Testimony, Trust, & Authority

Title Pages
University Press Scholarship Online
Oxford Scholarship Online
Testimony, Trust, and Authority

Benjamin McMyler
Print publication date: 2011

Print ISBN-13: 9780199794331
Published to Oxford Scholarship Online: September 2011
DOI: 10.1093/acprof:oso/9780199794331.001.0001
Title Pages
(p.i) Testimony, Trust, and Authority (p.ii)
(p.iii) Testimony, Trust, and Authority
(p.1) Testimony, Trust, and Authority (p.2)
(p.iv)
Oxford University Press, Inc., publishes works that further

Oxford University's objective of excellence
in research, scholarship, and education.
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France
Greece
Guatemala Hungary Italy Japan Poland Portugal
Singapore
Title Pages
South Korea Switzerland Thailand Turkey Ukraine
Vietnam
Copyright © 2011 Oxford University Press
Published by Oxford University Press, Inc.

198 Madison Avenue, New York, New York 10016
www.oup.com
Oxford is a registered trademark of Oxford University

Press.
All rights reserved. No part of this publication may be

reproduced,
stored in a retrieval system, or transmitted, in any form or
by any means,
electronic, mechanical, photocopying, recording, or
otherwise,
without the prior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data

McMyler, Benjamin.
Testimony, trust, and authority / Benjamin McMyler.
p. cm.
Includes bibliographical (p. ) references.
ISBN 978-0-19-979433-1 (alk. paper)
1. Testimony (Theory of knowledge) I. Title.
BD238. T47M36 2011
121'.3—dc22 2010052009
1 3 5 7 9 8 6 4 2
Printed in the United States of America
on acid-free paper
Introduction
Introduction
Benjamin McMyler
Why, if I accept what you say, on the basis of your saying it, do I
respond by saying “I believe you,” not “I believe what you say”? I
would like to say that the home of belief lies in my relation to
others.
(Cavell 1979: 391)
In the narrowest terms, this book is about the epistemology of

testimony, about the branch of the theory of knowledge concerned with
how we acquire knowledge and justified belief from the say-so of other
people. By all accounts, a great deal of what we know and believe is in
fact acquired in this way. Most of what we know about history, science,
and current events is acquired from the spoken and written word, from
being told things by people we trust and treat as authorities on these
matters. For many epistemologists, the sheer volume of knowledge and
belief acquired from the word of others is enough to make the topic of
testimony one of serious and legitimate epistemological concern. Given
that so much of what we know is actually acquired in this way, our
general epistemological theories about the nature of knowledge and
justification ought to have something to say about the kind of
knowledge and justification acquired from the word of others.
Introduction
The project of this book, however, is to demonstrate that the topic of

the epistemology of testimony is of much broader and deeper
philosophical significance than that of a mere subject matter to which
general epistemological theories should be applied. Suitably thought
through, the epistemology of knowledge and belief based on testimony
helps to reveal one of the ways in which the human mind is a
constitutively social phenomenon, one of the ways in which being the
kind of minded being that we (p.4) are involves participating in
interpersonal, social relationships. In the Anglo-American philosophical
tradition, interest in the role of sociality in the constitution of mind has
been largely confined to issues having to do with representational
content, with the ways in which a subject's social environment plays a
role in determining the content of her representational mental states.
This is not my concern in this book.1 Instead, I am concerned with the
way in which social relations play a justificatory role in the processes of
belief formation and knowledge acquisition. I am interested in how
knowledge and belief can be justified in virtue of relations of authority
and responsibility between persons and with what this reveals about
the nature of the human mind.
The thesis of this book is that the category of other persons plays a
distinctive and irreducible role in cognition. What other people tell us
plays a role in the processes of belief formation and knowledge
acquisition that is fundamentally unlike—that cannot be reduced to or
modeled on—the role played by other kinds of impersonal evidence. We
bear relations of epistemic dependence to others that are irreducibly
interpersonal in nature, relations that we do not bear to nonpersons,
and these relations have a distinctive kind of significance for us. We
could not give them up and still be the kind of minded beings that we
are. Such epistemic dependence is thus an important feature of what it
is to be a human being.
This is an extremely unorthodox philosophical thesis, but it is somewhat

surprising that it is so. In presenting and discussing these issues with a
variety of audiences, I have found that it often takes little effort to
convince nonphilosophers (and even many philosophers who are not
epistemologists) that there is something distinctive and sui generis
about the way in which knowledge and justified belief is acquired from
the testimony of others. Unlike knowledge based on other epistemic
sources such as perception, memory, and inference, knowledge based
on testimony involves “taking another's word for things,” “taking things
on the authority of another,” or “trusting another for the truth.” Indeed,
philosophical discussions of the epistemology of testimony are
themselves littered with references to trust and authority. For the vast
Introduction
majority of epistemologists, however, such terms are (and indeed must

be) mere placeholders to be replaced in the end by epistemological
concepts that are deemed vastly more palatable, concepts like that of
inductive evidence or reliable causal belief-producing processes. The
idea that there might be something sui generis about the epistemology
of testimony having to do with the way in which it is connected to the
notions of trust and authority is thus roundly dismissed as not only false
but verging on the nonsensical. We have a pretty clear idea of what
concepts are ultimately epistemologically palatable, so the thought
goes, and concepts like trust and authority are (p.5) not among them.
They may be useful in getting a relevant epistemological phenomenon
on the table, but in the end they must surely be replaced by something
more familiar.
In this way, dominant ideas about the general nature of knowledge and
justification have had a powerful influence on the way in which
philosophers have thought about the epistemology of testimony.
Accounts of the epistemology of testimony have generally been held
subservient to more general accounts of the nature of knowledge and
justification. There is nothing wrong with this in principle. The problem
is simply that this has resulted in mistaken accounts of the
epistemology of testimony. In this book I argue that the dominant
theories of the epistemology of testimony have significant difficulty
accounting for the relevant epistemic phenomena concerning the way
in which knowledge and justified belief is acquired from the testimony
of others. In essence, this is because the relevant epistemic phenomena
—for example, the way in which an audience is entitled to defer
epistemic challenges to its beliefs based on testimony—don't fit
squarely into standard epistemological categories. I articulate and
defend an alternative theory that does a much better job of accounting
for the relevant phenomena, but this alternative account does so only
by construing testimony as a genuinely sui generis epistemic source.
Testimony is a sui generis epistemic source in that explaining the
epistemology of testimony requires appealing to concepts and
principles unique to testimony. The account that I offer thus makes an
irreducible appeal to notions of trust and authority. On the account
developed here, the epistemology of testimony is extremely unorthodox,
but this is precisely why it is important and interesting. It is important
and interesting because it forces us to rethink the kinds of
considerations that are relevant to the acquisition of knowledge and
justified belief.
Introduction
I have tried in this book to focus as much as possible on the actual

phenomenon of testimonial knowledge and belief itself, leaving to one
side broader epistemological questions concerning the general nature
of knowledge and justification. I am here concerned with giving the
best possible explanation of our ordinary epistemic practices with
respect to the acquisition and dissemination of testimonial knowledge
and belief, leaving to one side the question of how all of this fits into
more general epistemological theorizing. There is thus a sense in which
I am here more concerned with testimony than I am with epistemology.
Nevertheless, over the course of this book I hope to demonstrate that
such unorthodox notions as trust and authority are deserving of serious
epistemological treatment. Even though references to trust and
authority are quite common in epistemological discussions of testimony,
epistemologists seldom pay serious attention to the extant philosophical
literatures concerning these concepts.2 These concepts have been of
sustained interest to moral, social, and (p.6) political philosophers, and
it is a virtue of the account of testimony presented in this book that it
shows how the epistemology of testimony connects to these broader
philosophical issues. At the very least, I hope here to demonstrate that
philosophers interested in the epistemology of testimony ought to pay
serious attention to the nature of trust and authority.
In a sense, then, I aim here to rehabilitate the intuitive idea that

testimonial knowledge and belief is distinctively based on “taking
another's word for things,” “taking things on another's authority,” or
“trusting another for the truth.” Beyond this, however, I want to provide
a diagnosis as to why, from a philosophical point of view, this can seem
so difficult to accept. I trace this difficulty to an idea that I call
epistemic autonomy: the idea that fully rational cognitive agents are
always solely epistemically responsible for the justification of their own
beliefs. In chapter 1, I argue that an extremely important though
largely unrecognized shift occurred in the Early Enlightenment period
concerning philosophical conceptions of testimony and testimonial
knowledge. Whereas prior to the Enlightenment testimonial knowledge
or belief was often taken to be the result of a cognitive capacity
distinctively connected to authority, figures like John Locke and David
Hume began to portray testimony as a kind of ordinary inductive
evidence, thereby severing the traditional connection between
testimony and authority. This shift in the way in which testimony was
conceived was a straightforward application to the epistemic realm of
broader Enlightenment suspicions concerning the place of authority in
political and religious affairs, but it is one that is seldom recognized. It
amounts to a substantive claim about the nature of theoretical
Introduction
rationality, that fully rational cognitive agents are epistemically

autonomous.
The standard contemporary theories concerning the epistemology of

testimony are typically cashed out in such a way as to simply assume
epistemic autonomy. The traditional debate about the epistemology of
testimony is typically cast as a debate between reductionism about
testimony and anti-reductionism about testimony. Reductionists about
testimony model the epistemology of testimony on the epistemology of
inductive inference. Reductionists hold that an audience's testimonial
belief is justified by the strength of an inference from a speaker's
testifying that p, through independently available considerations
concerning the speaker's trustworthiness, to the conclusion that p. Anti-
reductionists about testimony reject the idea that testimonial
justification is inferential and instead tend to model the epistemology of
testimony on the epistemology of perception. Anti-reductionists
typically hold that a speaker's testimony that p provides an audience
with a prima facie reason for believing that p analogous to the prima
facie reason for belief provided by perceptual representation.
Importantly, both knowledge based on inference and knowledge based
on perception are forms of firsthand knowledge. Both inferring that p
and perceiving that p involve a subject's coming to her own conclusion
about things. Intuitively, (p.7) however, knowledge and belief based on
testimony does not involve a subject's coming to her own conclusion
about things. Knowledge based on testimony is a form of secondhand
knowledge, meaning it is knowledge that is epistemically mediated by
the mind of a speaker in a way that knowledge based on ordinary
inference or perception is not. Standard reductionist and anti-
reductionist theories of testimony thus have difficulty making sense of
the way in which testimonial knowledge is secondhand in virtue of
being distinctively mediated by another mind.
In chapter 2, I provide an alternative, essentially interpersonal account

of the epistemology of testimony that can make sense of the way in
which testimonial knowledge is so mediated. This account does so,
however, only by rejecting epistemic autonomy. According to the
account of the epistemology of testimony developed here, learning from
testimony is a fundamentally social epistemic capacity, a capacity the
exercise of which is a cooperative undertaking between speaker and
audience. This is demonstrated by the fact that, in acquiring knowledge
and justified belief on the basis of testimony, an audience is entitled to
defer epistemic challenges to her testimonial beliefs back to the
original testifier. If an audience comes to believe that p on the basis of a
speaker's testimony, and if a third-party challenges the audience's
Introduction
belief by producing evidence that tells against p, then the audience is

entitled to defer the challenge back to the original speaker. The
audience can fulfill her epistemic burden with respect to meeting the
challenge by deferring to the testimonial speaker. This marks out
testimonial belief as epistemologically distinctive. Only testimonial
belief, belief justified by the authority of a speaker, admits of the
deferral of challenges. And the fact that testimonial belief admits of
deferral shows that testimonial believers are not solely epistemically
responsible for the justification of their own beliefs. Epistemic
autonomy is therefore mistaken.
Importantly, to reject epistemic autonomy is not to endorse gullibility or

blind obedience to authority. We shouldn't trust just anyone about just
anything, and hence we must be careful to assess the trustworthiness
of purported theoretical authorities. Nevertheless, when we do judge
that a speaker is trustworthy, and when we proceed to believe her
testimony on this basis, we are not then epistemically autonomous. We
are not then solely responsible for the justification of our testimonial
belief, and this is shown by the fact that we are entitled to defer certain
challenges to our testimonial belief back to the original testifier.
Unfortunately, giving up on epistemic autonomy might not seem like a

viable option. Epistemic autonomy is such an entrenched assumption
about the nature of theoretical rationality that it can be difficult to see
what exactly it would mean to give it up. In the final three chapters I
therefore work from a variety of directions to try to lessen its initial
appeal. In chapter 3, I attempt to further articulate the interpersonal
account of the epistemology of testimony developed in chapter 2 by
placing this account in the (p.8) context of a developing debate
concerning the epistemology of testimony that is decidedly different
from the traditional debate between reductionism and anti-
reductionism. This new debate concerns what we might call epistemic
dependence, and placing my interpersonal account in the context of
this new debate helps to highlight the way in which this account is
important and distinctive. According to my interpersonal account,
testimonial knowledge and belief is epistemically dependent on the
metaphysical category of other persons (on second persons, in
particular) and not merely on the speech acts or the beliefs of others.
Testimonial knowledge and belief is epistemically dependent on the
second-personal relations that we bear towards others—it involves
trusting a speaker for the truth—and this makes it unlike any other
form of knowledge and belief.
Introduction
In chapters 4 and 5, I consider and respond to two very general reasons

for thinking that this interpersonal account of the epistemology of
testimony cannot be correct. The first involves the account's appeal to
the attitude of trust. According to the interpersonal account of
testimony developed in chapters 2 and 3, testimonial knowledge and
belief distinctively involve trusting another person. However, one might
think that interpersonal trust relations are a matter for ethics, not for
epistemology.3 Such a thought is in fact encouraged by much
philosophical thinking about the nature of trust. Philosophers often
take the normative constraints on trust to be very different from the
normative constraints on belief, and so they often hold that there is an
inherent tension between trust and theoretical rationality. In chapter 4,
I argue that this is not the case. The account of testimony developed in
chapters 2 and 3 helps us to see that trust can itself be construed as a
species of belief and so as fully consistent with theoretical rationality.
Trust is nevertheless very different from other forms of belief in that it
involves beliefs that are epistemically supported by a particular kind of
reason, what I call a second-personal reason. This can explain much of
what motivates philosophers to argue that trust is very different from
ordinary belief, but it does so without having to claim that there is an
inherent tension between trust and theoretical rationality.
Finally, in chapter 5, I examine the notion of a second-personal reason

itself. One might object to the interpersonal account of testimony
developed in chapters 2 and 3 and to the account of trust offered in
chapter 4 by arguing, in effect, that there are no genuinely second-
personal reasons for belief, that the relations of authority and
responsibility between persons appealed to by my account of
testimonial belief in particular and trust-based belief more generally
are simply the wrong kind of thing to (p.9) play an irreducible role in
epistemically justifying belief. Along these lines, one might accept that
relations of authority and responsibility between persons can play a
genuine role in practical rationality, that there can be genuinely second-
personal reasons for action, but nevertheless deny that such relations
play an analogous role in theoretical rationality. Something like this line
of thought can be found in Stephen Darwall's recent work on the
second person. Darwall (2006b) makes a compelling case for the
existence of distinctively second-personal reasons for action, but in so
doing he explicitly denies that there are any genuinely second-personal
reasons for belief. In chapter 5, I argue that there is just as much
reason to think that second-personal considerations play an irreducible
role in theoretical rationality as there is to think that they play an
irreducible role in practical rationality and that Darwall's reasons for
thinking the contrary don't stand up to scrutiny. There is thus good
Introduction
reason to think that the rational significance of the second-person

actually spans whatever divide there may be between theoretical and
practical reason.
The upshot of this book is that human rationality is constitutively

dependent on social relations, on relations of authority and
responsibility that we bear towards others. Emphatically, this is not to
give up on the significance of personal autonomy. Rather, it is to force a
more nuanced appreciation of what such autonomy consists in for
essentially social creatures like us. Genuine personal autonomy does
not consist in our always being solely rationally responsible for the
justification of our beliefs and actions. It does not involve our always
coming to our own conclusion about what to think and do. The simple
fact is that we do not always come to our own conclusion about what to
think and do, and this is born out by our ordinary practice of deferring
challenges to beliefs and actions justified by the directives of other
persons in positions of theoretical and practical authority. Relations of
trust and authority between persons thus play a fundamental and
irreducible role in our lives as rational beings, and there is nothing
lamentable or unfortunate about this. Even though epistemic autonomy
is probably one of the most cherished ideals of the Enlightenment, it
stands in the way of appreciating how genuinely autonomous cognitive
agents are often, nevertheless, rationally dependent on social relations.
The issues broached in this book are deep and difficult, and I have done
little more than scratch the surface of what ultimately ought to be said
about them. I hope that what I say here proves at least somewhat
illuminating and instructive, but most of all, I hope that it encourages
others to do better.
Notes:
(1) Goldberg (2007) uses social externalist (or anti-individualist)
considerations concerning representational content to argue for anti-
individualist conclusions concerning the epistemology of testimony.
(2) Notable exceptions include Faulkner (2007a) and (2007b), Keren

(2007), and Zagzebski (manuscript).
(3) Alternatively, one might think that to the extent that interpersonal
trust is relevant to epistemology this is because interpersonal trust
amounts to something like a bet made on the basis of the consideration
of probabilities. I argue that this is mistaken as well.
Introduction
Testimony as a Philosophical Problem

Benjamin McMyler
DOI:10.1093/acprof:oso/9780199794331.003.0002
Abstract and Keywords

