further discussion of this last step appears in the paragraph after the next one.)
That is, if skepticism about the past is true, then despite having gone through
a complex argument for G, the agent is not justified in believing it. So it follows
from skepticism about the past that one cannot come to know, or be justified
in believing, a proposition by going through a complex argument for it.
I’ll now examine in more detail the last step of the argument just sketched.
That step was:
(*) If an agent does not have in their head the argument for G, is not
justified in trusting their apparent memory that they went through an ar-
gument for G, and has no independent reason for believing G, then they
are not justified in believing G, even if they did in fact go through a good
argument for G.
First, a preliminary remark. While I think (*) is highly plausible, it is worth

noting that my project does not actually require (*) to be true. All it requires
is that some external world skeptics would accept it. This is because, as noted
in the introduction, my overall aim is to show how it is possible for an external
world skeptic to reason their way out of skepticism. So it is not a problem for
my project if some contemporary epistemologists reject (*), as long as some
10 Pasnau (2014) recounts a lively debate, going back to the Middle Ages and beyond, con-
cerning which arguments can be grasped, in their entirety, all at once. (For example, Burge
(1993) denies that even single-step inferences can be grasped all at once.) Poston (2016)
contains an extensive discussion of how much is contained in the “fleeting present.”
external world skeptics would accept it. (*) is endorsed by many contemporary
epistemologists (myself included), and an external world skeptic has no spe-
cial reason to reject it.
That said, it will nonetheless be interesting to consider some views which
are, and some which are not, compatible with (*). One view that is incompat-
ible with (*) is the view that anyone who competently deduces some proposi-
tion P from known premises thereby comes to know P. This view has some
implausible consequences, however. Suppose that Candace carefully and cor-
rectly goes through a complex argument for P. When she finally infers P, she
no longer has in her head the steps of this argument; she merely seems to re-
member having gone through some argument or other for P (and she has no
independent reason to believe P). Then, Candace learns that she is under the
influence of a drug that makes one’s memory unreliable, in the following sense:
people who have taken this drug often have false apparent memories. Much of
what they seem to remember never, in fact, happened. It is plausible that upon
learning this, Candace should suspend judgment on whether she did, in fact,
go through an argument for P, and, consequently, suspend judgment on wheth-
er P is true. After all, she does not have in her head any argument for P, and she
has no independent reason to believe P. However, according to the view just
stated, since she did in fact competently deduce P from known premises, she is
justified in believing P. This, I think, is the wrong result.
A defender of (*) can see skepticism about the past as playing a role analo-
gous to the role played, in the example just given, by Candace’s knowledge of
having taken the drug. What the knowledge of the drug does, first and fore-
most, is make it unreasonable for Candace to trust her apparent memories. It
is this that then makes it unreasonable for her to believe propositions that she
has in fact derived via complex reasoning. Skepticism about the past has the
same effect. First and foremost, it makes it unreasonable for the agent to trust
their apparent memories. Just as in the drug case, this means that it is unrea-
sonable for the agent to believe the deliverances of complex reasoning. Skepti-
cism about the past, and knowledge of having taken a memory-distorting drug,
both have this result, because both make the agent unjustified in trusting their
apparent memories.
Some philosophers have suggested to me, in conversation, that (*) might also
be rejected by a proponent of the view that, although positive reason for doubt-
ing one’s memory can undermine justification in the results of complex reason-
ing, in the absence of such defeaters, one has a default entitlement to trust one’s
memory (and therefore complex reasoning). This view is defended by (among
others) Tyler Burge in a dispute with Roderick Chisholm (Burge (1993) and Ch-
isholm (1977)). Burge and Chisholm disagree about whether propositions about
the past are part of one’s justification for the conclusion of a complex a rgument.
248 Rinard
Chisholm affirms this; Burge denies it. Clearly, Chisholm would endorse (*).
Burge’s case is not initially so obvious, but I will argue that he would as well.
Although Burge denies that propositions about the past are part of one’s
justification for G, he does allow that there is some sense in which one relies on
one’s memory in complex reasoning. As noted above, he thinks that if one has
positive reason for doubting one’s memory, then one is not justified in trusting
complex reasoning. The antecedent of (*) states that the agent is not justified
in believing what she seems to remember. On Burge’s view, if this is true, then
it must be that she has positive reason for doubting her memory. Burge agrees
that, given this, the consequent of (*)—that trusting complex reasoning is not
justified—follows. So Burge would agree with (*).
What Burge would reject is the idea that philosophical arguments for skepti-
cism about the past constitute good positive reason for doubting one’s memory.
But it is not my aim to defend that claim. The aim of this section is merely to estab-
lish that if skepticism about the past is true, then so is skepticism about complex
reasoning; and, as I have argued, Burge should endorse that conditional claim.
One view that is incompatible with (*) is the view that coming to believe P
via a reliable process is always sufficient for justification in P, even if one has
good reason to doubt that one’s process was reliable. But this view is in direct
conflict with both skepticism about the external world and skepticism about
the past, so such skeptics would not endorse that view. (According to the re-
liabilist view, if one’s senses and memory are in fact reliable, one is justified
in trusting them, whereas according to the skeptic, no one is ever justified in
trusting one’s senses or memory.)
To summarize, this section has argued that the following claim should be
accepted by those external world skeptics who accept (*):
Claim ii: If it’s rational to accept skepticism about the past, then it’s
rational to accept the argument from skepticism about the
past to skepticism about complex reasoning.
4 It is Not Rational to Accept Skepticism about Complex Reasoning
This section argues that it is not rational to accept the argument, described
in Section 3, from skepticism about the past to skepticism about complex
reasoning.
First, notice that this argument is itself complex. I cannot consciously grasp,
all at once, why each step of the argument is plausible and how all the steps
come together to support the conclusion (see Section 6 for a related objection
and reply). But the conclusion of this argument is that it is not rational to accept
complex arguments. So there is a sense in which the argument is self-under-