This chapter argues that an important though largely unrecognized
shift occurred in the Early Enlightenment period concerning
philosophical conceptions of testimony and testimonial knowledge.
Whereas prior to the Enlightenment testimonial knowledge or belief
was often taken to be the result of a cognitive capacity distinctively
connected to theoretical authority, figures like John Locke and David
Hume began to portray testimony as a kind of ordinary inductive
evidence, thereby severing the traditional connection between
testimony and authority. This shift in the way in which testimony was
conceived was a straightforward application to the epistemic realm of
broader Enlightenment suspicions concerning the place of authority in
political and religious affairs, but it is one that is seldom recognized. It
amounts to a substantive claim about the nature of theoretical
rationality, that fully rational cognitive agents are epistemically
autonomous.
Keywords: knowledge, understanding, authority, testimony, faith, autonomy, evidence,

epistemic responsibility
Here are a few things that I know. I know that the copperhead is the
most common venomous snake in the greater Houston area. I know that
Napoleon lost the Battle of Waterloo. I know that, as I write, the

average price for gasoline in the US is $4.10 per gallon. And I know
that my parents recently returned home from a trip to Canada.
All of these things I know on the basis of what epistemologists call

testimony, on the basis of being told of them by another person or
group of persons. I know that the copperhead is the most common
venomous snake in the greater Houston area because I recently read
this in a guide to Houston area snakes. I know that Napoleon lost the
Battle of Waterloo because at some point I learned about this in school.
I know that the national average price for gasoline is $4.10 per gallon
because I just saw a report about it on the evening news. And I know
that my parents recently returned home from their trip to Canada
because I just talked to my mother on the phone.
As epistemologists use the term, ‘testimony’ refers to something much

broader than what goes on in a court of law or a governmental hearing.
None of the instances of testimonial knowledge listed above required a
speaker to do anything like swear an oath, and if it turns out that one of
the above speakers has lied, she cannot be brought up on charges of
perjury. Additionally, the term ‘testimony’ refers to something narrower
than mere statements or even, arguably, mere assertions. All of the
instances of testimonial knowledge listed above were acquired from a
speaker's not only making a statement but making a statement
explicitly intended to communicate information to (p.11) an audience.
They were all acquired from a speaker's explicitly telling something to
an audience. In this sense, we can say that, as epistemologists typically
use the term, ‘testimony’ refers to all those cases in which a speaker (or
a writer) tells something to an audience.1 The epistemologically
relevant category of testimony is the category of ordinary tellings.2
Testimony is without a doubt one very common source of knowledge,

but it is certainly not the only one. All of the facts listed above I might
have come to know in ways other than through testimony. If I had been
visiting my parent's house, I might have come to know that my parents
had returned home from their trip by perceiving their return first-hand.
If I had personally participated in the Battle of Waterloo, I might know
that Napoleon lost the battle on the basis of my memory of my own
experience there. And if I was a practicing herpetologist, I might have
come to know that the copperhead is the most common venomous
snake in the greater Houston area on the basis of calculations from my
own field research. Perception, memory, and inference are three other
very common sources of knowledge.
More strictly, perception, memory, and inference are three cognitive

capacities, powers, or abilities, the exercise of which is capable of
producing knowledge. One might think that a source of knowledge is
something like material from which knowledge can be gained or
derived. Testimony would appear to be a source of knowledge in this
sense. Perception, memory, and inference, however, are not materials
from which knowledge can be gained or derived. Instead, they are
capacities for gaining or deriving knowledge. Perception, memory, and
inference are cognitive capacities, the exercise of which is capable of
yielding states of belief that amount to knowledge. We might therefore
call them epistemic capacities. Though testimony (or ordinary telling) is
itself material from which knowledge can be gained or derived, learning
from testimony is certainly an epistemic capacity on a par with
perception, memory, and inference. Learning from testimony is a
cognitive capacity the exercise of which is capable of yielding states of
belief that amount to knowledge.
There is one important respect, however, in which the capacity for

learning from testimony is very different from the capacities of
perception, memory, and inference. Learning from testimony is a
fundamentally social capacity; it is what the Scottish (p.12)
Enlightenment philosopher Thomas Reid refers to as “a social operation

of mind” (2002: 68).3 Learning from testimony is an epistemic capacity
the exercise of which necessarily takes two—a speaker and an
audience. In contrast, perception, memory, and inference are
fundamentally solitary capacities. They are capacities the exercise of
which takes only one. In knowing that my parents have returned home
on the basis of perception, I am perceiving their return for myself. In
knowing that Napoleon lost the Battle of Waterloo on the basis of
memory, I am remembering my own experience there. And in knowing
that the copperhead is the most common venomous snake in the
greater Houston area on the basis of inductive inference, I am
performing my own calculations. In acquiring knowledge through
perception, memory, or inference, I am utilizing a cognitive capacity the
exercise of which is something I do for myself. In a slogan, I am coming
to my own conclusion about things. When I acquire knowledge by
learning from testimony, however, there is a basic and fundamental
sense in which I am not coming to my own conclusion about things. I
am not blindly adopting the conclusion of someone else, but neither am
I relying solely on my own cognitive resources. Instead, I am utilizing a
cognitive capacity that is essentially cooperative.
The task of this book is, first, to spell out the precise sense in which the
capacity for learning through testimony is a fundamentally social
capacity, and second, to begin to explore the implications that this has
for our understanding of the human mind. In this first chapter I attempt
to demonstrate that the epistemological problem of testimony is most
properly construed as a problem concerning what I will call theoretical
authority. I will use the term ‘theoretical authority’ to refer to a
particular kind of epistemic privilege had by persons or groups in virtue
of which the person or group is in a special position to convey
knowledge. Such authority is to be distinguished from practical
authority, from the authority in virtue of which, for example, a military
officer is in a position to declare a certain area off limits; but
fundamental to both notions of authority is the idea of privilege, of
singling out a particular person or group as the bearer of certain rights
and responsibilities not necessarily had by all. Theoretical authority is
such privilege as it pertains to the processes of belief formation and
knowledge acquisition.4
Of course, this assumes that there is an intelligible sense in which such

privilege applies to the processes of belief formation and knowledge
acquisition. Philosophical intuitions can be surprisingly ambivalent
about this. On the one hand, it seems plainly obvious that there are
theoretical authorities, people whom in comparison with others we
deem to be in a privileged position to convey knowledge. On the other
hand, it can seem that rational belief and knowledge really ought not to
be based on someone else's (p.13) authority. Ultimately, what we know
and believe is our own responsibility, and it would not only be mistaken
but outright dangerous to allow someone else to have authority over
our own beliefs.
I don't want to sort out these issues just yet. The task of the entirety of
this book is to explain in detail how these issues pertain to the
epistemology of testimony. In this chapter, I simply want to point to the
way in which concerns about theoretical authority form the background
against which epistemological questions about testimony have
historically arisen. This isn't as clearly the case with contemporary
debates. Contemporary debates about the epistemology of testimony
tend to be pursued in abstraction from explicit questions about
theoretical authority, and I think that this obscures what is really
important and interesting about the epistemology of testimony.
Historically, however, worries about theoretical authority have been the
deep motivation behind epistemological concerns about testimony.
In many respects, the epistemological problem of testimony is a

decidedly modern problem. Thought about the epistemology of
testimony first took the form that it now has in the Early Modern
period, and it is no coincidence that this period is marked by a deep
suspicion of authority. One of the defining ideas of the early
Enlightenment is the idea that authority and tradition are sources not
of truth and wisdom but of oppression and error. This is evident from
the motto of the Royal Society in Britain—“Nullius in Verba” (“On no
man's word”)—to the opening lines of Kant's “What is Enlightenment?”:
Enlightenment is man's emergence from his self-imposed

immaturity. Immaturity is the inability to make use of one's
understanding without guidance from another. This immaturity is
self-imposed when its cause lies not in lack of understanding, but
in lack of resolve and courage to use it without guidance from
another. Sapere Aude! “Have courage to use your own
understanding”—that is the motto of enlightenment. (1983: 33)
Such an idea clearly has practical and political implications, but it has
profound epistemological implications as well. If authority is conceived
first and foremost as a source of error, then it must be up to us, to each
of us, to determine what to believe. Genuine knowledge must be the
solitary achievement of each individual inquirer.
In his groundbreaking book on testimony, C. A. J. Coady argues that a

latent individualism is to blame for the historical neglect of the
epistemological significance of testimony.
In the post-Renaissance Western world the dominance of an

individualist ideology has had a lot to do with the feeling that
testimony has little or no epistemic (p.14) significance. It is a
commonplace that the political, social, and economic thought and
practice of the West have been profoundly influenced in recent
centuries by certain ideas and ideals stressing the powers, rights,
dignities, and autonomy of the individual person . . . It may be no
accident that the rise of an individualist ideology coincided with
the emergence of the theory of knowledge as a central
philosophical concern but, accident or not, the coincidence was
likely to cast into shadow the importance of our intellectual
reliance upon one another and hence obstruct a serious
examination of the issues this reliance raises. (1992: 13)
While I agree with the spirit of what Coady says here, we need to be
careful about how exactly to frame the issue. Coady seems to suggest
that an independently motivated “individualist ideology” simply
obscured the extent to which much real-world rational inquiry requires

that we rely on one another to gather, pool, and preserve information.
This doesn't get to the heart of the matter, however. The idea that the
acquisition and preservation of much everyday and scientific knowledge
requires reliance on others isn't of itself inconsistent with the kind of
epistemological individualism Coady has in mind, the individualism
paradigmatically associated with philosophers like John Locke and
David Hume. Both Locke and Hume readily accept that rational
scientific inquiry requires a great deal of reliance on others. As Hume
famously states in the Enquiry Concerning Human Understanding,
“there is no species of reasoning more common, more useful, and even
necessary in human life, than that which is derived from the testimony
of men, and the reports of eye-witnesses and spectators” (1999: 111).
The epistemological individualism that we find epitomized in the Early
Modern period thus isn't opposed to epistemic reliance on others.
Rather, it is opposed to theoretical authority, to the idea that knowledge
and belief can be justified by taking things on the authority of others.
Unfortunately, as far as I can tell, a substantive philosophical history of
these issues has yet to be written, and so in this chapter I will attempt
to provide a very preliminary sketch of this history.
Section 1.1 argues that, prior to the Enlightenment, beliefs based on

testimony were often excluded from the category of knowledge but not
because of any explicit worries about theoretical authority. Testimonial
belief was excluded from the category of knowledge because of the way
in which knowledge was itself conceived, but testimonial belief was
nevertheless taken to be a perfectly legitimate form of belief
distinctively based on theoretical authority. Section 1.2 argues that this
conception of the nature of testimonial belief was gradually
transformed in the early Enlightenment in such a way as to divorce
testimonial belief from any epistemologically distinctive connection to
theoretical authority. In concert with general Enlightenment suspicions
concerning the nature and legitimacy of authority, testimony came to be
conceived as just another kind of ordinary inductive evidence and
testimonial belief came to be conceived as (p.15) something for which
an audience was itself solely epistemically responsible. Section 1.3
argues that this shift in the way in which testimony was conceived
helps to explain why Hume and the Port Royal Logic draw from
seemingly similar premises completely opposite conclusions concerning
the rationality of belief in the occurrence of miracles on the basis of
testimony. And section 1.4 argues that this shift sheds new light on the
dispute between Hume's reductionism about testimony and Reid's anti-
reductionism, a dispute that sets the agenda for contemporary debates

about the epistemology of testimony.
1.1 Testimony, Knowledge, and Understanding

One of the earliest discussions in the Western philosophical tradition of
issues concerning the epistemology of testimony occurs in Plato's
Theaetetus. Socrates famously argues that knowledge cannot be
defined as true judgment by appealing to the difference between a
juror and an eyewitness.
Soc.
There is a whole art indicating to you that knowledge is not what you
say [namely true judgment]
.
Theaet.
How's that? What art do you mean?

Soc.
The art of the greatest representatives of wisdom—the men called

orators and lawyers. These men, I take it, use their art to produce
conviction not by teaching people, but by making them judge whatever
they themselves choose. Or do you think there are any teachers so
clever that within the short time allowed by the clock they can teach
adequately the truth of what happened to people who have been
robbed or assaulted, in a case where there were no eye-witnesses?
Theaet.
No, I don't think they possibly could; but they might be able to
persuade them.
Soc.
And by ‘persuading them’, you mean ‘causing them to judge’, don't

you?
Theaet.
Of course.
Soc.
Then suppose a jury has been justly persuaded of some matter which
only an eye-witness could know, and which cannot otherwise be known;
suppose they come to their decision upon hearsay, forming a true
judgment: then they have decided the case without knowledge, but,
granted they did their job well, being correctly persuaded?
Theaet:
Yes, certainly.
Soc.
But, my dear lad, they couldn't have done that if true judgment is the
same thing as knowledge; in that case the best juryman in the world
couldn't form a correct judgment without knowledge. So it seems they
must be different things. (201ac)
(p.16) Socrates here proposes, and Theaetetus agrees, that a juror can
form a true judgment based on the oratory of a lawyer without
acquiring knowledge. If this is true, then clearly it is enough to
dispense with the true judgment account of knowledge. If there can be
cases of true judgment that don't amount to knowledge, then
knowledge cannot be defined as true judgment. Still, the tone of the
passage seems to suggest that a juror can never, at least in a case in
which there are no eyewitnesses, acquire knowledge based on the
speech of a lawyer.5 Socrates claims that the lawyer might very well
persuade the juror to form a correct judgment but that the lawyer is not
in the position to genuinely teach the juror, where teaching appears to
involve transmitting knowledge.
As Myles Burnyeat has noted, this passage poses several interpretive

challenges (1980; 1990: 124–127). First, the passage seems to imply
that knowledge can otherwise be taught, something that Plato calls into
question elsewhere. In Book 7 of the Republic, for example, Socrates
claims that education is not what it is typically thought to be—actively
putting knowledge into a soul where there was none before. Instead,
education involves turning one's intellectual vision in the right direction
so that one is capable of seeing the truth for oneself (518bc). Moreover,
in the Meno Plato famously suggests that the difference between
knowledge and correct opinion concerning such contingent matters of
fact as the way to Larissa lies in the fact that genuine knowledge
requires that one has traveled the road oneself (97b). Again knowledge
seems to require that one see things for oneself, so if teaching involves
transmitting knowledge from one person to another, it looks like
knowledge cannot genuinely be taught. Nevertheless, in the passage
from the Theaetetus, Socrates seems to imply that the lawyer is
distinctively limited to persuasion, that the lawyer is barred from doing
something that is otherwise possible.
Second, it is unclear why exactly the lawyer is so limited. On the one

hand, Socrates makes explicit mention of the time constraints within
which Athenian orators were allowed to present their cases. If this is
the reason that the lawyer is limited to persuasion, then it may very
well be the case that, were these constraints lifted, the lawyer could
succeed in transmitting knowledge to the jury. On the other hand,
Socrates goes on to speak of a juror forming a judgment with respect to