mining. As I will argue in this section, the self-undermining character of this ar-
gument manifests itself in the fact that if one accepts it, one ends up believing
a proposition P while at the same time believing that it is not r ational to believe
P. This is an irrational combination of beliefs. So accepting the argument is not
rational, since doing so results in an irrational combination of beliefs.
To see why, suppose one were to accept the argument from skepticism about
the past to complex reasoning skepticism. Let P be the conclusion of this argu-
ment (i.e. the thesis of skepticism about complex reasoning). At the moment
one accepts P, one knows one is not accepting it on the basis of a simple argu-
ment. After all, if one were accepting it on the basis of a simple argument, one
would have all of the steps of that argument in one’s head at the moment one
accepts P. However, one can tell at the moment of acceptance that one does
not have in one’s head all the steps of an argument for P.
Since one knows one is not accepting P on the basis of a simple argument,
one knows that one of the two remaining possibilities obtains: either one is ac-
cepting P on the basis of a complex argument, or one’s acceptance of P is not
based on any argument at all. Since one is a skeptic about complex reasoning,
one believes that if the first possibility obtains, one’s belief in P is not ratio-
nal. Consider now the second possibility. Recall that P is the proposition that
skepticism about complex reasoning is true. Perhaps there are some proposi-
tions one could rationally believe without basing one’s belief on an argument
(“1 = 1,” perhaps), but if there are, skepticism about complex reasoning is not
among them. It is a highly surprising claim, far from obvious. So one also be-
lieves that if the second possibility obtains, one’s belief in P is not rational. So,
one believes that one’s belief in P is not rational, no matter which of these two
possibilities obtains.
That is, at the moment one accepts the conclusion of the argument for
skepticism about complex reasoning, one believes P and one also believes that
one’s belief in P is not rational. But this is not a rational combination of beliefs.
So it is not rational to accept the argument from skepticism about the past to
skepticism about complex reasoning, because doing so results in an irrational
combination of beliefs.
The foregoing relied on the following principle:
Anti-Denouncement: It is not rational to believe a proposition P while

also believing that it is not rational for one to be-
lieve P.
The idea is that it is not rational to denounce one’s own belief, in the sense of
believing it to be irrational. This principle is highly plausible. However, it is not
250 Rinard
universally accepted, and there is debate about whether it holds in all cases.11
It is beyond the scope of this paper to enter into this debate. I will simply de-
scribe one case that helps bring out the plausibility of this principle, and note
that the aim of this paper will be achieved if the overall argument given here
would be convincing to those skeptics who do accept Anti-Denouncement.
Imagine someone—let’s call him Randall—who’s wondering whether it
would be a good idea for him to invest his retirement savings in the stock mar-
ket. He’s sure of the following conditional claim: if the market will have an
average annual return of at least 8% over the next few decades, then he ought
to invest his savings in the market. Naturally, he then turns to the question of
whether the antecedent of this conditional is true.
Suppose that, upon careful consideration of the evidence available to him,
Randall concludes that, given his evidence, he definitely should not believe
that the market will return at least 8%. This belief, he is sure, would not be
rational, given his evidence. But suppose further that, despite this, he does be-
lieve that the market will return at least 8%, and decides on that basis to invest
his savings entirely in the market. I think that in this case, Randall should be
regarded as irrational. A rational person would not believe that the market will
return at least 8% while also believing at the same time that, given his evi-
dence, he should believe no such thing.
This judgment about this case suggests that Anti-Denouncement is true.
If Anti-Denouncement were false, it’s hard to see what could be wrong with
Randall’s beliefs. But it seems clear that they are not the beliefs of a rational
person.
This section defended the following claim:
Claim iii: It is not rational to accept the argument from skepticism
about the past to complex reasoning skepticism.
5 It is Not Rational to Accept External World Skepticism
To summarize, this section brings together the three claims defended in Sec-
tions 2–4. They entail that it is not rational to accept external world skepticism.
11 Weatherson (unpublished manuscript), Williamson (2011), and Lasonen-Aarnio (2014)

argue against related claims. Principles along these lines are defended in Feldman (2005),
Bergmann (2005), Horowitz (2014), and Greco (2014), and discussed in Christensen (2010),
Elga (2013), and Titelbaum (2015).
Claim i: If it’s rational to accept external world skepticism, then it’s
rational to accept skepticism about the past.
Claim ii: If it’s rational to accept skepticism about the past, then it’s
rational to accept the argument from skepticism about the
past to skepticism about complex reasoning.
Subconclusion: If it’s rational to accept external world skepticism, then
it’s rational to accept the argument from skepticism
about the past to skepticism about complex reasoning.
Claim iii: It is not rational to accept the argument from skep-
ticism about the past to skepticism about complex
reasoning.
Conclusion: It is not rational to accept external world skepticism.
6 Objections and Replies
Objection i:
The argument in Section 4 for the claim that it is not rational to accept the ar-
gument from skepticism about the past to skepticism about complex reasoning
rests on the assumption that this argument is complex. But it’s not clear that it is.
Reply:
I think it is plausible that this argument is complex; I, at least, am not able to
consciously appreciate, all at once, each individual step of the argument, why
it is plausible, and how exactly all the steps combine to support the conclusion.
However, my argument would go through even if this argument were simple,
since this argument is in fact only a small part of the overall argument for skep-
ticism about complex reasoning. The overall argument includes the argument
for skepticism about the past, and the arguments for the premises of that argu-
ment (from the parallel with external world skepticism).
That is, the entire argument for skepticism about complex reasoning in-
cludes the argument for external world skepticism, the argument linking
external world skepticism to skepticism about the past, the argument for skep-
ticism about the past, and the argument from skepticism about the past to
skepticism about complex reasoning. This argument is surely complex.12
12 As noted above, Pasnau (2014) presents a history of an extended debate in philosophy

concerning which arguments can be grasped in their entirety at once. Some, such as
Burge (1993), hold that we are unable to do this even for very short arguments.
252 Rinard
Objection ii:
It may be plausible that, for actual humans, the argument for skepticism
about complex reasoning is complex, since we are unable to hold this entire
argument in our heads at once. However, whether an argument is complex or
simple is agent-relative; it is (metaphysically) possible for there to be an agent,
who although in all other respects is just like us, is able to hold incredibly long
arguments in their head at once. In particular, they can hold in their head the
entire argument for skepticism about complex reasoning. For this agent, this
argument is simple, and so it would not be self-undermining for them to accept
it—there is nothing self-undermining about accepting a simple argument for
skepticism about complex reasoning. Since this agent is in all other respects
like us, they will find each premise of the argument individually plausible, and
so, since the argument is simple for them, they will accept it, and come to be-
lieve skepticism about the external world, the past, and complex reasoning.
In short: if there were an agent with certain enhanced cognitive abilities, they
would be a skeptic.
The reasoning just given (continues the objector) should be accepted by
anyone who accepts the argument in Sections 2–5. Such a person would then
be in the following peculiar situation: because they accept the argument in
Sections 2–5, they think it would not be rational for them to believe skepticism,
and so they don’t believe it. But, at the same time, they know that if there were
an agent just like them, except with certain enhanced cognitive abilities, that
agent would believe skepticism. This combination of beliefs is not rational, ac-
cording to the following principle:
Deference: If one (rationally) believes that a cognitively enhanced ver-