“some matter which only an eye-witness could know, and which cannot
otherwise be known.” This suggests at the very least that there are
some matters that could be at issue in a court of law that only an
eyewitness would be in a position to know and hence that could not be
known by a juror on the basis of hearsay. Interestingly, this seems to
imply that one can acquire knowledge of contingent empirical facts on
the basis of ordinary sense perception, something Plato seems to deny
in the Phaedo and in the Republic. It may also suggest, (p.17) however,
that a juror can never acquire knowledge based on hearsay, that
knowledge is not the kind of thing that can be acquired on the basis of
testimony. But why is this? If it is possible that contingent matters of
fact can be known, and if it is possible that, barring time constraints,
knowledge can genuinely be taught, then why can't knowledge of
contingent empirical facts be taught via testimony?
Much more could certainly be said about the interpretive issues

surrounding Plato's jury example in the Theaetetus. For our purposes,
however, it is instructive to note that Plato immediately goes on from
here to introduce the conception of knowledge as, roughly, true
judgment with an account. The above passage continues:
Theaet:
Oh, yes, Socrates, that's just what I once heard a man say; I had
forgotten, but now it is coming back to me. He said that it is true
judgment with an account [logos] that is knowledge. And he said that
things of which there is no account are not knowable (yes, he actually
called them that), while those which have an account are knowable.
(201cd)
Notoriously, the Greek word ‘logos’ has a wide array of meanings, but
here it certainly refers to something far more comprehensive than the
contemporary epistemologist's ‘reason’ or ‘justification.’ It likely refers
to something like a systematic explanatory account of the phenomenon
in question, something that yields robust scientific understanding. If
this is what is required to turn true judgment into knowledge, then
Plato's seeming dismissal of the possibility of testimonial knowledge
makes more sense. Plausibly, coming to believe that p on the basis of a
speaker's testimony that p does not typically involve acquiring this kind
of systematic understanding concerning p. Along these lines, Burnyeat
writes:
Much of what Plato says about knowledge and its relation to true
opinion falls into place if we read him, not as misdescribing the
concept which philosophers now analyze in terms of justified true
belief, but as analyzing a richer concept of knowledge tantamount
to understanding. (a) If knowledge (episteme) is understanding, in
many fields that does arguably require that one master for oneself
a proof or explanation; and here it is well to note that the Meno's
leading condition on knowledge, aitias logismos (98A), is Greek
for working out the explanation of something, not for assembling a
justification for believing it, which the slave already has at a stage
when Plato denies he has knowledge (85C). (b) If proof or
explanation is something that one has to work at for oneself, it is
natural that the notion of teaching should be problematic for
Plato. It is natural that he should vacillate between the view that a
good teacher will get his pupil to see things for himself . . . and the
Meno view that at least in morals and mathematics teaching does
not produce (p.18) knowledge, so that if there is an honest job for
teaching to do, it is in the transmission of practical skills, not of
theoretical knowledge (89D ff). (c) If, as Plato thinks, proof or
explanation rests ultimately on definition, we can see why he
should so often insist that definitional knowledge (knowledge of
Forms) is a prerequisite for knowing anything else. It is not that
without a definition you cannot be certain of anything or be
justified in believing it, but that you will not have an adequate
understanding of e.g. something's being beautiful or becoming
two (Phdo 100B ff), or even of a mathematical theorem for which
you have an acceptable proof (Rep. 533BE). The epistemic ideal
for which the Republic aims is most frequently described not in
terms of certainty but in terms of clarity, the total clarity of a
synoptic understanding of all the sciences. (1980: 186-187)
If, in the end, knowledge for Plato requires the kind of systematic
theoretical understanding found in the sciences, paradigmatically in
mathematics, then knowledge may very well be transmitted through
teaching, but it cannot be simply passed from one soul to another.
Imparting knowledge through teaching will involve helping the student
to see and understand for herself the connection between things. And
even if one can know such contingent empirical facts as the way to
Larissa, such knowledge will involve something analogous to scientific
knowledge, something like a detailed understanding of all of the twists
and turns along the way, something that one must see for oneself and
that cannot simply be taught. In this sense, Plato's conception of
knowledge as involving a kind of systematic scientific understanding
appears of itself to rule out the possibility of acquiring genuine
knowledge from testimony. Hearsay may be a very good reason for

believing a great many things, especially concerning contingent
empirical facts, and hence it may produce a good deal of true,
reasonable, and responsible judgment, as in the case of the jury, but
such judgment does not thereby attain the kind of systematic clarity
required for it to amount to knowledge.
This conception of knowledge as episteme or scientia, as involving a

systematic understanding of the elements in a comprehensive theory,
remained prevalent from antiquity through the Middle Ages and into
the Early Modern period. Such knowledge was primarily conceived as
pertaining to necessary truths, and it was typically taken to be acquired
through the routes of either demonstration or intellectual intuition.
Belief in contingent matters of fact based on testimony was often
classified as “faith” or “opinion” and was excluded from the category of
knowledge. Importantly, however, testimony was nevertheless
conceived as a perfectly good reason for belief and as a necessary
source of much true opinion, particularly concerning the events related
in scripture. In his Confessions, for example, Augustine notes that
believing things on the basis of testimony is necessary and legitimate
for carrying on our daily affairs, and he takes this as further (p.19)
reason to believe the testimony of the Bible (1961: 117).6 Due to the
systematicity requirement on genuine knowledge, testimony was not
deemed to be a genuine source of knowledge, but it was nevertheless
taken to be a perfectly legitimate reason for belief.
This was still very much the case for many seventeenth century figures
like Descartes. Descartes was much influenced by the Scholastic
conception of knowledge as scientia, and it is for this reason that he
dismisses the possibility of knowledge acquired by testimony. In Part II
of the Discourse on Method, Descartes denies that genuine knowledge
can be acquired from the testimony of books and teachers, but this is
because the firm, clear, unclouded knowledge to which he aspires
requires the kind of systematic understanding that can only be
achieved by seeing the connection between things for oneself. True
beliefs derived from reasonings that are merely probable, though
important and necessary in ordinary life, do not yield this kind of
systematic understanding (1985: 116–117). Again, it is the
systematicity requirement on knowledge conceived in terms of scientia
that rules out the possibility of knowledge based on testimony.7
1.2 From Human Faith to Inductive Evidence

The systematicity requirement on knowledge was gradually relaxed in
the Early Modern period, largely in concert with what Ian Hacking
(1975) has called “the emergence of probability.” As probabilistic
reasoning gradually came to occupy a central place in the conduct of
the sciences, epistemological conceptions of the nature of knowledge
were gradually relaxed to make room for the possibility of knowledge
based on non-demonstrative reasoning. One would expect that such a
transformation of the concept of knowledge would be especially
congenial to countenancing the possibility of testimonial (p.20)
knowledge, and in a sense, it was. Testimony soon came to be

conceived as just another ordinary consideration on the basis of which
one can acquire nondemonstrative knowledge. Importantly, however,
this involved a dramatic shift in the way in which testimony itself was
conceived, a shift that involved a rejection of theoretical authority.
While the ancient and medieval conception of knowledge as episteme or

scientia resulted in the exclusion of belief based on testimony from the
category of knowledge, it nevertheless allowed for the view that
learning from testimony was a distinctive kind of cognitive capacity
very different from the capacities of demonstration and intellectual
intuition. Learning from testimony was typically conceived as a capacity
for basing beliefs on theoretical authority. The proper exercise of this
capacity produced beliefs that amounted to faith or opinion, not
knowledge, but nevertheless this capacity for basing beliefs on
theoretical authority was taken to be perfectly legitimate. Beliefs based
on testimony were excluded from the category of knowledge, but this
was because they enjoyed a category of their own, a category that was
of the utmost importance for the conduct of ordinary life.
This kind of view is clearly advanced in the context of the notably

Cartesian epistemology of Arnauld and Nicole's Logic or the Art of
Thinking (also known as the Port Royal Logic). In chapter 12 of the
Logic, knowledge derived from authority, either human authority or
divine authority, is carefully distinguished from knowledge based on
reason.
Everything we have just said up to now concerns knowledge that

is exclusively human and based on rational evidence. But before
ending it will be good to discuss another kind of knowledge that
often is no less certain nor less evident in its own way, namely
knowledge derived from authority.
For there are two general paths that lead us to believe that
something is true. The first is knowledge we have of it ourselves,
from having recognized and examined the truth either by the
senses or by reason. This can generally be called reason, because
the senses themselves depend on a judgment by reason, or
science, taking this name more generally than it is taken in the
Schools, to mean all knowledge of an object derived from the
object itself.
The other path is the authority of persons worthy of credence who

assure us that a certain thing exists, although by ourselves we
know nothing about it. This is called faith or belief, following the
saying of St. Augustine: Quod scimus, debemus rationi, quod
credimus, autoritati [What we know we owe to reason, what we
believe, to authority]. (1996: 260)
It is noteworthy that the beginning of this passage appears to allow that

belief based on authority can actually amount to knowledge. This is
then taken back in the quote from (p.21) Augustine. More important,
however, is the way in which such knowledge or belief based on
authority is sharply distinguished from knowledge based on reason.
Whereas reason produces “knowledge we have of it ourselves,”
“knowledge of an object derived from the object itself,” faith produces
knowledge or belief “derived from authority” where “by ourselves we
know nothing about it.” Here belief based on human testimony would
appear to be a very particular kind of belief distinctively based on
theoretical authority.
As the Logic goes on, however, this distinction between faith and reason
quickly becomes problematic in a way that lays the foundation for much
subsequent thought about testimony. Hacking claims that it is in the
closing chapters of the Logic that the word ‘probability’ was first
applied to something epistemically measurable (1975: 73). He argues
that the medievals used the term ‘probable’ only with respect to
opinion based on authority, not with respect to knowledge, and that the
term was not originally applied to ordinary evidence, as we use it today.
In the medieval period, to say that an opinion was probable was to say
that it was approved or accepted by authorities (1975: 22). Hacking
argues that this medieval conception of probability was gradually
transformed in the Early Enlightenment through the work of empirics,
alchemists, and physicians. Probability was connected to the notion of a
sign; the notion of a sign was eventually applied to the natural world
via conceiving the natural world as a book authored by God, the
ultimate authority; and finally the notion of a sign was fully naturalized
and disconnected entirely from the notion of authority by the work of
figures like Pascal. Pascal was a close associate of Port Royal, and
Hacking speculates that he had a great deal of influence on the ideas
that gained expression in the Logic, especially with respect to
probability.
What Hacking calls “the probability chapters” of the Logic begin

immediately after the chapter that argues for the distinction between
knowledge based on reason and on faith, a chapter that Hacking
notably calls “lame and conventional” (1975: 75). The first of these
probability chapters contains the following rule concerning how to
employ one's reason when forming beliefs about contingent matters of
fact on the basis of human testimony:
In order to decide the truth about an event and to determine

whether or not to believe in it, we must not consider it nakedly
and in itself, as we would a proposition of geometry. But we must
pay attention to all the accompanying circumstances, internal as
well as external. I call those circumstances internal that belong to
the fact itself, and those external that concern the persons whose
testimony leads us to believe in it. Given this attention, if all the
circumstances are such that it never or only rarely happens that
similar circumstances are consistent with the falsity of (p.22) the
belief, the mind is naturally led to think that it is true. Moreover,
it is right to do so, above all in the conduct of life, which does not
require greater certainty than moral certainty, and which even
ought to be satisfied in many cases with the greatest probability.
(1996: 264)
The Logic here claims that an individual's acceptance of a speaker's

testimony that p should be rationally responsive to two kinds of
circumstances: internal circumstances, circumstances concerning the
proposition p itself, and external circumstances, circumstances
concerning the speaker who testifies that p. Hacking claims that
internal circumstances are a rudimentary formulation of the modern
notion of inductive evidence: “The ‘internal’ circumstances of the event
are those that bear on the place of the event in nature—whether it is
the sort of thing that tends to happen” (1975: 79). Internal
circumstances thus concern the likelihood or probability of p, of the
event or fact in question. By contrast, external circumstances concern
such things as the sincerity, competence, and reliability of the speaker
who has testified that p.
Hacking rightly notes that the Logic here lays the groundwork for the
eighteenth-century debate about belief in the occurrence of miracles on
the basis of testimony most famously associated with Hume's Enquiry

Concerning Human Understanding.
Hume was able to turn this chapter of the Logic on its head. In his
essay On Miracles he argued that no external circumstances could
ever suffice to render probable an event improbable enough to be
called a miracle. That thought created another flurry in the
concept of probability. The Port Royal Logic had already set the
conceptual scheme in which the debate was to be conducted
eighty years later. (1975: 79)
The chapter of the Logic immediately following the distinction between

internal and external circumstances argues that human testimony can
be sufficient to warrant belief in miracles, and Hume does certainly
turn this on its head. But what Hacking says here nevertheless glosses
over the fact that the conception of the kind of reason for belief
provided by testimony changes dramatically from the Logic to Hume's
discussion of miracles in the first Enquiry. The chapter of the Logic that
Hacking dismisses as “lame and conventional” sharply distinguishes
knowledge based on reason from knowledge or belief based on human
faith. The Logic quickly goes on to admit that we must often employ
reason in forming beliefs concerning contingent matters of fact on the
basis of human faith, and the distinction between internal and external
circumstances in the subsequent chapter offers a rule concerning how
to do this—the title of chapter 13 is “Some rules for directing reason
well in beliefs about events that depend on human faith.” The rule
states that in forming beliefs on the basis of human (p.23) faith, an
audience must be rationally responsive to considerations concerning (1)
the likelihood of the event in question and (2) the trustworthiness of the
speaker. In this sense, as the Logic states, “faith always presupposes
some reason” (1996: 261). Crucially, however, this is intended to be
completely consistent with the distinction between knowledge based on
faith and knowledge based on reason. Even though faith always
presupposes reason in the sense stated by the rule, the rule is intended
to be consistent with belief based on faith being different in kind from
belief based on reason. Even if an audience carefully assesses for
herself the likelihood of the event testified to and the trustworthiness of
the speaker, if she then goes on to believe what the speaker says, her
belief is nevertheless based on faith; it is derived from authority where
she alone knows nothing of it.
For Hume, however, belief based on testimony is no longer

distinguished from belief based on reason. Hume claims that belief
based on testimony is simply a species of ordinary belief based on

nondemonstrative causal reasoning.
This species of reasoning, perhaps, one may deny to be founded

on the relation of cause and effect. I shall not dispute about a
word. It will be sufficient to observe that our assurance in any
argument of this kind is derived from no other principle than our
observation of the veracity of human testimony, and of the usual
conformity of facts to the reports of witnesses. It being a general
maxim, that no objects have any discoverable connexion together,
and that all the inferences, which we can draw from one to
another, are founded merely on our experience of their constant
and regular conjunction; it is evident, that we ought not to make
an exception to this maxim in favour of human testimony, whose
connection with any event seems, in itself, as little necessary as
any other. Were not the memory tenacious to a certain degree;
had not men commonly an inclination to truth and a principle of
probity; were they not sensible to shame, when detected in a
falsehood: Were not these, I say, discovered by experience to be
qualities, inherent in human nature, we should never repose the
least confidence in human testimony. (1999: 111-112)
According to Hume, our beliefs based on testimony are based on

nothing more than our general experience of human nature and of the
conjunction between reports (or types of reports) and facts (or types of
facts). We must proportion our belief in the proposition advanced by
any particular piece of testimony in accordance with the regularity of
the conjunction we have experienced between the particular kind of
testimony and the particular kind of fact. Belief based on testimony is
thus simply belief based on a causal inference from a variety of factors
concerning the speaker's offering of the particular piece of testimony
and the particular matter of fact that is the object of the testimony to
the conclusion that the proposition advanced by the testimony is (p.24)
true. For Hume, then, testimony is no longer conceived as a kind of

reason for belief distinctively connected to theoretical authority.8
The shift that occurs between the way in which testimony is conceived
in the Logic and in Hume's first Enquiry is important and dramatic and
can be found in much Early Enlightenment thought. One of the clearest
examples of the way in which testimony was divorced in the
Enlightenment from its traditional connection to authority can be found
in the work of Locke. Like many of his fellow moderns, Locke is very
attracted to the classical conception of knowledge as scientia, and as
we've seen, this in itself is enough to rule out the possibility of
knowledge based on testimony. If the paradigm of knowledge involves
the kind of systematic understanding characteristic of mathematical

proof, then belief based on testimony will quite plausibly be excluded.
However, Locke is keenly aware of the need for rational belief that falls
short of knowledge so conceived, and so in the Essay Concerning
Human Understanding he sets out to present rules for the proper
conduct of the understanding when going about forming beliefs that fall
short of knowledge—specifically, beliefs based on probabilistic
reasoning. For Locke, all beliefs based on probabilistic reasoning fall
short of knowledge strictly conceived—in the context of Locke's
epistemology this is due to the fact that knowledge requires intuition of
the certain connection between ideas, something that is lacking in the
case of belief based on probabilistic considerations—but nevertheless
belief based on broadly nondemonstrative inference can rise to the
level of what he calls “assurance,” something at least approaching
knowledge in its degree of certainty (1975: 655).
Locke goes on to list two general grounds for belief based on

probabilistic reasoning: “the conformity of any thing with our own
Knowledge, Observation, and Experience” and “the Testimony of
others, vouching their Observation and Experience” (1975: 666).
Testimony is thus one of the two general grounds of probable belief. In
fact, Locke's leading example of the distinction between genuine
knowledge and merely probable belief concerns the difference between
knowing a mathematical proposition on the basis of working through a
proof and believing the same proposition on the basis of the testimony
of a respected mathematician (1975: 654). In this sense, Locke clearly
holds that testimony amounts to an important and perfectly legitimate
source of rational belief based on the assessment of probabilities. Due
to its merely probable character, belief based on testimony cannot rise
to the level of knowledge, but if the testimony is sufficiently credible, it
can nevertheless rise to the level of assurance.
Significantly, however, we also find in Locke some of the

Enlightenment's most florid indictments of belief based on authority. As
Nicholas Wolterstorff notes, “nothing so (p.25) effectively stirs up
eloquence in Locke as this subject of the assent to authority” (1996:
100). In Book I of the Essay, Locke writes:
This I am certain, I have not made it my business, either to quit,

or follow any Authority in the ensuing Discourse: Truth has been
my only aim; and where-ever that has appeared to lead, my
thoughts have impartially followed, without minding, whether the
footsteps of any other lay that way, or no. Not that I want a due
respect for other Mens Opinions; but after all, the greatest
reverence is due to Truth; and, I hope, it will not be thought
arrogance, to say, That, perhaps, we should make greater
progress in the discovery of rational and contemplative
Knowledge, if we sought it in the Fountain, in the consideration of
Things themselves; and made use rather of our own Thoughts,
than other Mens to find it. For, I think, we may as rationally hope
to see with other Mens Eyes, as to know by other Mens
Understandings. So much as we our selves consider and
comprehend of Truth and Reason, so much we possess of real and
true Knowledge. The floating of other Mens Opinions in our
brains makes us not one jot the more knowing, though they
happen to be true. What in them was Science, is in us but
Opiniatry, whilst we give up our Assent only to reverend Names,
and do not, as they did, employ our own Reason to understand
those Truths, which gave them reputation . . . In the Sciences,
every one has so much, as he really knows and comprehends:
What he believes only, and takes upon trust, are but shreds; which
however well in the whole piece, make no considerable addition to
his stock, who gathers them. Such borrowed Wealth, like Fairy-
money, though it were Gold in the land from which he received it,
will be but Leaves and Dust when it comes to use. (1975: 100-101)
As I have said, Locke is very attracted to the classical conception of

knowledge according to which genuine knowledge requires the kind of
systematic understanding characteristic of mathematical proof. This
conception of knowledge is itself enough to rule out the possibility of
genuine knowledge based on the testimony of authorities, and surely
this is part of what is going on in this passage. However, the rhetoric of
the passage seems to go further, suggesting that beliefs based on trust
are necessarily somehow counterfeit or fraudulent. This goes well
beyond the simple claim that, due to general constraints on the
category of knowledge, beliefs based on testimony cannot amount to
knowledge.
Similar rhetoric can be found in Locke's treatment of written testimony

in his later work Of the Conduct of the Understanding. There Locke
distinguishes between two things that are “recorded” by an author in
writing, “matters of fact” and “reasonings,” and he argues that
reasonings are “most properly the business of those who pretend to
improve their understanding and make themselves knowing by

reading” (1996: 199). (p.26) What Locke has in mind here is that
writing can include both cases of testimony, where the author comes
out and tells us certain facts, and cases of argument, where the author
presents us with a train of reasoning that is intended to lead us to a
certain conclusion.9 While Locke admits that we often do need to rely
on an author's testimony, he thinks that it is her arguments that are of
real epistemological significance.
We are mightily beholden to judicious writers of all ages for those

discoveries and discourses they have left behind them for
instruction, if we know how to make a right use of them; which is
not to run them over in a hasty perusal, and perhaps lodge their
opinions or some remarkable passages in our memories, but to
enter into their reasonings, examine their proofs, and then judge
of the truth or falsehood, probability or improbability of what they
advance, not by any opinion we have attained of the author, but by
the evidence he produces and the conviction he affords us, drawn
from things themselves. Knowing is seeing, and, if it be so, it is
madness to persuade ourselves that we do so by another man's
eyes, let him use ever so many words to tell us that what he
asserts is very visible. Until we ourselves see it with our own eyes,
and perceive it by our own understandings, we are as much in the
dark and as void of knowledge as before, let us believe any
learned author as much as we will. (1996: 200-201)
Again, if Locke's claim here is simply that genuine knowledge requires

the kind of systematic understanding that can only be acquired by
working through an argument or proof for oneself, then it is quite
plausible that genuine knowledge cannot be acquired from testimony.
Still, the rhetoric seems to suggest that there is something dangerous
and illegitimate about belief based on authority itself, that regardless of
whether such belief can amount to genuine knowledge, there is
something about the activity of taking things on authority that is by its
very nature a misuse of one's understanding.
In this respect, Locke goes beyond the claim that testimony is not a
source of genuine scientia, also rejecting assent to authority as a basis
for merely probable belief. In Book IV of the Essay Locke explicitly lists
deference to authority as one of the four “wrong measures of
probability,” one of the four ways in which people make wrong use of
their reason in forming beliefs on the basis of probability.
The fourth and last wrong Measure of Probability I shall take note
of, and which keeps in Ignorance, or Errour, more People than all
the other together, is . . . the (p.27) giving up our Assent to the
common received Opinions, either of our Friends, or Party;
Neighborhood, or Country. How many Men have no other ground
for their Tenets, than the supposed Honesty, or Learning, or
Number of those of the same Profession? As if honest, or bookish
Men could not err; or Truth were to be established by the Vote of
the Multitude: yet this with most Men serves the Turn. The Tenet
has had the attestation of reverend Antiquity, it comes to me with
the Pass-port of former Ages, and therefore I am secure in the
Reception I give it: other Men have been, and are of the same
Opinion, (for that is all is said,) and therefore it is reasonable for
me to embrace it. A Man may more justifiably throw up Cross and
Pile for his Opinions, than take them up by such Measures. (1975:
718)
As we've seen, Locke takes testimony to be one of the two general

grounds for belief based on probabilistic reasoning. Nevertheless, he
here claims that deference to authority is a misuse of one's reason
when tasked with forming such beliefs. This can seem puzzling. If
testimony is a perfectly legitimate ground for belief, then what is so
unreasonable about deference to theoretical authority?
The tension here is typical of much Early Enlightenment thought.

Discussing Early Modern scientific practice, Steven Shapin claims that
“the rejection of authority and testimony in favor of individual sense-
experience is just what stands behind our recognition of seventeenth-
century practitioners as ‘moderns,’ as ‘like us,’ and, indeed, as
producers of the thing we can warrant as ‘science’ ” (1994: 201). He
continues:
A few examples suffice to illustrate the topic and highlight the

presumed relationship between individualism and epistemic
virtue. William Harvey's On the Generation of Animals directed
readers “to strive after personal experience, not to rely on the
experience of others,” even urging them “to take nothing on trust
from me.” John Evelyn's Sylva condemned writers “who receiving
all that came to hand on trust, to swell their monstrous Volumes,
have hitherto impos'd upon the credulous World, without
conscience of honesty.” It was “base” and “servile” to subject our
divinely sensing faculties to the “blind Traditions” of authority. A
manuscript note by Boyle on the “Use of Reason in Natural
Philosophy” asserted that “the great Reverence men usually give
to humane Authority is undeserved” and that “Humane Testimony
ought not to be of force against either right Reason or
Experience.” It was, Boyle wrote, “improper” to “urge and rely on
Testimonys for matters, whose Truth or Falsehood may be proved
by manifest Reason or easy Experiment.” Boyle urged
practitioners to follow his example in scrupulously distinguishing
“betwixt the matters of fact, they deliver as upon their own
knowledge, and those, which they have upon trust from
others.” . . . It was a rhetoric which insisted that no source of
factual information possessed greater reliability or inspired
greater confidence (p.28) than the direct experience of an
individual. The legitimate springs of empirical knowledge were
located in the individual's sensory confrontation with the world.
(1994: 201-202)
Shapin immediately goes on to note, however, that these indictments of

authority typically appear alongside a keen appreciation of the need for
reliance on testimony in ordinary scientific practice.
This sort of individualist rhetoric, taken by itself and at face value,

would count as a massive misrepresentation of scientific practice.
In fact, seventeenth-century English natural historians and
natural philosophers, writing in different moods and for other
purposes, showed themselves well aware that it was. Many of the
same practitioners who produced some of the most vigorous
individualistic methodological pronouncements also displayed
keen appreciation that there was a proper, valuable, and
ineradicable role for testimony and trust within legitimate
empirical practices. (1994: 202)
Shapin characterizes this as a largely unrecognized contradiction

between theory and practice, but as is evident in the case of Locke, this
tension is present in Early Modern epistemological theory itself. On the
one hand, many Early Modern figures are keenly aware of the need for
epistemic reliance on testimony. On the other hand, one of the defining
characteristics of the age—and part of what makes these figures
“moderns,” as Shapin puts it—is the deep suspicion of epistemic
appeals to authority. Genuine knowledge (and fully rational belief
falling short of knowledge) is something that must be acquired “for
ourselves.” Doing our epistemic best requires that we are careful to use
our own reason in deciding what to believe.
In the most general terms, the idea here is that knowledge and justified
belief can only be acquired by coming to our own conclusion about
things. Taking something on the authority of others is conceived as
something expressly opposed to coming to our own conclusion about
things. Therefore knowledge and justified belief cannot be acquired on
the basis of theoretical authority. The requirement that knowledge must
be the result of coming to our own conclusion about things explicitly
rules out the possibility of knowledge based on authority. But how then
are we to construe our epistemic reliance on testimony? How can we
make our epistemic reliance on testimony consistent with the general
requirement that we come to our own conclusion about things?
The answer here, an answer clearly present in the case of Locke, is that
we must divorce testimony from its traditional connection to authority.
Belief based on testimony is thus no longer conceived as the result of a
distinct cognitive capacity for (p.29) forming beliefs on the basis of
theoretical authority. It no longer amounts to what the Logic calls
human faith. Testimony is a perfectly legitimate and necessary source
of belief, but it is so only if we are careful to “employ our own Reason”
in forming testimonial beliefs rather than “give up our Assent only to
reverend Names.” And “employing our own reason” here no longer
simply means being rationally responsive to considerations of the
likelihood of the fact testified to and the trustworthiness of the speaker,
as it does for the Logic. (Recall that the Logic's rule for the use of
reason is a rule for using reason in forming beliefs distinctively based
on human authority.) Instead, it now means treating a speaker's
testimony as a consideration that is to be weighed on just the same
scale as any other ordinary evidence. Given a particular instance of
testimony to the proposition p, an audience must place the instance of
testimony alongside other available evidence for and against p, assess
the probability of p on this body of evidence, and then proportion its
level of confidence in p to p's probability on the evidence. In this sense,
testimony is simply a species of ordinary inductive evidence.
For Locke, then, testimonial belief is not the result of its own distinctive
cognitive capacity. Instead, testimonial belief is the result of a very
general capacity for broadly nondemonstrative inference. This capacity
takes as input both testimonial and nontestimonial input, and it outputs
both testimonial and nontestimonial belief. Most importantly, the
functioning of this capacity doesn't recognize any distinctions between
testimonial and nontestimonial input and output. As far as the capacity
for nondemonstrative inference is concerned, there is no difference
between testimonial and nontestimonial evidence and there is no
difference between testimonial and nontestimonial belief. Testimony is
not distinctively connected to theoretical authority in any
epistemologically relevant way, and neither is testimonial belief.
Testimony amounts to a species of evidence on the basis of which an
audience is in the position of having to draw an ordinary inductive
inference to the conclusion that the proposition advanced by the
testimony is true, and belief based on testimony is a species of belief
based on ordinary inductive inference.
The point of Locke's construing testimonial belief as simply a species of

belief based on inductive inference is to make testimonial belief
consistent with the general requirement that doing our epistemic best
requires that we come to our own conclusion about things. For Locke,
assessing the probability of a given proposition on the available
evidence involves nothing beyond the employment of an individual's
own reason, and he sees it as the duty of every individual who has the
time and opportunity to exercise her understanding properly that she
be careful to so employ her own reason with regard to every piece of
testimony that she confronts. As Wolterstorff puts it, “Locke's picture of
the community of responsible believers is the picture of a democracy in
which each listens to his or her own inner voice of Reason and no one
treats any voice outside (p.30) himself or herself as authoritative—
unless his or her Reason tells him or her to do so” (1996: 87).10 This
can be construed as a point about epistemic responsibility. In order to
be a fully epistemically responsible inquirer, each individual must be
sure to come to her own conclusion about the content of a speaker's
testimony. Each individual inquirer must exercise her own reason such
that she is solely epistemically responsible for the justification of her
testimonial beliefs. This is consistent with epistemic reliance on the
testimony of others, but only if such reliance is something that the
individual inquirer is herself solely responsible for, only if such reliance
involves nothing beyond the exercise of the individual's own reason.
Locke thus accepts a thesis that we might call epistemic autonomy.11

According to epistemic autonomy, fully rational cognitive agents are
always solely epistemically responsible for the justification of their own
beliefs. Fully rational cognitive agents are always in the position of
coming to their own conclusion about things. This isn't to say that other
people can't be causally or even morally responsible for putting a
subject in a position to, for example, infer correct or incorrect
conclusions from evidence. The claim of epistemic autonomy is only
that, from an epistemological perspective, whether a given subject
knows or justifiably believes that p is “up to her” in the sense that she is
solely responsible for the epistemic conclusions concerning p that she
draws from within her own epistemic position.12
Thus formulated, epistemic autonomy is fully consistent with epistemic

reliance on the testimony of others.13 An audience can acquire
knowledge (or for Locke, belief approaching knowledge) on the basis of
a speaker's testimony as long as the audience (p.31) is careful to come
to its own conclusion about the content of the speaker's testimony, as
long as the audience is still solely epistemically responsible for the
justification of its testimonial belief. The audience can succeed in
remaining so responsible by treating a speaker's testimony that p as
ordinary inductive evidence that counts in favor of p. While the
audience's knowledge may still loosely be said to be “based on the
authority of the speaker,” the audience is here in the position of
inferring its own conclusion about p and thus has not given over to the
speaker any of its own epistemic responsibility. The audience's
knowledge based on testimony is thus not based on authority in the
sense that Locke finds so objectionable.14
This helps to clarify what Locke and so many other Enlightenment

figures are railing against when they criticize knowledge and belief
based on authority. They object to belief based on authority only insofar
as it involves a subject's giving over to the authority epistemic
responsibility for the justification of her own beliefs. Theoretical
authority is objectionable only in so far as it is inconsistent with
epistemic autonomy. The indictments of theoretical authority so
common in the Early Enlightenment are thus affirmations of epistemic
autonomy, and as such they help to usher in a new conception of the
epistemic significance of testimony, one that divorces testimony from its
traditional connection to theoretical authority. The rejection of
theoretical authority as a distinctive kind of reason for belief brings
with it a new conception of testimony as just another kind of ordinary
inductive evidence.15 This then explains how the Enlightenment
rejection of authority can coexist with an explicit appreciation of the
need for reliance on testimony. If testimony is construed as nothing

more than ordinary inductive evidence, then forming beliefs on the
basis of testimony doesn't involve any reliance on theoretical authority
and is completely consistent with epistemic autonomy.
1.3 On Miracles: Hume Versus Port Royal

In order to further illustrate the significance of the Enlightenment
conception of testimony as ordinary inductive evidence, it will prove
instructive to examine in some detail Hume's famous argument against
belief in the occurrence of miracles on the (p.32) basis of testimony.
This will amount to a slight digression from our examination of the
historical development of philosophical conceptions of knowledge and
belief based on testimony, but the digression is an illuminating one. The
shift that occurs in Early Enlightenment thinking from construing
testimony as a distinctive source of belief based on human authority to
construing testimony as ordinary inductive evidence is something that
has not been well understood, and failing to recognize this shift has the
effect of shielding from view much of what is at stake in philosophical
discussions concerning issues connected to testimony. This is
particularly true of Hume's argument against belief in miracles on the
basis of testimony in Section X of the first Enquiry. The considerations
that Hume appeals to in the course of making this argument appear to
be very similar to considerations advanced by other Early Modern
figures who proceed to derive from them very different conclusions. In
particular, the Port Royal Logic lays much of the groundwork for
Hume's argument, but it then goes on to argue that belief in the
occurrence of miracles on the basis of testimony can be perfectly
justified. I think that the key to explaining this lies in appreciating that
the premises appealed to by Hume and by the Logic are not that similar
after all. The difference lies in the fact that, as we've seen, Hume is
working with a very different conception of the epistemology of
testimony from that found in the Logic. Appreciating this helps us to
understand not only why Hume draws the particular conclusion that he
does, but it also helps us to develop a sense of what is at stake between
the competing conceptions of testimony present in the Logic and in
Hume's first Enquiry.16
As we've seen, the Logic holds that knowledge or belief based on

human testimony is distinctively based on theoretical authority. It
amounts to human faith and must be distinguished from knowledge
based on reason. Nevertheless, faith always presupposes some reason.
In particular, in deciding whether to believe a speaker's testimony that
p, an audience must be careful to assess (1) the probability of p and (2)
the trustworthiness of the speaker. Rational responsiveness to
considerations of the probability of p and of the trustworthiness of the
speaker is required for belief in the speaker's testimony to be
warranted, and such rational responsiveness is particularly crucial
when it comes to testimony to the occurrence of miracles. Without it,
the Logic warns, we are in danger of falling into the extremes of either
credulity or skepticism, of believing all testimony to the occurrence of

miracles or believing none. In order to avoid these extremes, we should
believe only those instances of testimony to the occurrence of miracles
that are rendered credible by the surrounding circumstances. (p.33)
Piety does not oblige a person of good sense to believe all the
miracles reported in the Lives of the Saints or in Metaphrastes,
because these authors are full of so many fables that there is no
reason to be sure of anything based merely on their testimony . . .
But I maintain that persons of good sense, even if they are devoid
of piety, ought to recognize as authentic the miracles St.
Augustine relates in his Confessions or in City of God as having
taken place before his eyes, or about which he testifies to having
been particularly informed by the persons themselves to whom
these things happened. (1996: 267)
The Logic holds that with respect to each individual instance of

testimony to the occurrence of miracles, we must assess the internal
and external circumstances surrounding the testimony in an effort to
determine whether the testimony is credible. If we are careful to do
this, we will find that some testimony to the occurrence of miracles is
credible, while other testimony clearly is not.
Similarly, Hume argues that in accepting any piece of testimony we

must be rationally responsive to two kinds of considerations: (1)
considerations concerning the probability of the fact or event testified
to—concerning the extent to which the fact or event corresponds with
past experience—and (2) considerations concerning the trustworthiness
of the speaker. Clearly, these two kinds of considerations roughly
correspond to the Logic's distinction between internal and external
circumstances. However, from here Hume goes on to argue for a
conclusion completely contrary to that drawn by the Logic. He argues
that no testimony to the occurrence of miracles has been sufficiently
credible to warrant belief in the occurrence of miracles.
In very brief outline, Hume claims that miracles are by definition

violations of the laws of nature and hence events that run counter to all
previous experience and observation. This amounts to strong evidence
against the occurrence of the miracle on the basis of considerations of
type (1) above. As such, the standards for accepting testimony to the
occurrence of miracles are extremely high. We must have even stronger
evidence on the basis of considerations of type (2) above if the truth of
the testimony is to amount to a probability, let alone a proof. In the case
of miracles used to establish the foundation of religion, Hume claims
that these standards cannot be met. This is due to the great variety in
religious belief and to our previous experience with deception in the