sion of oneself would believe P, then rationality requires one
to believe P.
Deference is very plausible. For example, suppose you’re uncertain about

whether Goldbach’s conjecture can be proved. You then learn that if there
were a version of yourself with enhanced cognitive abilities—specifically, en-
hanced mathematical abilities—that enhanced agent would believe that there
is a proof of Goldbach’s conjecture. Plausibly, upon learning this information,
you are rationally required to believe there is such a proof. This suggests that
Deference is true. If so, then it would not be rational to accept the argument
given in Sections 2–5.
Reply:
I agree with the objector that Deference is plausible, and I agree that the
Goldbach’s Conjecture case shows that something in the vicinity of this
rinciple must be true. Nevertheless, we have independent reason to think

p
that, as stated, Deference is false. I will argue that the properly revised version
of Deference does not have the consequence that anyone who accepts the ar-
gument in Sections 2–5 has an irrational combination of beliefs.
First, though, note that the objector is not obviously correct in assuming
that an enhanced agent would believe skepticism. It might be that if one were
enhanced in this way, one would no longer believe the skeptic’s premises. Also,
even supposing that the enhanced agent would believe skepticism, the non-
skeptic may have good reasons for thinking that an ideally rational agent would
not.13 (Since the enhanced agent is just like us in every respect other than this
particular enhancement, they are not ideally rational.) Nevertheless, for the re-
mainder of my reply I will assume the objector is right that the enhanced agent
would accept skepticism, since I think the objection fails in any case, because
the key principle on which it relies does not hold in the case of the non-skeptic.
The following consideration shows that Deference, as stated, is false. As the
objector noted, we have good reason to doubt that an enhanced agent actually
exists. However, according to Deference, we are rationally required to believe
that such an agent does exist. This is because we know that if there were an
enhanced agent, they would know that they are enhanced in a certain way,
and so they would believe that an enhanced agent (namely, themselves) exists.
According to Deference, one should believe whatever one knows an enhanced
agent would believe, so according to Deference, one should believe that an en-
hanced agent actually exists. Clearly this is the wrong result, and so Deference,
as stated, is false.14,15
Nevertheless, the Goldbach’s conjecture case shows that some principle in
the vicinity of Deference must be true. The crucial question, then, is this: will
the correct version of Deference (whatever it is) still entail that it would not be
rational to decline skepticism on the basis of the argument in Sections 2–5? Or
is this just another example in which Deference, as originally formulated, gets
the wrong result?
13 Assuming the non-skeptic believes that skepticism is false (rather than merely failing to
believe it’s true), they must believe that one of the skeptic’s premises is false. It might
seem plausible that an ideally rational agent would believe every true necessary proposi-
tion. If so, then the non-skeptic must think that an ideally rational agent would disbe-
lieve whichever premise of the skeptical argument is in fact false, and so would not be a
skeptic.
14 Similar arguments appear in Plantinga (1982).
15 One might respond by modifying Deference as follows: One should believe whatever an
enhanced agent would advise one to believe. Presumably an enhanced agent would not
advise you to believe that an enhanced agent exists. The question is then whether they
would advise you to believe skepticism. The rest of the reply to this objection can be seen
as a reason for thinking they would not.
254 Rinard
I think we have independent reason for thinking that the second possibility
obtains. This is because there is another counterexample to Deference, and the
most natural explanation for why Deference fails in this case has the conse-
quence that Deference fails in the case of the non-skeptic as well.
Suppose one learns that, if there were an enhanced version of oneself, that
enhanced agent would believe that Deference is false. In particular, the en-
hanced agent would believe the following: the fact that an enhanced version
of oneself believes P is never a good reason for believing P. According to Defer-
ence, upon learning this, one should come to believe that Deference is false.
But that is clearly not the rational response to the situation. To do this would
be self-undermining. It would not be rational to believe, on the basis of Defer-
ence, that Deference is false. To do so would be to believe a proposition P (that
one should never adopt a belief on the basis of Deference) while believing that
one should not believe P (because the basis for one’s belief in P is that Defer-
ence says one ought to believe it).
This suggests that Deference fails in cases in which, if one were to believe
what one believes the enhanced agent would believe, one’s position would be
self-undermining (in the sense that one would believe P while believing that
one should not believe P).
But this is true of the person who accepts my argument. Suppose this non-
skeptic were to adopt the belief, on the basis of Deference, that skepticism is
true. That is, suppose they were to reason as follows: an enhanced agent would
believe skepticism. One should believe whatever one believes an enhanced
agent would believe. So I should believe skepticism.
If they accept skepticism on the basis of this argument, their position is self-
undermining,16 because the above argument for skepticism about complex
reasoning is complex. (This is because it relies on the assumption that an en-
hanced agent would believe skepticism, and the argument for this is complex.)
I take this to show that Deference gives the wrong result in the case of the
agent who, on the basis of my argument, gives up the belief that skepticism is
true. This is because we have independent reason to believe that Deference
16 The argument for this can be spelled out in more detail as follows. They believe a propo-
sition, P (skepticism about complex reasoning). They know that they do not believe P
on the basis of a simple argument (the argument in the above paragraph is not simple,
because it relies on the claim that an enhanced agent would believe skepticism; the argu-
ment for this is complex.). They also know that P is not the kind of proposition that could
be rationally believed on the basis of no argument. The only remaining possibility is that
they believe P on the basis of a complex argument (this is in fact the case); but since they
accept skepticism about complex reasoning, they believe that in this case they should not
believe P. So they believe P while believing that they should not believe P.
fails in cases in which following it would lead one into a self-undermining posi-
tion, and this is true in the case of the non-skeptic.
More can be said to explain why Deference fails in such cases. I think the
plausibility of principles like Deference stems from a picture we have about
the role of idealized agents in epistemology. According to this picture, the ra-
tionality of one’s position increases as one’s position becomes more similar,
overall, to the position of an idealized agent. One important respect of similar-
ity concerns the contents of one’s beliefs. Other things equal, adopting beliefs
that are shared by an idealized agent makes one’s position more rational. That
is why this picture makes Deference seem plausible.
But this very same picture also explains why Deference fails in certain cases.
Similarity in the contents of one’s beliefs is not the only kind of similarity that
counts.17 Moreover, sometimes, for limited agents, becoming more similar in
the content of one’s beliefs involves becoming less similar in another impor-
tant respect. Deference fails to take this into account; it focuses on only one
respect of similarity.
The position of the limited agent, the non-skeptic, differs from the p osition
of the enhanced agent in that the former does not believe skepticism, but the
latter does. However, the positions of the limited agent and the enhanced
agent are similar in the following important respect: both positions are not
self-undermining. If the limited agent were to adopt the enhanced agent’s
belief, their position would become less similar in this important respect,
because their position would now be self-undermining. Deference fails in this
case because matching beliefs would make the limited agent overall less simi-
lar to the enhanced agent, because it would make the limited agent’s position
self-undermining, unlike the position of the enhanced agent.
So we see that the motivating idea behind Deference also helps explain
why Deference fails in certain cases, like the case of the non-skeptic. The
motivating idea is that one’s position should be as similar as possible to the
enhanced agent’s position. The problem is that Deference focuses on only
one respect of similarity. Usually, this doesn’t matter, because becoming more
similar in this respect doesn’t usually make one less similar in other respects.
But occasionally, as in the case of the non-skeptic, it does. In such cases, Def-
erence fails.
17 For example, consider a complex mathematical theorem M which one has no reason for
believing. One shouldn’t believe M, even though an enhanced agent would (one doesn’t
know that an enhanced agent would believe it). This example makes the general point
that the rationality of one’s position depends on more than just the overall similarity of
the contents of one’s beliefs to the contents of the beliefs of an enhanced agent.
256 Rinard
7 It is Not Rational to Suspend Judgment on External World