name of religion. In the case of nonreligious miracles, Hume claims that
these standards could possibly be met but that as a matter of historical
fact they never have been met and we have little reason to believe that
they will be met in the future.17
In contrast to the Logic, then, Hume thinks that the probative value of
testimony to the occurrence of miracles is simply insufficient to
outweigh the inherent improbability (p.34) of miraculous events. On
this score there seems to be a straightforward difference of opinion.
Whereas the Logic holds that the trustworthiness of some of those who
have testified to the occurrence of miracles, most notably Saint
Augustine, is sufficient to make the strength of their testimony
outweigh the inherent improbability of miraculous events, Hume holds
that no testimony has ever been sufficient to do so. It is an interesting
feature of Hume's argument that, even though he was surely aware of
the very similar discussion of these issues in the Logic, he doesn't
acknowledge the fact that so many other Enlightenment figures appeal
to very similar premises while drawing from them completely different
conclusions.18 Hume doesn't seem to allow for the possibility that one
might find some particular historical testimony to the occurrence of
miracles far more credible than he himself does. In this sense, Hume's
argument leaves open the possibility that one might simply disagree
about the credibility of particular instances of testimony to the
occurrence of miracles.
I suspect that there is more going on here, however. It might be the

case that the very different conclusions drawn by Hume and the Logic
concerning the trustworthiness of particular instances of testimony to
the occurrence of miracles are at least partially affected by the very
different ways in which they conceive of belief based on testimony. For
the Logic, belief based on human testimony is a piece of human faith; it
is belief distinctively based on theoretical authority. Testimony thus has
a very distinctive kind of probative value. For Hume, belief based on
testimony is just a species of belief based on ordinary causal reasoning.
The probative value of testimony is thus no different from the probative
value of other ordinary events that we have come to associate together
causally.
This means that, from Hume's point of view, assessing the reliability of
an instance of testimony is no different from assessing the reliability of
any other piece of ordinary evidence. It involves calculating the
likelihood that this particular piece of testimony is correlated with the
event that is the content of the testimony. For the Logic, on the other
hand, testimony is a kind of evidence distinctively connected to
theoretical authority, and so assessing the reliability of testimony is not

the same as assessing the reliability of other ordinary evidence.
Assessing the reliability of testimony involves determining whether the
speaker should be treated as an authority, as a person whose word it is
reasonable to trust.
In this respect, it shouldn't be a surprise that in arguing for the

credibility of some instances of testimony to the occurrence of miracles
the Logic goes out of its way to appeal to the very person of Augustine,
to his character and honesty as an individual. (p.35)
So the only remaining basis for skepticism would be to doubt the

testimony itself of St. Augustine, and to suppose that he altered
the truth to legitimize the Christian religion in the minds of the
pagans. No one can say this with the slightest plausibility.
First, it is not at all likely that a judicious person would have

wanted to lie about such things when he could have been
convicted of lying by countless witnesses, which could only have
brought disgrace on the Christian religion. Second, no one was a
greater enemy of lies than this Saint, especially in religious
matters, since he wrote entire books proving not only that it is
never permissible to lie, but that it is a horrible crime to do it
under the pretext of attracting people to the faith more easily.
(1996: 269)
One needn't think Augustine a liar to discount his testimony. One might
think that he was honestly mistaken about what he reported. And of
course one might have a very different opinion of the character of
Augustine than the writers of the Logic. The point is simply that the
Logic takes the kind of assessment of the reliability of a speaker
required when forming beliefs on the basis of testimony to involve a
very direct assessment of the person of the speaker. Given what we
know about the speaker, we must decide whether to treat this
particular person as an authority on these matters.
These kinds of personal assessments are largely lacking from Hume's

discussion. Instead, Hume offers several very general reasons for
thinking that no testimony, no matter how trustworthy the testifier, has
ever been sufficient to justify belief in miracles. First, there have never
been enough concurring trustworthy witnesses to a miracle (1975:
116). Second, belief in miracles can be easily explained by the way in
which the agreeable passions of pleasure and surprise often lead us to
believe things contrary to reason (1975: 117). Third, belief in
miraculous events is found chiefly among “ignorant and barbarous
nations” (1975: 119). Fourth, the miracles reported by different

religions all conflict with one another (1975: 121). And fifth, a great
many stories about the occurrence of miracles have proven false (1975:
126). Hume does consider particular instances of historical testimony
to the occurrence of miracles. He cites in particular a miracle
attributed to Vespasian that was reported by Tacitus. He also mentions
a more recent miracle reported by one Cardinal de Retz. Hume admits
that the persons testifying to these miracles appear to be trustworthy,
however he dismisses these reports nonetheless by appealing to the
very general kinds of considerations listed above.19
(p.36) In this respect, there appears to be a significant difference of

emphasis between the way in which Hume and the Logic conceive of
the assessment of the reliability of testimony to the occurrence of
miracles, a difference of emphasis directly connected to their very
different conceptions of the nature of belief based on testimony. The
Logic tends to construe the assessment of the reliability of testimony as
the assessment of a person. Judging a given instance of testimony
reliable involves judging the speaker to be an authority on the matter.
Believing the testimony then involves believing the speaker, what the
Logic calls human faith. In contrast, Hume tends to construe the
assessment of the reliability of testimony as the assessment of the
reliability of an event. A speaker's testimony is construed as an ordinary
empirical event, and the task for the audience is to assess the likelihood
that this speech event is correlated with the fact or event that is the
content of the testimony. Considerations of the trustworthiness of the
individual speaker certainly have a role to play here, but the role that
they play is one of increasing or decreasing the likelihood that the
particular speech event is a reliable guide to the truth. Judging a given
instance of testimony reliable thus involves judging that the speech
event is likely correlated with the truth, and believing the testimony
involves nothing more than believing in the content of the testimony on
the basis of an ordinary inductive inference from consideration of the
likelihood of this correlation.20
This difference in emphasis between conceiving of the assessment of

the reliability of testimony as the assessment of a person and as the
assessment of an event may help to explain why Hume and the Logic
draw such different conclusions concerning the rationality of belief in
miracles on the basis of testimony. The more that we see the
assessment of the reliability of testimony as a matter of determining
whether we can treat the speaker as an authority, whether we can put
ourselves in her hands and trust her for the truth, then the more we
might be inclined to take the authority of the speaker to override the
improbability of the content of her testimony. On the other hand, the

more that we see the assessment of the reliability of testimony as a
matter of determining whether the testimony considered as a mere
event is likely correlated with the truth, then the more we might be
inclined to take the inherent implausibility of the content of the
testimony to outweigh the likelihood that the speech act is correlated
with such a miraculous event. Of course, there is no direct connection
between these different ways of conceiving the assessment of the
reliability of testimony and particular verdicts on the question of
whether we should believe testimony to the occurrence of miracles.
One could agree with the Logic that assessing the reliability of
testimony is essentially a matter of assessing the trustworthiness of a
person and still conclude that we shouldn't believe in the occurrence of
miracles on the basis of testimony, and one (p.37) could agree with
Hume that assessing the reliability of testimony is essentially a matter
of assessing the reliability of an event and still conclude that we have
good reason to believe some testimony to the occurrence of miracles.
Nevertheless, it is plausible that the differences here have at least
some effect on the very different conclusions drawn by the Logic and by
Hume.
Putting things briefly and roughly, the Logic gives the category of other
persons much more epistemological significance than does Hume.
Hume takes testimony to be epistemologically significant. Again, he
holds that “there is no species of reasoning more common, more useful,
and even necessary to human life, than that which is derived from the
testimony of men, and the reports of eye-witnesses and
spectators” (1975: 111). But Hume takes the epistemological
significance of testimony to be that of a mere event, an event that may
or may not be a reliable indicator of the truth. For the Logic, testimony
has a very different epistemological significance that is distinctively
connected to theoretical authority, and this at least opens the door to
the possibility that the probative value of a particular instance of
human testimony may be able to outweigh the inherent improbability of
the content of the testimony. In this sense, what is ultimately at stake
between Hume and the Logic concerns much more than the particular
verdict that one should draw concerning the rationality of belief in the
occurrence of miracles on the basis of testimony. What is at stake
between Hume and the Logic concerns whether and to what extent the
category of other persons plays a distinctive role in human rationality.
1.4 Inference, Perception, and Sociality: Hume Versus Reid

The conception of the epistemology of testimony present in the Port

Royal Logic is seldom mentioned in contemporary discussions of Early
Modern debates about testimony. In a sense, this is unsurprising. As I
have already suggested, contemporary discussions of the epistemology
of testimony tend to be pursued in abstraction from any explicit
questions about the nature of theoretical authority. The Logic assigns
an ineliminable role to theoretical authority in the acquisition of
knowledge and belief based on testimony, and so it naturally falls
outside the purview of most contemporary discussions. Not only does
this neglect a significant portion of the history of philosophical
conceptions of testimony, but it also distorts much of what is at issue in
Early Modern debates about testimony. At stake between the very
different conceptions of the epistemology of testimony offered by the
Logic and by Locke and Hume is a commitment to epistemic autonomy.
Recognizing this has an important and interesting consequence. Not
only does it make sense of how figures like Locke can acknowledge the
tremendous degree to which we must rely on the testimony of others
while at the same time (p.38) railing against deference to authority, but
it also sheds new light on the influential epistemological dispute
between Hume and Reid.
The Early Modern debate concerning the epistemology of testimony is

typically cast as a debate between Hume's reductionism about
testimony and Reid's anti-reductionism about testimony. Hume is
typically said to reduce the capacity for learning from testimony to the
operation of other epistemic capacities, particularly the capacity for
inference, while Reid is said to deny the possibility of such a reduction.
For Reid, the capacity for learning from testimony is just as basic and
noninferential as the capacity for perception. In one sense, then, Reid
clearly rejects the conception of the epistemology of testimony
advanced by figures like Locke and Hume. In another, much less
recognized sense, however, the manner in which Reid rejects Hume's
reductionist conception of the epistemology of testimony is further
testament to the way in which epistemic autonomy was established as a
central ideal of the Enlightenment.
In An Inquiry into the Human Mind, Reid draws an analogy between

what he calls “original” and “acquired” perception and what he calls
“natural” and “artificial” language (1997: 190–195). For Reid, acquiring
knowledge on the basis of both perception and testimony is a matter of
the interpretation of signs, a process whereby the sign immediately and
noninferentially produces in an audience belief in the thing signified.
Some perception is original, the result of an innate interpretive
capacity, and some perception is acquired, the result of an interpretive
capacity acquired only through repeated experience and custom.

Analogously, Reid thinks that there are some human signs such as
features of the face, modulations of the voice, and movements of the
body that we have an innate capacity to interpret and other signs, those
of speech and writing, that are established by the will of men but that
through experience and custom we come to be able to interpret. Like
knowledge based on acquired perception, knowledge based on
testimony is the result of an audience's acquired capacity to
immediately and noninferentially pass from a sign, the speaker's
testimony that p, to belief in the thing signified, belief that p. Just as we
are naturally disposed to form perceptual beliefs on the basis of the
deliverances of the senses, so we are naturally disposed to form
testimonial beliefs on the basis of the testimony of others.
In order to explain how a speaker's testimony can immediately and

noninferentially produce in an audience belief in the content of the
testimony, Reid posits two fundamental principles of the human mind:
the principle of veracity and the principle of credulity (1997: 193–194).
According to the principle of veracity, we all have a natural tendency to
tell the truth—to sincerely express what we believe and to believe what
is true. According to the principle of credulity, we all have a
corresponding natural tendency to believe what others tell us—to
believe that they are sincere and to believe that what they say is true.
This disposition is tempered by our experience with instances of
insincerity in others, but it is due to the fact that these two principles
are part of the (p.39) natural constitution of the human mind that a
speaker's testimony that p can immediately and noninferentially
produce in an audience the belief that p.21
The accounts of the epistemology of testimony offered by Hume and

Reid thus appear to be diametrically opposed. Like Locke, Hume takes
testimonial knowledge to be a species of knowledge based on
nondemonstrative inference, while Reid takes testimonial knowledge to
be just as immediate and noninferential as knowledge based on
perception. Nevertheless, our discussion up to this point of the
historical development of philosophical conceptions of testimony helps
us to see that, for all that has been said so far, the accounts offered by
Hume and by Reid have one important thing in common: they are both
consistent with epistemic autonomy. Since Reid takes testimonial
knowledge to be the result of a natural disposition to noninferentially
form beliefs on the basis of testimony, analogous to the case of acquired
perception, it is not clear that this provides any substantial role for
theoretical authority. Just as being naturally disposed to form beliefs on
the basis of the deliverances of the senses doesn't involve a subject's
giving up any responsibility for the justification of her perceptual

beliefs, so being naturally disposed to form beliefs on the basis of
testimony needn't involve giving up to an authority any of one's own
epistemic responsibility. Inferring things for oneself and seeing things
for oneself are two paradigmatic ways in which one comes to one's own
conclusion about things, and if learning from testimony is analogous to
perception in this sense, then an audience can still be solely
epistemically responsible for the justification of its testimonial beliefs.
Contemporary debates about the epistemology of testimony have for

the most part taken their cue from this dispute between Hume and
Reid, and so it shouldn't be a surprise that in so doing they have largely
sidestepped questions about the nature and legitimacy of theoretical
authority. The contemporary epistemological debate between
reductionism about testimony and anti-reductionism about testimony
pitches those who seek to reduce the capacity for learning from
testimony to the operation of other epistemic capacities, particularly
the capacity for inference, against those who deny the possibility of
such a reduction, claiming instead that the capacity for learning from
testimony is just as basic and noninferential as the capacity for
perception. Not only are both positions in this debate consistent with
epistemic autonomy, the debate itself can be interpreted as largely
fueled by a commitment to epistemic autonomy. Epistemic autonomy
holds that taking something on authority, insofar as this is opposed to
coming to one's own conclusion about things, is not a genuine
epistemic capacity. Taking something on authority is not a cognitive
capacity the exercise of which can yield beliefs amounting to
knowledge. This then sets a project for epistemological accounts (p.40)
of testimony. If one thinks that knowledge and justified belief cannot be

acquired by taking things on authority, then vindicating learning from
testimony as a genuine epistemic capacity requires making out how one
can acquire knowledge and justified belief from the testimony of others
in such a way that one is still coming to one's own conclusion about
things. The epistemic capacities of perception and inference both
involve coming to one's own conclusion about things, so it is natural to
suppose that vindicating learning from testimony as a genuine
epistemic capacity will involve making out how it is similar to the
capacities of perception or inference. If knowledge and justified belief
can only be acquired by coming to one's own conclusion about things,
and if perception and inference are the paradigmatic ways in which one
comes to one's own conclusion about things, then vindicating testimony
as a genuine source of knowledge requires likening the capacity of
learning from testimony to the capacities of either perception or

inference.
This sets the framework in which contemporary debates about the

epistemology of testimony have typically been pursued. If one is
attracted to epistemic autonomy, then this framework is a natural one.
Importantly, however, it isn't clear that Reid himself is motivated in this
way. There are points in Reid's writing that are reminiscent of the Port
Royal Logic, points at which Reid explicitly contrasts individual reason
and authority. In the Inquiry he writes:
It is the intention of nature, that we should be carried in arms

before we are able to walk upon our own legs; and it is likewise
the intention of nature, that our belief should be guided by the
authority and reason of others, before it can be guided by our own
reason. The weakness of the infant, and the natural affection of
the mother, plainly indicate the former, and the natural credulity
of youth, and the authority of age, as plainly indicate the latter.
The infant, by proper nursing and care, acquires strength to walk
without support. Reason hath likewise her infancy, when she must
be carried in arms; then she leans entirely upon authority, by
natural instinct, as if she was conscious of her own weakness; and
without this support, she becomes vertiginous. When brought to
maturity by proper culture, she begins to feel her own strength,
and leans less upon the reason of others; she learns to suspect
testimony in some cases, and to disbelieve it in others; and sets
bounds to that authority which she was at first entirely subject.
But still, to the end of life, she finds a necessity of borrowing light
from testimony, where she has none within herself, and of leaning
in some degree upon the reason of others, where she is conscious
of her own imbecility. (1997: 195)
Importantly, what Reid says here can in fact be interpreted in such a

way as to make it consistent with epistemic autonomy. If “leaning on
authority” means nothing more (p.41) than being naturally disposed to
form noninferential beliefs on the basis of what one is told, analogous
to the way in which one is naturally disposed to form noninferential
beliefs on the basis of perceptual representation, then this looks to be
completely consistent with epistemic autonomy. Contemporary
reductionists and anti-reductionists about testimony both commonly
claim that testimonial knowledge is justified by trust or authority, but
typically they don't go so far as to give these concepts any substantial
epistemological weight. Ultimately, the epistemology of testimony is
likened to that of perception or inductive inference, to the operation of
epistemic capacities that paradigmatically involve a subject's coming to

her own conclusion about things.
Reid, however, clearly holds that even though the capacities of learning
from testimony and acquired perception are analogous in virtue of
being noninferential, they are nevertheless disanalogous in other
important respects, and this may open the door to providing a more
substantive role for theoretical authority. In the Essays on the
Intellectual Powers of Man, Reid claims that testimonial knowledge and
belief is disanalogous to knowledge and belief based on perception in
that it is the result of what he calls a social operation of mind.
Some operations of our mind, from their very nature, are social,
others are solitary.
By the first, I understand such operations as necessarily suppose