Skepticism
So far, I have presented an argument, which could persuade an external world

skeptic, for the claim that believing external world skepticism is not rational.
The former skeptic may now come to believe that skepticism is false. After all,
consider their intellectual history. Before encountering the skeptical argument,
they had a typical collection of ordinary beliefs, including the belief that they
knew many things about the world. Then, upon hearing the skeptical argu-
ment, they were convinced by it, and gave up the belief that they had external
world knowledge. Once they accept the argument given here, they may simply
revert back to the position they were in before encountering the skeptical ar-
gument. They now see that accepting this argument was a mistake; it was not
rational for them to do so. A natural response would be to re-adopt the posi-
tion they would have maintained, had they not made that particular mistake.
However, some think that, rather than reverting to their original belief that
they know many things about the world, the former skeptic should now sus-
pend judgment on skepticism. They think the former skeptic should reason
as follows: I’ve seen that it’s not rational to accept the argument for external
world skepticism, because doing so commits one to an irrational combination
of beliefs. However, this doesn’t change the fact that the premises of the skep-
tical argument are highly compelling. They are so compelling that it couldn’t
possibly be rational to believe that one of them is false, so it couldn’t be ra-
tional to believe that external world skepticism is false. The only remaining
option is to suspend judgment on external world skepticism, so that is what
rationality requires.
According to this line of thought, the skeptic should suspend judgment on
skepticism while believing, on the basis of the argument just given, that it’s
rational to do so. I’ll call someone in this position a confident suspender. (Later
we’ll encounter an unconfident suspender, who suspends judgment on skepti-
cism while suspending judgment on whether it’s rational to do so.)
The position of the confident suspender may sound very reasonable. How-
ever, I will argue that it is not rational. It has a defect very similar to the one
that undermined the position of the original external world skeptic. First, note
that suspending judgment on external world skepticism commits one to sus-
pending judgment on other kinds of skepticism as well. Earlier I argued that
accepting external world skepticism commits one to also accepting skepticism
about the past and thereby skepticism about complex reasoning. Similarly,
suspending judgment on skepticism about the external world commits one to
suspending judgment on skepticism about the past and thereby skepticism
about complex reasoning. Let’s assume the confident suspender does so.
Now, however, we can begin to see where the problem lies. The confident
suspender believes a proposition P—the proposition that rationality requires
them to suspend judgment on external world skepticism—on the basis of the
argument sketched a few paragraphs back. This argument is complex. (It relies
on the claim that it’s not rational to believe external world skepticism, and the
argument for this (in Sections 2–5) is complex.) So the confident suspender
believes P on the basis of a complex argument, while suspending judgment
on skepticism about complex reasoning. That is, they believe P while suspend-
ing judgment on whether believing P is rational. In doing so, they violate a
plausible principle I call Belief Endorsement, which says, roughly, that rational
agents endorse their own beliefs, in the sense of believing them to be rational.
More precisely:
Belief Endorsement: Rationality prohibits combinations of attitudes of

the following kind: One believes P, but one takes
some doxastic attitude, other than belief, toward
the proposition that belief in P is rational.18
Note that Belief Endorsement does not say that, whatever beliefs one initially
happens to have, one must believe that they are rational: it may be that one has
some irrational beliefs, in which case Belief Endorsement allows that rational-
ity may require one to give up those irrational beliefs, rather than adopt the (in
these cases mistaken) belief that they are rational.
Belief Endorsement is highly plausible; it entails Anti-Denouncement (from
Section 4) and can be seen as a natural generalization of it. But, as noted above,
the confident suspender has a combination of attitudes that violate Belief
Endorsement. The confident suspender believes P—that rationality requires
suspension of judgment on external world skepticism—while suspending
judgment on whether it is rational to have that belief.
What about the position of the unconfident suspender? Both the confident
and unconfident suspender suspend judgment on external world skepticism,
and they both also suspend judgment on skepticism about the past and skepti-
cism about complex reasoning. The confident suspender got into trouble by
combining these attitudes with the belief that rationality requires suspend-
ing judgment on external world skepticism. The unconfident suspender seeks
to avoid this trouble by not believing this proposition. Instead, they suspend
judgment on it.
18 Here I am working within a framework in which there are three doxastic attitudes: belief,
disbelief, and suspension of judgment.
258 Rinard
Unfortunately the unconfident suspender thereby gets into a closely related

kind of trouble. They suspend judgment on external world skepticism while
suspending judgment on whether they ought to suspend judgment on external
world skepticism, thereby violating the following principle:
Endorsement: Rationality prohibits combinations of attitudes of the

following kind: One takes doxastic attitude D toward P,
but one takes some doxastic attitude, other than belief,
toward the proposition that taking D to P is rational.
Endorsement is the generalization of Belief Endorsement to doxastic attitudes

other than belief. In epistemology, as elsewhere, we should aim for simplic-
ity and elegance in our theorizing. The simplest theory will treat all doxastic
attitudes alike: if Endorsement is true for belief, it should be true for other
doxastic attitudes as well. If so, then the unconfident suspender fares no better
than the confident suspender.
There is another unfortunate feature of the unconfident suspender’s posi-
tion. Recall, first, that this position was adopted on the basis of complex consid-
erations of the sort described above. Their position of radical uncertainty was
not adopted out of the blue, for no reason whatsoever; rather, it was prompted
by seeing how skepticism is self-undermining. But, since they suspend judg-
ment on propositions about the past, and because these considerations are
complex, they know nothing of them now. They are unsure of many things,
but they have no idea why. Having adopted this position, they can no longer
see any reason for maintaining it. It is hard to see how such a position could
be rational.
Suppose, for example, that the unconfident suspender happens to catch
sight of their hands. They have a vivid experience as of a hand, and are inclined
to believe it. Since their doxastic position is so impoverished, they possess no
reason why they should not believe it.
Now, if a skeptic had this experience, they would have at the ready a com-
pelling argument for why it is that one shouldn’t believe that one has a hand,
namely their original argument for skepticism. And the confident suspender
would have at the ready an argument for why one should suspend judgment on
the proposition that one has a hand. But the unconfident suspender, since they
believe so little, has no doxastic resources with which to resist the inclination
to believe that they have a hand.
This section has argued that suspending judgment on external world skepti-
cism is not rational. The confident suspender is irrational because they violate
Belief Endorsement. The unconfident suspender violates a generalization of

Belief Endorsement, and their position has other unfortunate features.19
8 Rationality Requires Disbelieving Skepticism
I have now argued that believing skepticism is not rational (Sections 2–5) and
that suspending judgment on skepticism is not rational (the previous section).
What rationality requires, I claim, is that we believe that skepticism is false.
This claim would be mistaken if the skeptical problem were an instance of a
doxastic dilemma. A doxastic dilemma is a situation in which rationality pro-
hibits believing P, rationality prohibits disbelieving P, and rationality prohibits
suspending judgment on P. That is, a doxastic dilemma is a situation in which
there is no doxastic attitude one could rationally take towards P. My view is that
doxastic dilemmas are impossible. If so, then skepticism can’t be an instance of
one.
It is central to our concept of rationality that rationality constitutes an ideal
to which one could coherently aspire, and by which one could be guided. But
if doxastic dilemmas were possible, rationality could not play this role in those
cases. We cannot be guided by the voice of rationality if rationality tells us
to neither believe, nor disbelieve, nor suspend judgment on P. We could not
coherently aspire to conform to the requirements of rationality if there were
doxastic dilemmas. Since I regard its ability to play this guidance-giving role
as constitutive of rationality, I conclude that there are no doxastic dilemmas.
19 It has been suggested to me that Endorsement should be rejected because there could be
a rational agent who is so confused and uncertain that they suspend judgment on P and
also suspend judgment on whether it’s rational to suspend on P. However, in my view,
such an agent would not be rational. Consider the sorts of circumstances that could make
it rational to suspend judgment on P. Perhaps there are conflicting considerations, some
of which favor P, and some of which favor not-P. Or perhaps the agent hasn’t yet worked
out whether their evidence on balance favors P over not-P. Whatever circumstances make
it rational to suspend judgment on P, presumably a rational agent would be able to recog-
nize when they are in these circumstances, and recognize that such circumstances call for
suspending judgment on P. Of course, this need not prevent them from also recognizing
(when it’s true) that it may not be rational for other people to suspend judgment on P, or
that it may not be rational for them to suspend judgment on P in the future (for example,
they may expect that further deliberation will result in their coming to rationally believe
or disbelieve P). But whenever it is true that, in their present circumstance, rationality
requires them to suspend judgment on P, a rational agent will recognize this, and so will
not also suspend judgment on whether they ought to suspend on P.
260 Rinard
One might object that rationality could play its guidance-giving role even
if there were a doxastic dilemma. What rationality would tell us to do, in such
a case, is to take no doxastic attitude whatsoever toward the proposition in
question.
I will now argue that rationality cannot require that one take no attitude to-
ward some proposition. In order to do so, I’ll distinguish various ways it could
be the case that one takes no attitude toward some proposition P. One way is
to be cognitively deficient in a way that renders one unable to understand P.
One might lack one of the concepts contained in P, for example. But, I claim,
such cognitive deficiency could not be a requirement of rationality, which rep-
resents an ideal way that an agent might be.
A second way to take no attitude toward P is to be capable of understanding
P, but to have never, in fact, ever consciously entertained the question wheth-
er P. Again, it could not be a requirement of rationality that one inhabit so
unreflective a state. Having and pursuing curiosity, and engaging in inquiry
and reflection, are intellectual virtues, and, as such, could not be prohibited
by rationality.
At this point, some will think we have exhausted all possible ways of tak-
ing no attitude toward a proposition. Some hold that, once an agent considers
the question whether P, if she neither believes P nor disbelieves P, then she
suspends judgment on P. If so, then my argument for the claim that rationality
cannot require taking no attitude is complete.
However, some may have a more demanding conception of what it is to
suspend judgment on a proposition.20 Some may hold that one does not count
as suspending judgment on P unless one has concluded deliberation with a
settled neutrality toward P. On this view, one can take no attitude toward P
while one is in the process of deliberating on whether P; or, it may be that one
did deliberate on whether P in the past, but one’s deliberation was cut short
before coming to a proper conclusion, for example due to distraction, frustra-
tion, boredom, etc. Might rationality require that one inhabit a state of this
kind?
I do not think that rationality could require that one start deliberating on
whether P, but then stop due to some cause such as frustration, boredom, dis-
traction, etc. Some of these are intellectual vices. For example, stopping de-
liberation out of frustration or boredom is an intellectual vice, and not the
sort of thing that could be required by rationality. Stopping deliberation due
to distraction by some more urgent project need not be a vice; but it does not
20 See Friedman (2013) for a discussion of different conceptions of suspended judgment.