an intercourse with some other intelligent being. A man may
understand and will; he may apprehend, and judge, and reason,
though he should himself know of no intelligent being in the world
besides himself. But, when he asks information, or receives it;
when he bears testimony, or receives the testimony of another;
when he asks a favour, or accepts one; when he gives a command
to his servant, or receives one from a superior: when he plights
his faith in a promise or contract; these are acts of social
intercourse between intelligent beings, and can have no place in
solitude. They suppose understanding and will; but they suppose
something more, which is neither understanding nor will; that is,
society with other intelligent beings. They may be called
intellectual, because they can only be in intellectual beings: But
they are neither simple apprehension, nor judgment, nor
reasoning, nor are they any combination of these operations.
(2002: 68)
Here Reid claims that the social operations of mind—like giving and
receiving testimony, giving and receiving promises, giving and
receiving commands, and asking and receiving favors—are mental
capacities that “necessarily suppose an intercourse with some other
intelligent being.” This is supposed to distinguish these capacities from
the capacities for apprehension, judgment, and reasoning. Pretty
clearly, then, this is supposed to distinguish the capacity of learning
from testimony from the capacities of both inference and perception.
(p.42) For Reid, the social operations of mind are just as much a part of
our natural constitution as the solitary operations, and the social
operations cannot be reduced to or explained in terms of the solitary
operations (2002: 69). The social operations of mind are irreducibly

social, meaning that the social character of these social operations is
not something that can be reduced to some combination of the
operation of the solitary capacities. Unfortunately, however, he doesn't
have much to say about the precise sense in which these social
operations are irreducibly social. There are in fact a variety of ways in
which the solitary capacities for inference and perception may require
“an intercourse with some other intelligent being.” As twentieth-
century developments in semantics and philosophy of mind have shown,
the reference of our terms and the very content of our representational
mental states may depend heavily on what Hillary Putnam (1975) has
called a “linguistic division of labor.” If this is true, then the
determination of the content of our beliefs may be said to be “social” in
the sense of being constitutively dependent on our linguistic community
(Burge 1979). Additionally, it is plausible that what we are in a position
to perceive or infer from evidence often depends heavily on the society
in which we have been raised. The kinds of things of which we are
perceptually aware and the kinds of inductive generalizations that we
draw are plausibly affected by what we have been taught and by the
kinds of awareness and inference that we have experienced in our
fellows. Surely, if we had no intercourse with other intelligent beings,
our beliefs based on perception and inference would be very different.
If the kind of sociality characteristic of learning from testimony is

supposed to distinguish learning from testimony from the capacities for
inference and perception, then it must be different from any of the
kinds of sociality mentioned above. Coady has claimed that what
distinguishes the social operations of mind are that they are essentially
“interactive” (2004: 192), and I think that this is on the right track. It is
important to recognize that Reid is here concerned with social
operations of mind—with social mental capacities, powers, or abilities—
meaning that the sociality he has in mind pertains to the actual exercise
of the mental capacities themselves, not merely to the output of these
capacities. In this sense, Reid holds that it is the capacity for learning
from testimony itself that is irreducibly social, not merely the beliefs
that result from the exercise of this capacity. Learning from testimony
is irreducibly social in that the very exercise of the capacity is somehow
interactive or cooperative and so couldn't take place without
intercourse with other intelligent beings.
Reid doesn't have much to say about the precise sense in which the
exercise of social capacities is interactive or cooperative, but he does
make an intriguing reference to issues concerning what we might call

the second person.
In all languages, the second person of verbs, the pronoun of the

second person, and the vocative case in nouns, are appropriated
to the expression of social operations (p.43) of mind, and could
never have had place in language but for this purpose: Nor is it a
good argument against this observation, that, by a rhetorical
figure, we sometimes address persons that are absent, or even
inanimated beings, in the second person. For it ought to be
remembered, that all figurative ways of using words or phrases
suppose a natural and literal meaning of them. (2002: 70)
Reid here claims that, when expressed in language, the social

operations of mind often utilize second-personal grammatical devices.
For example, the speech acts of commanding, promising, and telling
are often performed by using grammatical constructions involving the
second person pronoun—“I command you to fall in line,” “I promise you
that I will pick up the kids,” “I tell you: the weather here has sure been
hot lately.” Second-personal grammatical devices are markers of
address. The second person pronoun refers to the addressee of the
utterance in which it appears. If the social operations of mind typically
utilize these markers of address when they are expressed in language,
then this might be reason to believe that the operation of these
capacities itself requires an addresser and an addressee. Promises,
commands, testimony, and requests all must be addressed to an
audience, and they all call for their audience to recognize them as
being so addressed. The sense in which the social operations of mind
are interactive or cooperative might thus have to do with the way in
which they necessarily require not only two individuals interacting but
two individuals interacting in a particular way, as addresser to
addressee.22
If this is right, then it much more radically distinguishes Reid's view of

testimony from that of Locke and Hume. Recall that both Locke and
Hume take testimonial belief to be the result of a very general capacity
for broadly nondemonstrative inference. In this sense, there is no
distinctive cognitive capacity for learning from testimony. Learning
from testimony is reduced to the operation of other epistemic
capacities. Reid holds that learning from testimony is a distinctive
cognitive capacity very different from the capacity for inference, but
the discussion of the social operations of mind distinguishes learning
from testimony from the capacity for acquired perception as well. Not
only is the capacity of learning from testimony noninferential, as the
analogy with acquired perception demonstrates, but it is also
irreducibly social. And it is social in the sense that the exercise of the
capacity requires two individuals relating to one another as addresser
to addressee. In this sense, we might call learning from testimony a
second-personal epistemic capacity.
Reid doesn't squarely address the issue that is at stake between Hume
and the Port Royal Logic, the issue of epistemic autonomy.
Nevertheless, his account of the social (p.44) character of learning
from testimony (and in particular his second-personal characterization
of the sociality at issue) does raise the question of the relation between
the social nature of learning from testimony and the issue of theoretical
authority. Unfortunately, this question has received almost no attention
in contemporary debates concerning the epistemology of testimony. For
the most part, contemporary debates have taken their cue from the
dispute between Hume and Reid, but in so doing they have construed
this dispute as one centering on the question of whether the capacity
for learning from testimony is inferential. Not only does this sidestep
Reid's own concern with the social character of learning from
testimony, it also completely ignores the issue of theoretical authority.
In the next chapter I will argue that the standard contemporary
accounts of the epistemology of testimony fail for precisely this reason.
Notes:
(1) In this book I will use the terms ‘speaker’ and ‘audience’ to refer to
the roles of disseminator and recipient of testimonial knowledge and
belief. I thus intend these terms to apply to both oral and written
testimony. Also, in order to prevent pronoun confusion, I will typically
use the feminine ‘she’ to refer to the role of speaker and the neuter ‘it’
to refer to the role of audience.
(2) This characterization of the speech act relevant to the acquisition of

testimonial knowledge is not uncontroversial. As we'll see in chapter 3,
Lackey (2006c) holds that the speech act relevant to the acquisition of
testimonial knowledge should be construed as the much broader
category of ordinary statements.
(3) I discuss Reid's view of the epistemology of testimony in section 1.4

below.
(4) I return to the relation between theoretical and practical authority

in chapter 5.
(5) It is unclear why Socrates presents the case as one in which there
are no eyewitnesses. If there were eyewitnesses to the crime, would
this mean that the jury could indeed acquire knowledge?
(6) No doubt, this connection between nonreligious and religious

testimony, or as the Port Royal Logic puts it, between human faith and
divine faith, is part of what worries Enlightenment opponents of
theoretical authority.
(7) I have here focused on one very general way in which historical
conceptions of the general nature of knowledge served to rule out the
possibility of knowledge based on testimony. Knowledge conceived in
terms of episteme or scientia was taken to involve a kind of systematic
understanding that plausibly isn't a feature of ordinary testimonial
belief. There are, however, other perhaps allied ways in which the
category of knowledge was historically conceived that also served to
exclude belief based on testimony. For example, the idea that
knowledge requires a kind of personal acquaintance with the facts,
something registered by the German word ‘kennen,’ also plausibly
excludes testimonial belief. Descartes seems to express something like
this idea in Part VI of the Discourse where he writes “no one can
conceive something so well, and make it his own, when he learns it
from someone else as when he discovers it himself” (1985: 146). In a
similar spirit, Marcel Proust writes in In Search of Lost Time, “But
things of which we have not had a direct intuition, which we have
learned only from other people, are such that we no longer have the
means, we have missed the chance, of conveying them to our inmost
soul; its communications with the real are blocked and so we cannot
profit by the discovery, it is too late” (1999: 936). Here again, beliefs
based on testimony may be perfectly reasonable and legitimate, but
they do not involve the kind of personal acquaintance with the facts
required for genuine knowledge.
(8) In section 1.3 I argue that this difference between the way in which
Hume and the Logic conceive of knowledge and belief based on
testimony goes some distance towards explaining why they draw such
very different conclusions concerning the rationality of belief in
miracles on the basis of testimony.
(9) In chapter 2 I argue that the distinction between testimony and

argument is crucial for understanding the epistemology of testimony.
(10) A similar sentiment motivates the philosophical anarchism of the

eighteenth-century novelist and political theorist William Godwin. See
especially Chapter VI of Book II of his Enquiry Concerning Political
Justice (1971).
(11) I have chosen to use the term ‘autonomy’ rather than terms like
‘individualism’ or ‘egoism’ in order to highlight the connection between
this thesis and worries about authority. Political philosophers often
worry about an apparent conflict between authority and autonomy—
see, for example, Wolff (1990) and Raz (1979)—and I think the same
general worry is at play here, though cast in an epistemological
register. Nevertheless, I don't want to build too much into my use of the
term. I take it that the thesis I am concerned with here roughly
corresponds to what Zagzebski (2007) calls “weak epistemic egoism.”
(12) I am here deliberately leaving to one side general epistemological

questions about the relationship between epistemic responsibility,
knowledge, and justification. I feel that the claims made in the text are
intuitive enough to stand on their own and thus that any general
account of the relationship between these concepts will have to be in a
position to make sense of these claims.
(13) Fricker (2006) discusses an ideal of epistemic autonomy, though

her formulation of this ideal is much stronger than the position I am
concerned with here. Fricker's ideal of the autonomous knower is of
one who “relies on no one else for any of her knowledge” (2006: 225). It
is an important feature of the position I am calling epistemic autonomy,
however, that it is consistent with epistemic reliance on others.
Epistemic reliance on others is consistent with epistemic autonomy as
long as, in so relying, the subject is solely epistemically responsible for
the justification of her beliefs.
(14) We can begin to see here the way in which commitment to

epistemic autonomy can serve to empty notions like trust and authority
of their theoretical content. This is common in contemporary
discussions of the epistemology of testimony. Though testimonial belief
is often taken to be “based on authority” and to involve “trusting
others,” commitment to epistemic autonomy requires that these notions
cannot be given any distinctive explanatory role.
(15) This new conception of testimony is clearly displayed in the work

of classical probability theorists like Bernoulli, Condorcet, Laplace, and
Poisson who sought to quantitatively apply the calculus of chances to
eye-witness testimony in an effort to formalize rules for the rational
acceptance of testimony. As Daston notes, “By the turn of the
nineteenth century, the probability of testimony was featured in texts
and treatises on the calculus of probabilities as a standard application,
along with the probabilities of games of chance, annuities, and the
probability of causes” (1988: 339).
(16) To be clear, I will not here be concerned with the particular verdict
that one draws concerning the rationality of belief in the occurrence of
miracles on the basis of testimony. I am interested only in the
conception of testimony and of the assessment of the reliability of
testimony at work in Hume's argument.
(17) For a detailed recent account of Hume's argument, see Fogelin

(2003).
(18) Locke also holds that belief in the occurrence of miracles on the
basis of testimony is warranted despite the inherent improbability of
miraculous events. In Locke's case, however, this doesn't have anything
to do with his having a conception of belief based on testimony that
differs from Hume's. Locke's account of belief based on testimony is in
all essential respects identical to Hume's. Nevertheless, he seems to
hold that testimony to the occurrence of miracles is the one case where
the inherent implausibility of the fact or event testified to actually
contributes to the strength of the testimony (1975: 667).
(19) Hume applauds Cardinal de Retz for not believing the story he
himself relates:He considered justly, that it was not requisite, in order
to reject a fact of this nature, to be able accurately to disprove the
testimony, and to trace its falsehood, through all the circumstances of
knavery and credulity which produced it. He knew, that, as this was
commonly altogether impossible at any small distance of time and
place; so was it extremely difficult, even when one was immediately
present, by reason of the bigotry, ignorance, cunning, and roguery of a
great part of mankind. He therefore concluded, like a just reasoner,
that such an evidence carried falsehood upon the very face of it, and
that a miracle, supported by any human testimony, was more properly a
subject of derision than of argument. (1975: 124)
(20) For more on the difference between construing our epistemic

reliance on the testimony of others as dependence on a person and as
dependence on an event, see chapter 3.
(21) For a good discussion of some of the complications concerning

Reid's position here, see Van Cleve (2006).
(22) I discuss the relevance of address to the epistemology of testimony

in much more detail in chapters 2 and 3.
Knowing at Second Hand

Benjamin McMyler

Print ISBN-13: 9780199794331
DOI: 10.1093/acprof:oso/9780199794331.001.0001

Benjamin McMyler
DOI:10.1093/acprof:oso/9780199794331.003.0003

Both reductionists and anti-reductionists about testimony commonly
describe knowledge acquired from testimony as knowledge acquired at
second hand. This chapter argues that appreciating the distinctive
sense in which testimonial knowledge is secondhand supports anti-
reductionism over reductionism but also that it supports a particular
kind of anti-reductionism very different from that typically offered in
the literature. Testimonial knowledge is secondhand in the demanding
sense of being justified by the authority of a speaker, where this serves
to parcel out epistemic responsibility for the audience's belief between
speaker and audience. The epistemic credentials of testimonial
knowledge are in this sense importantly interpersonal.
Keywords: testimony, authority, epistemic responsibility, deference, epistemic buck-

passing, address, the second person
Epistemologists have traditionally sought to understand empirical

knowledge on the model of either perception or inference. The
paradigmatic models according to which a subject acquires knowledge
about the world around her have traditionally been taken to consist
either in the direct experience of the facts or in the inference from such
experiences. These paradigmatic models have had a powerful influence
on the way in which epistemologists have thought about knowledge
acquired from testimony. Recent debates about the epistemology of

testimony have pitched those who, in the spirit of Locke and Hume,
advocate the reduction of the capacity of learning from testimony to the
operation of other epistemic capacities (particularly the capacity for
inference) against those who, in the spirit of Reid, seek to vindicate
learning from testimony as an epistemic capacity just as irreducible and
“basic” as perception. The guiding question on both sides of this debate
is one of what exactly constitutes the justification relevant to
testimonial knowledge and belief. For some this justification must look
like that typical of knowledge and belief acquired from inference, while
for others it must look more akin to the basic, noninferential kind of
justification relevant to perception.
In this chapter I want to offer an alternative account of the

epistemology of testimony, and I propose to do so through examining a
feature of testimonial knowledge that has been underappreciated in
recent discussions. Despite their differences, parties on both sides of
the aforementioned debate commonly describe testimonial knowledge
as knowledge (p.46) acquired at second hand.1 Still, little thought has
been given to what exactly it means to describe testimonial knowledge
in this way. At the very least, to say of an item of knowledge that it is
secondhand is to say that it has been somehow inherited or gained from
another. If I know something at second hand, then another person, a
speaker, has played an epistemologically significant role in mediating
my epistemic relation to the facts. But what exactly is this role of the
speaker, and how does it serve to make testimonial knowledge
secondhand?
In the course of this chapter, I want to explore what we might call the
secondhandness of testimonial knowledge. Again, this aspect of the
epistemology of testimony has been largely ignored in contemporary
discussions. While the contemporary debate has focused largely on the
kind of justification we have in knowing based on testimony, the
question of what distinguishes secondhand from firsthand knowledge
has received little explicit attention, remaining largely a casualty of the
focus on justification.
We can begin to see how this is so by considering the following two

situations:
(1) I am having car trouble and just happen to have a close

friend who is a mechanic. I ask her to look my car over, and after
a close inspection she tells me that I need a new transmission. I
later relate this fact to a third party who asks me how I know. I
respond, “A friend of mine who happens to be a mechanic told

me so.”
(2) As I am walking down the street I run into a colleague of
mine whom I don't know very well. I ask her how she is doing,
and she tells me that she has been working day and night trying
to meet a paper deadline. Later that day I hear someone remark
that the aforementioned colleague hasn't been around much
lately, and I interject that she has been working on a paper.
When asked how it is that I know this, I respond, “She told me
so.”
Both of these situations involve typical cases of knowledge by

testimony.2 Assuming that what the friend and colleague say is true, by
telling me they allow me to gain (p.47) knowledge of the content of
their respective assertions. If it is true that my car needs a new
transmission, and if it is true that my colleague has been working on a
paper, then their telling me so (somehow) puts me in a position to know
this. Still, the two situations might elicit different intuitions about what
exactly is involved in knowing by testimony. In the first situation, my
knowing that my car needs a new transmission depends heavily on my
knowing that the person telling me is in a position to know this kind of
thing (she is a mechanic) and my knowing that she is sincere (she is a
friend, not someone looking to make a profit off of my ignorance about
automobiles). It is only on the strength of my rational responsiveness to
such considerations about the speaker that I can properly be said to
know based on her testimony. In the second situation, however,
considerations about the speaker's competence and sincerity do not
seem to play as integral a role. I don't know my colleague that well, and
though it wouldn't make much sense for her to be mistaken about what
she has been doing (which isn't to say that this is impossible), I don't
have any special grounds for ruling out the possibility that she is lying
to me. Still, it seems plausible that if what she tells me is true then her
telling me puts me in a position to know it.
Such intuitions are at the heart of contemporary debates about the