seem to be the sort of thing that rationality could require. Surely there could
be a perfectly rational agent with all the time in the word to sit and ponder.
The only remaining possibility, compatible with one’s failing to take a dox-
astic attitude toward P, is that one deliberate on whether P forever (or at least
until one ceases to exist). But, I claim, this could not be a requirement of ra-
tionality either. In this case my argument relies on the following principle: if
rationality requires one to φ, then it must be possible for one to φ while ra-
tionally believing that rationality requires φ-ing. I claim that it is not possible
for a rational agent to deliberate on P while believing that what rationality
requires is that one deliberate on P forever.
The reason has to do with the nature of deliberation. Deliberation is essen-
tially a goal-directed activity. One deliberates on whether P with the aim of
concluding this deliberation with some sort of stable attitude toward P (be it
belief, disbelief, or suspension of judgment). One is trying to figure out what
attitude to take toward P. But one can’t rationally engage in a goal-directed
activity while believing that achieving that goal is rationally forbidden. One
can’t rationally deliberate on P while believing that concluding this delibera-
tion is rationally forbidden. I conclude that rationality cannot require that one
deliberate on some proposition forever.
The upshot is that rationality cannot require that one take no doxastic at-
titude at all towards P, and so, there can be no doxastic dilemmas, compatible
with rationality’s uniformly playing its guidance-giving role.
In this section I have made a prima facie case for the claim that there are no
doxastic dilemmas. This entails that skepticism is not an instance of a doxastic
dilemma, which is a necessary condition for it to be the case that, as I claim,
rationality requires disbelieving skepticism. I do not claim that my prima facie
case against the possibility of doxastic dilemmas should convince everyone.
There is much more that could be said on this point, but it is beyond the scope
of this paper to enter further into the details of that debate. As before, the aims
of this paper will be satisfied if my argument can convince those skeptics who
agree that doxastic dilemmas are impossible.
9 Is the Anti-Skeptical Position Unstable?
Suppose, then, that on the basis of the arguments given here, the former
skeptic comes around to endorsing their original view that they know many
things about the external world. One might worry that their newfound anti-
skepticism is vulnerable to instability. Since the premises of the original skepti-
cal argument remain compelling, will they not be inclined to adopt them again,
262 Rinard
to once again become a skeptic, again notice the problem, again reject skepti-
cism….and again accept the skeptical premises, repeating the cycle forever?21
This outcome is not inevitable. Having gone through the loop once, the for-
mer skeptic may rationally maintain their non-skeptical position, even when
presented with the skeptical arguments they once endorsed.
This is because there is an important difference between a seasoned non-
skeptic—a former skeptic who once endorsed the argument for external
world skepticism, discovered the irrational position to which it lead, and con-
sequently re-adopted the anti-skeptical stance—and a naïve non-skeptic, who
has never encountered skeptical arguments. The naïve non-skeptic may enter
innocently into external world skepticism before seeing the incoherence that
awaits. But the seasoned non-skeptic has been down that road before. When
presented with the argument for external world skepticism, they see not just
the prima facie plausibility of its premises, but also the irrational position to
which those premises lead. With this outcome vividly before them, they can,
even while acknowledging the plausibility of the premises, nonetheless ratio-
nally decline to accept them.
10 Conclusion
I have argued, using only premises that an external world skeptic could accept,
that rationality requires us to believe that external world skepticism is false. At
several points I have taken a stand on issues that remain controversial, such as
the role of memory in complex reasoning; which arguments can be grasped,
in their entirety, all at once, by ordinary humans; the status of various level-
bridging principles (Anti-Denouncement, Belief Endorsement, Endorsement);
and the possibility of doxastic dilemmas. There is more to be said on these
and other points. In each case, however, a prima facie compelling case can be
made; and an external world skeptic would not be committed to the opposing
view just in virtue of their skepticism. Thus, the line of reasoning presented
here could rationally persuade an external world skeptic, who shares the views
endorsed here on these orthogonal issues, to give up their skepticism.
In contrast, many contemporary epistemologists regard the skeptic as a
hopeless case, and the attempt to reason with the skeptic as a lost cause. The
skeptic is portrayed as someone so far gone that there’s no point in trying to
save them now. The best we can do is try to prevent others, who are not yet
21 Discussions of loop effects in different, but related, contexts appear in Hume (1888: 187)
and Plantinga (1993, 1994).
skeptics, from succumbing to the same fate. Thus Williamson’s observation that
most responses to skepticism are “better at prevention than cure”22 and Byrne’s
remark that “the sceptic doesn’t need an argument; she needs treatment.”23
The upshot of this paper is that this view of the situation is misguided. We
need not regard the skeptic as someone who can’t be reasoned with. Each
premise of the argument given here could be accepted by an external world
skeptic. Once a skeptic accepts the conclusion of this argument—that ratio-
nality requires the belief that skepticism is false—they should then adopt that
belief. It is possible to reason one’s way out of skepticism.24
Bibliography
Bergmann, M. 2005. “Defeaters and Higher-Level Requirements,” The Philosophical

Brueckner, A, 1992. “Problems With the Wright Route to Scepticism,” Mind 101: 308–317.
Burge, T, 1993. “Content Preservation,” The Philosophical Review 102: 457–488.
Byrne, A. 2004. “How Hard are the Sceptical Paradoxes?” Noûs 38: 299–325.
Chisholm, R. 1977. Theory of Knowledge, 2nd Edition. Englewood Cliffs, N.J.: Prentice
Hall.
Christensen, D. 2010. “Rational Reflection,” Philosophical Perspectives 24: 121–140.
Descartes, R. 1996. Meditations on First Philosophy with Selections from the Objections
and Replies. Cambridge: Cambridge University Press.
Dretske, F. 1970. “Epistemic Operators,” Journal of Philosophy 67: 1007–1023.
Elga, A. 2013. “The Puzzle of the Unmarked Clock and the New Rational Reflection
Principle,” Philosophical Studies 164: 127–139.
Feldman, R. 2005. “Respecting the Evidence,” Philosophical Perspectives 19: 95–119.
Friedman, J. 2013. “Suspended Judgment,” Philosophical Studies 162: 165–181.
Fumerton, R. 1995. Metaepistemology and Skepticism. Rowman and Littlefield.
Greco, D. 2014. “A Puzzle about Epistemic Akrasia,” Philosophical Studies 167: 201–219.
Horowitz, S. 2014. “Epistemic Akrasia,” Noûs 48: 718–744.
22 Williamson (2000, 27).

23 Byrne (2004, 301).
24 Many thanks to Selim Berker, Alex Byrne, Nick DiBella, Adam Elga, Andrew Graham,
David Gray, Stephen Maitzen, Julia Markovits, Kevin McCain, Ram Neta, Robert Pasnau,
Ted Poston, Agustin Rayo, Alison Simmons, Amia Srinivasan, Robert Stalnaker, Jonathan
Vogel, Timothy Williamson, Alex Worsnip, Stephen Yablo, and audiences at Oxford Uni-
versity, the University of Missouri – Columbia, and the Ranch Metaphysics Conference.
Special thanks to Miriam Schoenfield and Roger White for extensive comments on mul-
tiple drafts.
264 Rinard
Hume, D. 1888. A Treatise of Human Nature. Oxford: Clarendon Press.