epistemology of testimony. The main parties in the debate tend to fall
into two camps depending on whether their guiding intuitions are taken
from the first type of situation above or from the second. Those who
take their guiding intuitions from the first type of case are typically
called reductionists about testimony, while those who take their guiding
intuitions from the second type of case are typically called anti-
reductionists. Reductionists about testimony hold that there is nothing
about the features of testimony itself that serves to justify an audience's
knowledge based on testimony.3 Instead, an audience's justification in
knowing based on testimony must be explained in terms of the

operation of other epistemic capacities. As in the case of the mechanic
above, the audience must have independently available reasons for
believing the word of a speaker, reasons that typically consist of beliefs
concerning the sincerity and competence of the speaker that are
gathered independently of the particular piece of testimony at issue. An
audience's (p.48) justification in knowing by testimony then consists in
an inference from these independently available considerations to the
conclusion that things actually are as the speaker says. In the case of
the mechanic, I have independent, prior knowledge that the mechanic
is competent when it comes to diagnosing automotive problems, and I
have independent knowledge that she is acting in every way sincere.
This knowledge is stored and recalled through memory, and it then
functions as a premise in an inference to the conclusion that what the
speaker says (that my car needs a new transmission) is true. Though
the operation of the capacities for perception and memory may thus be
required to yield the premises from which the truth of what the speaker
says is inferred, it is the capacity for inference that does the real
epistemological work of justifying an audience's testimonial belief. In
this sense, reductionism about testimony is committed to the thesis that
(R) An audience's justification in knowing based on testimony

consists in the strength of an inference from a speaker's saying
that p, through independently available considerations concerning
the speaker's trustworthiness, to the conclusion that p.4
As Elizabeth Fricker puts it, testimonial knowledge “is inferential in the

sense that it requires a substantial structure of justification to support
it” (1987: 74). According to Fricker, one can hold that testimonial
knowledge is inferential in this respect without requiring that the
audience's belief be arrived at by an explicit act of conscious inference.
The inferential mechanism at work may function below the level of the
subject's conscious awareness (Fricker 1995: 405). The reductionist
claim is simply that the audience's justification in believing what the
speaker says is constituted by the strength of the inferential connection
specified in (R).
Anti-reductionists about testimony hold that the justification of an

audience's knowledge based on testimony cannot be cashed out in
terms of independent observation and inference.5 Anti-reductionists
thus appeal to situations like that of the colleague above in which our
purchase on independently available considerations of the speaker's
trustworthiness seems insufficient to justify a belief amounting to
knowledge. Since we can still acquire knowledge by testimony in such
situations, claims the anti-reductionist, (p.49) testimonial knowledge
cannot be reduced to independent observation and inference. In this

sense, anti-reductionism about testimony is committed to the thesis
that (R) is false. In at least some cases, the justification of an audience's
testimonial knowledge does not consist in the strength of an inference
from the speaker's saying that p, through independently available
considerations of the speaker's trustworthiness, to the conclusion that
p. But if there are cases in which (R) is false, then there must be
something about the features of testimony itself that serves to justify an
audience's testimonial knowledge in these cases—testimony must be a
basic, autonomous, or sui generis source of knowledge.6 Of course, it is
important in knowing by testimony that an audience has no positive
reason to doubt what the speaker says, but absent such defeating
conditions the features of testimony itself serve (somehow) to justify
the audience's belief.
While reductionism, through its commitment to (R), offers a relatively

robust account of what it is that justifies testimonial knowledge and
belief, namely the strength of an inference, it is less clear what justifies
testimonial knowledge and belief on an anti-reductionist account. If an
audience's testimonial knowledge is not justified by the strength of an
inference, then it must somehow be justified by the features of
testimony itself. There must be something about the activity of giving
and receiving testimony that is capable of producing knowledge and
justified belief in an audience all on its own. Most anti-reductionists
claim that what justifies an audience's belief based on testimony is the
audience's understanding or comprehension of the force and content of
the speaker's testimony. As Peter Graham puts it, “Comprehension as
such, like perceptual representation, confers non-inferential
justification on belief” (2006: 84). In understanding the force and
content of a speaker's testimony, an audience is provided with a prima
facie justification for belief in the content of the testimony. This prima
facie justification can be defeated by evidence of the incompetence or
insincerity of the speaker, but absent such defeaters it can justify an
audience's testimonial knowledge all on its own.
The central element of the dispute between reductionism and anti-

reductionism about testimony thus concerns how to characterize the
justification we have for knowing things on the say-so of others.
Reductionists hold that the only way to be justified in knowing based on
testimony is to be in possession of an argument. In knowing that (p.50)
p based on a speaker S's assertion that p, an audience's justification

consists in the cogency of an argument from “S asserted that p” to “p.”
Anti-reductionists hold that our justification is not inferential—it does
not consist in the cogency of an argument from “S asserted that p” to

“p”—and therefore the features of testimony itself must (somehow)
serve to justify an audience's testimonial knowledge.7
Importantly for our purposes, reductionists take the role that inference
plays in the justification of testimonial knowledge to be what explains
its peculiar secondhandness. In fact, the intuitive difference between
knowing at first and at second hand is sometimes used by reductionists
as an argument for (R).8 What it means for testimonial knowledge to be
distinctively secondhand, according to reductionists, is that it is passed
on to us by a speaker in such a way that our justification in knowing
based on the speaker's testimony requires an inferential appeal to
independently available considerations about the speaker.
Reductionism thus gets an account of secondhandness for free from its
inferential conception of the justification required for testimonial
knowledge.
(p.51) Something similar is true of anti-reductionism, though the anti-

reductionist account of the way in which testimonial knowledge is
mediated is generally less clear. In defending testimony as a “basic” or
“autonomous” source of knowledge, one that need not be supported by
an inferential structure of justification, anti-reductionism stresses the
similarities between testimony and other basic sources of knowledge
(typically perception), this largely to the neglect of what it is that
makes testimony a distinctive source of knowledge and belief. Anti-
reductionists think that testimonial knowledge is clearly different from
knowledge based on perception. Knowledge noninferentially justified by
comprehension of a speaker's testimony doesn't involve a direct
experience of the facts, and hence there must still be a sense in which
testimonial knowledge is mediated by the speaker's testimony. Perhaps
we can say that testimonial knowledge is mediated in virtue of being
comprehension-based. Still, since anti-reductionists are first and
foremost concerned to stress the analogies between testimony and
perception, the sense in which comprehension explains the
secondhandness of testimonial knowledge remains obscure. Perception
is, after all, a paradigmatic form of firsthand knowledge, and so any
adequate anti-reductionist account of testimony must have something
substantial to say about the distinctive way in which testimonial
knowledge is mediated.
In this chapter I want to explore in much more detail the peculiar sense
in which testimonial knowledge and belief is mediated in virtue of being
secondhand. This will allow me to argue that the reductionist claim (R)
is in fact inconsistent with the secondhandness of testimonial
knowledge. Since I will be arguing that reductionism is mistaken, I will
be defending a broadly anti-reductionist view of testimony. However,

the account I will be defending is very different from the anti-
reductionist views typically found in the literature.9 My account will be
in the unique position of not only providing a substantive conception of
the secondhandness of testimonial knowledge, but also of formulating
its rejection of reductionism in terms of this conception and thereby in
terms of the distinctively mediated character of testimonial knowledge
and belief. Understanding the distinctive way in which testimonial
knowledge is secondhand requires shifting our attention to the
relations of authority and responsibility that exist between a testimonial
speaker and audience. I hope to show that the really significant
(p.52) epistemological problem of testimony concerns whether and in
what way the category of other persons plays an irreducible role in the
processes of belief formation and knowledge acquisition.10
Section 2.1 argues that the proper object of explanation for the
epistemology of testimony is knowledge that is secondhand in the
demanding sense of being based on or justified by the authority of the
speaker. Sections 2.2 and 2.3 outline the peculiar features of epistemic
authority and responsibility that characterize the secondhandness of
such knowledge. When an audience's belief is justified by the authority
of a speaker, the audience is entitled to defer responsibility for meeting
certain epistemic challenges to its belief, to pass the epistemic buck
back to the testimonial speaker. Section 2.4 argues that reductionism is
inconsistent with this feature of testimonial belief. The reductionist
holds that testimonial knowledge and belief is justified by the strength
of an inference, but an epistemic agent is solely epistemically
responsible for the inferences that she makes from within her own
epistemic position. Therefore, reductionism is inconsistent with a
testimonial audience being entitled to defer epistemic responsibility for
her testimonial belief back to the testimonial speaker. In short,
reductionism construes testimonial knowers as being in the position of
a detective deducing a conclusion or a doctor framing a diagnosis, but
this is inconsistent with testimonial knowers being entitled to pass
justificatory responsibility for their testimonial beliefs back to the
testimonial speaker. Section 2.5 considers a very general response to
this argument from secondhandness. It is open to the reductionist to
respond to this argument by denying that testimonial knowledge and
belief involves an epistemic entitlement to pass the epistemic buck.
Section 2.5 argues that this response is intuitively implausible but also
that there is a deep philosophical worry lurking behind this response, a
worry concerning whether relations of authority between persons
should be allowed to play an irreducible role in cognition. The final
three chapters of the book work from a variety of directions to try to

alleviate this worry.
2.1 What Is Testimonial Knowledge?

The term ‘testimony’ is conventionally understood by philosophers to
refer broadly to something like ordinary everyday informative or
purportedly informative statements. In this broad or “natural” sense,
the class of testimony includes far more than the (p.53) “formal”
testimony offered in the context of a court of law or governmental
hearing.11 In its broad sense, testimony encompasses the general kind
of giving of information from one person (or group) to another that is
apt for inclusion alongside perception, memory, and inference as one of
the typical ways in which knowledge and justified belief can be
acquired. Testimonial knowledge is then itself construed broadly as the
particular kind of knowledge acquired from testimony understood in
this broad sense.
This is certainly an intuitively compelling way of carving up the

philosophical landscape. Still, epistemological accounts of testimony
are often unfortunately vague concerning the class of knowledge that
they are out to explain. Specifically, epistemologists typically endorse
conceptions of testimony and of testimonial knowledge that are just too
broad to capture what is epistemologically significant about that form
of knowledge apt to be classed alongside knowledge gained from
perception, memory, and inference.
Epistemologists typically portray paradigmatic instances of testimonial

knowledge as being acquired from a particular type of speech act.
Fricker characterizes this central type of speech act as “serious
assertions aimed at communication” (1994: 137). Similarly, Robert Audi
claims that the epistemological problems of testimony pertain to
“roughly, saying or affirming something in an attempt to convey
(correct) information” (1997: 405). Such characterizations are quite
broad. Nevertheless, they outline a class of speech acts narrower than
that of mere declarative statements. On these characterizations, the
statement must be performed with a particular intention—the intention
to “communicate” or “convey information” to an audience—if it is to be
one from which testimonial knowledge can be acquired.12 If a speaker
is merely guessing, supposing, or bullshitting, then an audience cannot
gain testimonial knowledge from what she says. (p.54) Of course, an
audience may gain knowledge from a speaker's guessing, supposing, or
bullshitting—it may even gain knowledge of the content of what is said
if it has independent reason to believe that what is here being guessed,
supposed, or bullshitted is reliably correlated with the truth—but such
knowledge isn't properly testimonial.13 At the very least, testimonial
knowledge is knowledge concerning the content of a speaker's
assertion based on the speaker's openly attempting to communicate or

convey this information to the audience.
Even so, there is reason to believe that the class of serious assertions
aimed at communicating or conveying information is not itself a unitary
epistemic category. The knowledge that an audience gains from a
speaker's arguing is very different from the knowledge that an
audience gains from a speaker's telling, though both of these activities
involve making serious assertions aimed at communicating or
conveying information. Both arguing and telling are speech acts from
which an audience can gain knowledge of the content of what is said
based on the speaker's attempting to convey this information to the
audience, but intuitively, the epistemic credentials of knowledge based
on a speaker's arguing look very different from the epistemic
credentials of knowledge based on a speaker's telling. When a speaker
argues that p, I may thereby come to know that p, but it looks like in
such a case I have reached my own conclusion that p. The speaker's
argument may have convinced me that p, but my justification is not (or
at least should not be) that I have heard it from the speaker. My
justification for believing that p is that I have come to my own
conclusion about it, and if I am asked how I know that p, I will rehearse
the argument; the ground I appeal to will not be the speaker's having
told me so. This is precisely the purpose of argumentation. In
paradigmatic cases, providing argument is a means of helping the
audience to see things for itself. Knowledge based on a speaker's
telling, however, looks very different. When I am told by a speaker that
p, I may thereby come to know that p, but in such a case I don't seem to
have come to my own conclusion about p. This is because knowledge
based on tellings, unlike knowledge gained from argumentation,
involves the citing of an authority. If I come to know something based
on a speaker's telling, and if someone (p.55) asks me how I know, I will
cite the authority—“So and so told me so.” Knowledge based on a
speaker's telling, knowledge justified by citing an authority, thus
exhibits a kind of dependence on the speaker that knowledge based on
arguing does not. This distinction is of course only intuitive, but it is
incredibly natural; and in the course of this chapter I will explore some
of its implications for the epistemology of testimony.14
Before we move on, there are several points concerning the distinction
between knowledge based on a speaker's arguing and knowledge based
on a speaker's telling that merit discussion. First, the validity of the
distinction doesn't require denying that in the course of making an
argument a speaker may very well tell various things. That is, coming
to believe what has been argued may involve taking various things on
the authority of the speaker. In arguing that taxes need to be raised, a

politician may cite various figures purporting to show that certain
public programs are underfunded, and most of the audience, not being
in a position to check these figures firsthand, will have to believe them
(if they do) on the politician's authority. Nevertheless, the point of the
speech act of arguing is that the audience come to believe the
conclusion of the argument, here that taxes need to be raised, on the
basis of its own understanding of the argument rather than on the basis
of the authority of the speaker.
Second, a speaker may use argument to “back up” something she has
told where the argument is not intended to allow the audience fully to
see things for itself but rather to reassure it that it can indeed take
what the speaker has said on her authority. I might tell someone who
has never been to Chicago that Chicago is a great place to live and then
proceed to give a number of reasons why this is the case. Whatever the
strength of the argument I can muster in support of this conclusion, my
audience's coming to know that Chicago is a great place to live will still
involve an ineliminable appeal to my authority on the matter. My use of
argument is intended to assure the audience that I am authoritative
about this issue (that I can be trusted), and in this sense my audience's
justification in knowing that Chicago is a great place to live still seems
to involve an appeal to my authority on the matter.
Third, there seem to be cases in which an audience may accept a

conclusion based on a speaker's argument but accept the argument
based on the authority of the speaker rather than on any real
understanding of the argument. I may accept the conclusion of a proof
offered by a mathematician based on the mathematician's ability to
produce the proof but where (unfortunately) I am simply unable to
follow the proof myself. If (p.56) the proof is sound, I think that I can
legitimately be said to have knowledge of the conclusion of the proof
despite my inability to follow the proof, but this is because my belief in
the soundness of the proof is based on the authority of the speaker.15
Fourth, there may also be cases in which we believe a proposition

partly on the authority of a speaker and partly on the basis of a
speaker's argument.16 This may in fact characterize many cases of
learning in which a student first believes what her teacher says based
on the teacher's authority and then gradually transitions to believing
the same thing on the basis of arguments that the teacher presents.
This is often important in educational contexts insofar as the goal of
education is often not simply knowledge but understanding. At any
point on the road to understanding, the student's belief might be said to
be based partly on authority and partly on argument. This may also
characterize cases similar to that of the mathematician above in which

an audience believes what the mathematician says despite being able to
follow all of her arguments. The audience may be able to follow some of
her arguments, and her belief might be partly based on her own
understanding of those arguments, but her belief might still be partly
based on the authority of the mathematician.
Fifth, knowledge based on a speaker's telling need not involve being

told explicitly. Seriously communicating or conveying the information
that p need not involve explicitly saying that p. A speaker may, for
whatever reason, imply that it is raining outside without coming out
and saying that it is raining outside, and the audience may on this basis
come to know that it is raining outside in such a way that the
justification of its knowledge involves an appeal to the authority of the
speaker.17 Since for my purposes speech acts are to be individuated
epistemically, such a speaker can still be considered to have performed
the speech act of telling.18
(p.57) Sixth, much knowledge based on the authority of others isn't

justified by the citing of a particular authority, but rather by the fact
that we've simply heard it somewhere, from someone. I know that the
Aya Sofya in Istanbul was built by the Roman Emperor Justinian, but I
don't know where exactly I heard or read this. Such knowledge looks to
be based on appeal to authority, even though there is no particular
authority that I am in a position to cite. Relatedly, it seems that
communities, both scientific and otherwise, can function as epistemic
authorities just as well as individuals. A person's knowledge of a
particular scientific theory can thus be based on the authority of a
scientific community such that, if asked, the subject will appeal to the
authority of the community. Such knowledge is like knowledge based on
a speaker's telling in that it is justified by appeal to authority, but the
authority appealed to is somehow communal.19
Finally, there are vast realms of knowledge that are acquired from the
assertoric speech of others but whose epistemic credentials don't
appear to be constituted either by argument or by authority—for
example, basic linguistic and conceptual knowledge (e.g., that triangles
have three sides) and general nonobservational knowledge of the world
around us (e.g., that the earth existed long before I was born). Both
these types of knowledge are typical of those discussed by Wittgenstein
in On Certainty, and though they may be dependent on speakers or
communities of speakers to varying degrees, the fact that their
justification doesn't seem to involve an appeal either to argument or to
authority suggests that they require an epistemological treatment