Lasonen-Aarnio, M. 2014. “Higher-Order Evidence and the Limits of Defeat,” Philoso-
phy and Phenomenological Research 88: 314–345.
McCain, K. 2012. “A Predictivist Argument Against Scepticism,” Analysis 72: 660–665.
Moore, G.E. 1962a. “A Defense of Common Sense,” In G.E. Moore (ed.), Philosophical
Papers, 32–59. New York: Collier Books.
Moore, G.E. 1962b. “Proof of an External World.” In G.E. Moore (ed.), Philosophical Pa-
pers, 127–150. New York: Collier Books.
Moore, G.E. 1962c. “Four Forms of Skepticism.” In G.E. Moore (ed.), Philosophical Pa-
pers, 196–225. New York: Collier Books.
Nozick, R. 1981. Philosophical Explanations. Cambridge: Cambridge University Press.
Pasnau, R. 2014. Isaiah Berlin Lectures, Lecture 5.
Plantinga, A. 1982. “How to be an Anti-Realist,” Proceedings and Addresses of the Ameri-
can Philosophical Association 56: 47–70.
Plantinga, A. 1993. Warrant and Proper Function. Oxford: Oxford University Press.
Plantinga, A. 1994. “Naturalism Defeated,” unpublished manuscript.
Poston, T. 2016. “Acquaintance and Skepticism about the Past.” In B. Coppenger & M.
Bergmann (eds.), Intellectual Assurance: Essays on Traditional Epistemic Internal-
ism, 183–201. Oxford University Press.
Pritchard, D. 2001. “Scepticism and Dreaming,” Philosophia: Philosophical Quarterly of
Israel 28: 373–390.
Pryor, J. 2000. “The Skeptic and the Dogmatist,” Noûs 34: 517–549.
Rinard, S. 2013. “Why Philosophy Can Overturn Common Sense.” In T. Gendler & J.
Hawthorne (eds.), Oxford Studies in Epistemology, 185–213. Oxford: Oxford Univer-
sity Press.
Russell, B. 1921. The Analysis of Mind. Public Domain.
Titelbaum, M. 2015. “Rationality’s Fixed Point.” In T. Gendler & J. Hawthorne (eds.),
Oxford Studies in Epistemology, 253–294. Oxford: Oxford Univeristy Press.
Tymoczko, T. & J. Vogel 1992. “The Exorcist’s Nightmare,” Mind 101: 543–552.
Weatherson, B. (unpublished manuscript). “Do Judgments Screen Evidence?” available
online at [http://brian.weatherson.org/JSE.pdf].
Williamson, T. 2000. Knowledge and Its Limits. Oxford University Press.
Williamson, T. 2011. “Improbable Knowing.” In T. Dougherty (ed.), Evidentialism and its
Discontents, 147–164. Oxford University Press.
Wilson, J. 2012. “The Regress Argument against Cartesian Skepticism,” Analysis 72:
668–673.
Wright, C. 1991. “Scepticism and Dreaming: Imploding the Demon,” Mind 100: 87–116.
Index
Apraxia 65–67, 71–72 Fallibilism 2, 74, 106–115, 117–121, 123–124,

Anti-Denouncement principle 249–250, 147–150
257, 262 Forgotten evidence 164–165, 170–171
Autonomy 1, 22, 63
Generativism 162–163, 170–172
Belief Argument 29 Gettier problem 2, 129, 144–152, 155, 180–181
Belief-dependent processes 181–183 Good Samaritan experiments 217–218,
Belief Endorsement principle 257–259, 262 222–223
Belief-independent processes 182–183
Bootstrapping 11 Independence principle 84–86, 88
Bottleneck Argument 35–40 Infallibilism 107–108, 110–111, 124
Internalism 2, 17, 161, 166–174, 176, 180,
Causal Constraint 34 182–184, 236
Causal Inference (inf) 28–31, 33–35 Instrumental knowledge 24, 32–34, 38–43
Closure 35, 37–38, 40, 65, 128, 137–138,
140–141, 229, 243–244 Justification 2, 7–15, 19, 21–22, 26, 28–30, 33,
Commonsensical claims 187–189, 193, 195, 47, 49, 51–53, 58, 74–75, 85–93, 96, 99,
197–199 101–102, 105–108, 113, 119, 128–136, 140,
Criteria of truth 143–152, 155, 158–159, 161–174, 176–177,
Detection 8–11 179–186, 204, 227–228, 231, 234, 240,
Inference 9–11 243–248
Paradigmatic 8–11, 13–14, 19–21
Knowledge 1–3, 7, 12–13, 17, 24–25, 27,
Deceiver Argument 24, 26, 28–30 29–43, 47–53, 56, 58–60, 64–67, 71,
Deference principle 252–255 74–76, 94, 99–100, 102, 105–124, 128–136,
Disagreement 138, 140–141, 143–152, 155–160, 161–163,
Peer 2, 60, 83–85 165, 167, 174, 180–182, 184–185, 204–205,
Religious 2, 83, 87–91, 100–101 213–214, 224, 227, 230, 232–233, 236,
Dispositional belief 171, 177–178, 181, 186 240, 243–244, 247, 256
Disposition to believe 10–11, 16, 177–185, 205,
205, 216 Logically opaque 188, 190, 193
Eliminativism 28–30, 34, 49 Memory 2, 128, 134, 140, 161–174, 236, 241,
End of Inquiry (ei) 115 245–248, 262
Endorsement principle 258–259, 262 Moorean dogmatism 24, 26, 35, 38–43,
Epistemic conservatism 232, 234–236 64–65, 71, 76, 195–198, 227, 234–236, 242
Epistemic Priority of Experience (epe)
24–26, 28–31, 34–35, 43 New evil demon problem 167, 183
Equal Weight View 1–2, 83–85, 87–102
Equipollence 30–32, 34–35, 37, 67–68 Perception Argument 31–32, 34–35, 37
Eudaimonia 209–210 Phenomenal Conservatism 2, 169, 172
Evidentialism 2, 30, 55, 143, 164, 168–169, 171 Pragmatism 2, 105–106, 110–112, 114,
Explanationism 120, 124
First Philosophy (Ex-1st) 230–232, Pragmatist Theory of Truth 106, 111–118, 121,
234–236 123–124
Second Philosophy (Ex-2nd) 232–236 Preservationism 162–163, 165, 170–172
266 Index
Problem of the Criterion 1, 7–23, 234 133–134, 136–137, 140–141, 143–144,

147–149, 151, 159, 196, 204–205, 232–233,
Rationality 55, 67, 71, 83, 87, 94–99, 118, 196, 236, 240–246, 248, 250–252, 256–258,
240–241, 252, 255–263 261–263
Regress 7–8, 46–49, 51, 109, 184 Extraordinary 2, 128–160
Reliabilism 182–184, 203–206, 213, 216 Moorean response See Moorean
Reliability 8, 10–11, 14, 16–17, 19, 21, 146, 196, dogmatism
205–206, 215 Ordinary 128–129
Representative Theory of Perception Pragmatist response See Pragmatism
(rtp) 26–31, 33–35, 43 Pyrrhonian 1, 67–69, 105
Sage Trickle Down Argument 1, 47, 49–53, 56–60

Skeptical 71–73, 76–78 Trivialism 56–60
Stoic 66, 72, 214
Sensitivity 17, 36–37, 40–43, 212, 224 Underdetermination 24–25, 229, 243
Situationism 203, 217–219, 222, 224
Skepticism Virtue
About complex reasoning 3, 241, 245, Explanatory 230, 232–234, 236
247–252, 254, 256–257, 262 Intellectual 2, 203–206, 210, 212–219,
About the past 162, 164, 229, 241–252, 224, 260
256–258 Moral 69, 203, 206, 210, 212, 215–217
About virtue 2, 205, 213, 217–224 Skill model of 2, 203, 206–216
Academic 1, 65–67, 71–77 Reliabilism 203–206, 213–214, 216–217,
Cartesian See Skepticism: External world 224
Explanationist response 3, 227–236 Responsibilism 203, 206–216, 224
External world 1, 3, 24–43, 64, 128–129,

McCain e Poston (2018) - The Mystery of Skepticism

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

McCain e Poston (2018) - The Mystery of Skepticism

Uploaded by

Copyright:

Available Formats

The Mystery of Skepticism

Brill Studies in Skepticism

Diego Machuca (conicet)

John Greco (Saint Louis University)

The titles published in this series are listed at brill.com/bss

Copyright 2019 by Koninklijke Brill nv, Leiden, The Netherlands.

This book is printed on acid-free paper and produced in a sustainable manner.

1 The Mystery of Skepticism: an Introduction 1

2 Procedural Reasons and the Problem of the Criterion 7

3 Is Cartesian Skepticism Too Cartesian? 24

4 Hume’s Certain Doubts: Why We Should Worry Too 46

5 Skepticism as a Way of Life 63

6 Disagreement Skepticism and the Rationality of Religious Belief 83

7 Pragmatist Anti-Skepticism: at What Price? 105

8 Extraordinary Skepticism 128

9 Internalism, Memory, and Skepticism 161

10 Inferential Internalism and the Problem of Unconscious

11 The Philosopher’s Doom: Unreliable at Truth or Unreliable at

12 The Skill Model: a Response to Skepticism about the Intellectual

13 Two Strategies for Explaining Away Skepticism 227

14 Reasoning One’s Way Out of Skepticism 240

Scott Aikin—Vanderbilt University

The Mystery of Skepticism: an Introduction

© koninklijke brill nv, leiden, ���9 | doi:10.1163/9789004393530_002

significance of peer disagreement to religious disagreements (Chapter 6). Ac-

Poston’s exploration of explanationist responses to skepticism in Chapter 13.

1 Thanks to Siddharth Srikakolapu for editorial assistance in formatting this volume.

Procedural Reasons and the Problem

(1a) A proposition P can be justified for a person S only if there is a criterion

© koninklijke brill nv, leiden, ���9 | doi:10.1163/9789004393530_003

(4b) ∴ No criterion can be justified.1

2.1 Paradigmatic Criteria of Truth

(DC) Necessarily, for all propositions P, if P has X, then P is true.

Inference criteria are paradigmatic criteria according to which standing in a

(IC) Necessarily, for all P and Q, if P is true and P stands in R to Q, then

Let us say that a proposition P is directly or indirectly connected by relation R

(HC) Necessarily, for all P, if P has X or there is a Q such that Q has X

(CD) Necessarily, for all propositions P, if I clearly and distinctly per-

An example of an inference criterion is this principle that would underwrite

(SS) If x% of observed Ss have been Ps under the relevant conditions,

(1aP)  proposition P can be justified for a person S only if there is a

2.2 Reliability Claims

intuitions, memories, and so on. We have, for example, dispositions to form

(MR)  ecessarily, for all P, if the belief that P is activated by procedure

(1aR) A proposition P can be justified for a person S only if there is a

An aspect of this version of the problem is whether having justified beliefs

2.3 Epistemic Principles

(E) Necessarily, for all P, if _____, then P has Φ.8

There are at least two kinds of epistemic principles. Generation principles

(G) For all P, if P has X, then P has Φ.

This principle of Chisholm’s is a generation principle:

(M1) If the property of being F is self-presenting, if S is F, and if S be-

Another generation principle is this proposition about the connection be-

(CJ) If P coheres with the system of propositions accepted by S, then P

Transmission principles state that propositions have a normative epistemic

An example of a transmission principle is this proposition to the effect that

(ΦT) For all P, if P has Φ, then P is (probably) true

jointly entail a paradigmatic criterion:

(RT) For all P, if P has X, then P is (probably) true.

Likewise, any set of propositions that includes both a generation principle

For all P, if P has X or there is a Q such that Q has X and Q is directly or

Given (ΦT)—propositions that have Φ are (probably) true—a hybrid paradig-

(HC) F or all P, if P has X or there is a Q such that Q has Χ and Q is di-

(1aE) A proposition P can be justified for person S only if there is an

© koninklijke brill nv, leiden, ��9 | doi:10.1163/9789004393530_002

© koninklijke brill nv, leiden, ��9 | doi:10.1163/9789004393530_003

(IC) Necessarily, for all P and Q, if P is true and P stands in R to Q, then

(HC) Necessarily, for all P, if P has X or there is a Q such that Q has X

(CD) Necessarily, for all propositions P, if I clearly and distinctly per-

(SS) If x% of observed Ss have been Ps under the relevant conditions,

(1aP) proposition P can be justified for a person S only if there is a

(MR) ecessarily, for all P, if the belief that P is activated by procedure

(1aR) A proposition P can be justified for a person S only if there is a

(M1) If the property of being F is self-presenting, if S is F, and if S be-

(CJ) If P coheres with the system of propositions accepted by S, then P

(HC) F or all P, if P has X or there is a Q such that Q has Χ and Q is di-

(1aE) A proposition P can be justified for person S only if there is an

(PR1AA) I f it were false that (AA) I am an above-average employee, then

(PR2AA) I f I were to believe that (AA) I am an above-average employ-

significance of different types of epistemic justification—the value of epis-

© koninklijke brill nv, leiden, ��9 | doi:10.1163/9789004393530_004