separate from that of clear cases of either arguing or telling.
These qualifications notwithstanding, there is an intuitive distinction

between clear cases of knowledge based on a speaker's arguing and
clear cases of knowledge based on a speaker's telling, and this
distinction is of the utmost importance for the epistemology of
testimony. A broad conception of the central type of speech act from
which testimonial knowledge can be gained, a conception like that of
Fricker or Audi that encompasses both arguing and telling, is in danger
of putting in the same box items of knowledge with very different
epistemic credentials. I have argued that knowledge based on a
speaker's telling is dependent on the speaker in a way that knowledge
based on a speaker's arguing is not. The point is that this difference in
respect of dependence makes for an epistemic difference. The kind of
justification appropriate to knowledge based on a speaker's arguing
involves appeal to the cogency of the argument. The kind of justification
appropriate to knowledge based on a speaker's telling involves the
citing of (p.58) an authority. The task at hand, then, is to begin to spell
out what the epistemic credentials of knowledge based on the authority
of a speaker actually consist in, what justifying a claim by citing an
authority actually amounts to. In the end, I will argue that coming to
understand the epistemic credentials of knowledge based on the
authority of a speaker requires rejecting (R), but for now the prima
facie difference between knowledge based on a speaker's telling and
knowledge based on a speaker's arguing should be enough to call into
question the utility of a broad conception of testimony. An epistemology
of testimony based on a broad conception of the central type of speech
act from which testimonial knowledge is to be gained is in danger of
eliding a substantive epistemic distinction, one that is of the utmost
importance in our everyday commerce with each other. It is frequently
a point of worry and contention whether and to what extent we are
taking the information a speaker has conveyed to us on her authority,
and any adequate epistemology of testimony must be in a position to
account for this.
I think the most prudent epistemological move here is to disambiguate,

so I will adopt a narrow conception of the central type of speech act
relevant to the epistemology of testimony.20 I want to explore what is
involved in knowing things based on a speaker's tellings, where the
justification of this knowledge involves appeal to the authority of the
speaker. I think this is, and always has been, the most pressing problem
for the epistemology of testimony, especially since this is the kind of
knowledge that is most naturally describable as secondhand. If a
speaker argues that p, and if I come to (p.59) know that p on the basis
of my assessment of the cogency of this argument such that, if asked, I
will justify my knowing that p by appeal to the argument, then it doesn't
seem natural to describe this knowledge as secondhand. There is
certainly a sense in which such knowledge is acquired from a speaker,
but the way in which it is acquired doesn't seem to make it aptly
describable as secondhand. We can thus formulate the following
principle concerning the secondhandness of testimonial knowledge:
(S) An item of knowledge is testimonial just in case it is

secondhand in the demanding sense of being justified by appeal
to the authority of the speaker.
One can be said to know by testimony, in this sense, only when the
ground, basis, or justification of one's knowledge, that which is the
response to the question “How do you know?,” is that one was told by
someone else.
In itself, (S) leaves quite open what exactly it is that justifies knowledge
and belief based on testimony. Both reductionists and anti-reductionists
about testimony should therefore want to accept (S). They will simply
give alternative accounts of what being justified by appeal to the
authority of the speaker amounts to. According to reductionists,
knowledge justified by appeal to the authority of the speaker will be
justified by the strength of an inference from the speaker's testimony
that p, through independently available considerations of the speaker's
trustworthiness, to the conclusion that p. According to anti-
reductionists, knowledge justified by appeal to the authority of the
speaker will be justified (somehow) by features of the testimony itself
rather than by the cogency of an argument. Still, the distinction
between knowledge based on telling and knowledge based on arguing
can begin to help us assess the adequacy of these two competing
accounts.
According to reductionism, the way in which an audience's knowledge

based on a speaker's telling is justified by appeal to the authority of the
speaker is formally identical to the way in which knowledge based on a
speaker's arguing is justified by the audience's assessment of the
cogency of the argument. The only difference is that in the case of
knowledge based on a speaker's telling both the fact that the speaker
has said that p and considerations concerning her trustworthiness serve
as premises in the requisite argument to the conclusion that p. In the
case of knowledge based on a speaker's arguing, there is no need for
these premises concerning the speaker. The way in which knowledge
based on a speaker's telling is justified by appeal to the authority of the
speaker is thus a matter of considerations concerning the speaker

needing to function as premises in the requisite argument to the
conclusion that p. According to reductionism, an audience's relation to
the testimony of a speaker is therefore an essentially juridical one. The
epistemic credentials of an audience's testimonial knowledge are akin
to those of a juror in a court of law, meaning that the audience is in the
position (p.60) of assessing and evaluating the evidence provided by
the speaker and then drawing its own conclusion.21
It is certainly true that an audience must be rationally responsive to

considerations of a speaker's trustworthiness if it is to so much as be a
candidate for testimonial knowledge. If an audience believes what a
speaker says while flying in the face of considerations pointing to the
insincerity or incompetence of the speaker, the audience clearly
shouldn't be credited with testimonial knowledge even if what the
speaker says happens to be true. In this sense, an audience must come
to its own conclusion about considerations of a speaker's
trustworthiness. However, in knowing that p based on a speaker's
testimony that p, the more the audience appears to be in the position of
drawing its own conclusion about p, about the content of what is said
rather than about the trustworthiness of the speaker, then the less it
appears that the audience's knowledge that p can be properly
construed as being justified by appeal to the authority of the speaker.
The task of a juror in a court of law is precisely not to take anything on
the authority of the witness (unless, perhaps, it is an expert witness),
and so it looks like reductionism about testimony is in a poor position to
account for the intuitive epistemic distinction between knowledge
based on arguing and knowledge based on telling. This is not yet to say
that reductionism is false. It may turn out that our epistemic relation to
testimony really is so juridical. In sections 2.4 and 2.5 I will argue that
this can't be right, but for now I simply want to note that reductionism
about testimony looks to be in danger of collapsing the intuitive
epistemic distinction between knowledge based on arguing and
knowledge based on telling.
This isn't to say that anti-reductionism is much better off. According to

anti-reductionism, knowledge justified by appeal to the authority of the
speaker is justified not by the cogency of an argument but rather
(somehow) by the features of testimony itself. Anti-reductionism can
accept the intuition that in order to so much as be a candidate for
testimonial knowledge an audience must be rationally responsive to
considerations of the speaker's trustworthiness, and anti-reductionism
can thus accept that in knowing by testimony an audience must come to
its own conclusion about considerations of the speaker's
trustworthiness. The point is simply that an audience's testimonial

knowledge isn't justified by such responsiveness to the speaker's
trustworthiness but rather by appeal to the authority of the speaker.
Anti-reductionism thus isn't in danger of collapsing the intuitive
distinction between knowledge based on arguing and knowledge based
on telling, but it nevertheless leaves rather mysterious how exactly it is
that knowledge based on a (p.61) speaker's telling can be justified by
appeal to the authority of the speaker. Reductionists at least provide an
account, though a problematic one, of how the authority of a speaker
can figure into the justification of an audience's testimonial knowledge
—via premises concerning the speaker entering into the argument that
constitutes this justification—but anti-reductionists are generally
content to reject the reductionist view without providing an account of
their own of how the authority of a speaker can play a substantive
epistemic role. Claiming that comprehension of a speaker's testimony
provides a prima facie justification for belief does nothing to answer
this epistemological question, and so standard anti-reductionist
accounts of testimony appear equally unable to provide an adequate
account of the way in which testimonial knowledge is secondhand.
2.2 Secondhandness and the Epistemic Right of Deferral

I want to offer an alternative account of the way in which knowledge
and belief can be justified by appeal to the authority of the speaker, an
account that (1) doesn't run the risk of making our epistemic relation to
testimony look too juridical to be appropriately secondhand, and (2)
actually provides an account of how the authority of a speaker can play
a substantive epistemic role. This will require fleshing out what is
involved in (S) and in the end will provide the materials for showing
that (R) is actually inconsistent with (S) as properly understood.
Imagine the following scenario that is, I think, typical of our ordinary
epistemic practice with respect to testimony. Alfred and Sylvia are
chatting in a café, while Mary, an acquaintance of theirs, sits at a
nearby table reading a book. Alfred and Sylvia begin to discuss a party
that took place the previous weekend but which neither of them
attended.
Alfred:
George was there, and he made a real fool of himself.

Sylvia:
How do you know?

Alfred:
Mary told me so.

Sylvia:
But I talked to George the day before the party, and he told me he
would never go to such a thing.
Alfred:
Don't ask me; Mary's the one who told me. [turning to Mary] Mary, are
you sure that George was at the party last weekend? He told Sylvia
that he wasn't going.
Our scenario might now continue in any of the following three ways
depending on the possible responses available to Mary.
Mary's first possible response: Well, I thought it was him that I

saw. I guess it was rather dark and crowded. Maybe he really
wasn't there.
(p.62)
Mary's second possible response: Oh yeah, I talked to him for a

good fifteen minutes.
Mary's third possible response: Well, Max told me he was there.
I take it that this scenario proceeds schematically as follows: Alfred

enters a claim to know; Sylvia requests a justification for the claim;
Alfred provides such a justification; Sylvia challenges the claim; Alfred
responds to the challenge by deferring it to Mary; and finally Mary
either fails to meet the challenge, meets the challenge, or defers the
challenge herself. What is important and interesting here is that when
Alfred justifies his claim to know that George was at the party by citing
an authority, and when Sylvia challenges him by producing evidence
that counts against this claim, Alfred has the right to defer this
challenge back to the authority, to Mary, whereupon Mary, not Alfred,
seems responsible for meeting it. The way in which Mary is able to
respond to the challenge then seems to directly affect the justification
of Alfred's belief. If Mary fails to meet the challenge, as in her first
response, then Alfred's justification appears to be dissolved. If Mary
succeeds in meeting the challenge, as in her second response, then
Alfred's justification looks to be upheld. Finally, if Mary defers the
challenge, as in her third response, then both Alfred's and Mary's
justification seems dependent upon Max's ability to meet the challenge.
Whether Alfred and Mary remain justified in believing that George was
at the party depends on whether Max is capable of meeting the
challenge.22
These features of our ordinary epistemic practice point to the way in

which testimonial knowledge and belief involves what we might call an
epistemic right of deferral. It is a general feature of epistemic agency
that mature epistemic agents are under a standing obligation to
respond to relevant epistemic challenges to what they believe, either by
meeting the challenge or by giving up their belief. What is so peculiar
about testimonial knowledge is that, insofar as the justification
appropriate to testimonial knowledge involves the citing of an authority,
a testimonial audience is entitled to defer relevant challenges back to
the original speaker. This is a direct result of the way in which, in
believing that p on the basis of a speaker's testimony that p, an
audience isn't coming to its own conclusion about p. When Alfred
repeats what he has come to know based on Mary's testimony, and
when Sylvia challenges him by producing evidence that tells against
what it is that he has claimed to know, he has the right to defer the
challenge back to Mary. If Alfred were coming to his own conclusion
about things, then there would be little sense to be made of his
entitlement to defer to Mary a challenge to his own epistemic
conclusion. If Alfred had claimed (p.63) to have seen George at the
party for himself, for example, then he clearly would have been fully
responsible for meeting Sylvia's challenge. In knowing based on
perception, a subject is coming to her own conclusion about things, and
this means that it is part of the knower's epistemic burden to meet
challenges to this conclusion. In knowing based on testimony, however,
an audience is not coming to its own conclusion about things, and as a
result the epistemic burden with regard to meeting challenges to the
audience's testimonial knowledge is shared out between speaker and
audience.
Even if the burden of meeting challenges is shared out in this way

between speaker and audience, the audience doesn't have the right to
defer all kinds of challenges back to the original speaker. As I have put
it, the audience has the right to defer only those challenges that involve
the production of evidence against what it has claimed to know. If an
audience claims to know that p based on testimony, it can defer only
those challenges that involve the production of evidence against p. In
the above example, Alfred claims to know that George was at the party,
and Sylvia challenges him by producing evidence that counts directly
against George's being at the party. In this case, Alfred clearly has the
right to defer the challenge back to Mary. If the challenge takes a
different form, however, then he doesn't have this right. If Sylvia
challenges Alfred's claim to know that George was at the party by
pointing out that, unbeknownst to Alfred, Mary is an inveterate liar,
then Alfred does not have the right to defer the challenge back to Mary.
In such a case, Mary's being an inveterate liar would not tell directly
against George's being at the party, but only against Mary's
trustworthiness and hence against Alfred's ability to know that George
was at the party on the basis of Mary's testimony. As we've seen, an
audience must come to its own conclusion about considerations of a
speaker's trustworthiness, and this means that it is part of the
audience's epistemic burden to meet challenges to this conclusion. Still,
this does not mean that in knowing based on testimony the audience is
coming to its own conclusion that what the speaker says is true, and
this is exhibited by the fact that the audience's epistemic burden with
regard to a challenge involving the production of evidence against what
it has come to know by testimony is met by deferring the challenge to
the original speaker.
I have claimed that Alfred can fulfill his epistemic burden with respect
to Sylvia's challenge by deferring the challenge back to Mary and that
the nature of this deferral is such that, if Mary is unable to meet the
challenge, then Alfred is no longer justified in believing that George
was at the party. But what if Mary isn't available to meet the challenge?
What if there is no practical way in which Sylvia can go and confront
Mary with the challenge? Is Alfred's belief still justified? And what if, as
is the case with much testimonial knowledge with respect to historical
matters, the original speaker from which an item of testimonial
knowledge was acquired is long since deceased? Can we still maintain
our justification in the face of challenges by deferring to such speakers?
These are difficult questions. In the end, I think a lot will turn on the
details of the particular case—for example, whether one's testimonial
knowledge is justified by an (p.64) individual who happens to be a
party in a conversation or by an amorphous epistemic community. I
have focused on paradigmatic cases of testimonial knowledge acquired
from a speaker's telling, and in these cases I think it is clear that, as
long as the testimonial speaker is capable of meeting a challenge were
she to be presented with it, a testimonial audience can maintain her
justification by deferring a challenge back to the testimonial speaker
even if the speaker is not in fact present to meet the challenge. When it
comes to cases in which the testimonial speaker is incapable of meeting
the challenge were she to be presented with it, though for nonepistemic
reasons such as death or impairment, things become less clear.
Whether the audience can maintain her justification by simply deferring
the challenge may depend to a significant extent on the strength of the
challenge, on the degree to which it tells against the audience's
testimonial belief. And when in comes to such things as testimonial
knowledge of historical facts, this knowledge may in fact be justified by
an epistemic community, by the community of historians, such that any

living member of this community will be in a position to meet the
challenge.
Again, these are difficult issues, but an adequate epistemology of

testimony should recognize that there is in fact a good deal of grey area
here. Moreover, an adequate epistemology of testimony should be in a
position to say something substantive about why, in fact, this area is
grey. The precise extent to which an audience's beliefs are based on the
authority of a speaker or group of speakers is something that varies
widely and that is often difficult to discern. Insofar as this is the case,
the degree to which the audience is entitled to defer particular
challenges to her beliefs, challenges that may themselves have degrees
of strength, will vary widely as well. I have abstracted away from all of
this in offering my account of deferral, but I think that the account I
have offered does illuminate what is at stake in these more complicated
cases. So, while acknowledging the difficulties here, I will simply say
that the epistemic burden with respect to meeting challenges to an
audience's testimonial knowledge is shared out between speaker and
audience in the sense that (1) the audience can fulfill its epistemic
burden with respect to the challenge by deferring the challenge to the
original speaker, but (2) the original speaker then still bears the burden
of meeting the challenge such that if she is epistemically incapable of
meeting it then the audience's justification is lost.23
(p.65) 2.3 The Epistemological Problem of Testimony Revisited

I have argued that testimonial knowledge—knowledge justified by
appeal to the authority of a speaker—is distinguished by the fact that it
involves an epistemic right of deferral. Testimonial knowledge and
belief is distinctively epistemically mediated in such a way that an
audience is entitled to defer challenges to its belief back to the original
speaker. This then gives us a new way of framing the epistemological
problem of testimony. The overarching problem for epistemological
accounts of testimony is to give an adequate explanation of the precise
sense in which testimonial knowledge, unlike any other form of
knowledge, is mediated in virtue of being secondhand. But if the
epistemic mediation distinctive of testimonial knowledge and belief is
characterized by the fact that it generates an epistemic right of
deferral, then explaining the mediation characteristic of testimonial
knowledge will involve explaining the right of deferral. In virtue of what
is a (p.66) testimonial audience epistemically entitled to defer
challenges to its testimonial knowledge and belief back to the original
speaker? The epistemological problem of testimony can thus be recast

as a problem of explaining the epistemic right of deferral.
Recasting an epistemological problem in this fashion has some

precedent. Much recent work on self-knowledge has approached the
epistemological problem of self-knowledge in terms of the problem of
first-person authority or first-person privilege.24 Explaining the
epistemology of self-knowledge is thus understood to involve explaining
the peculiar kind of privilege that a subject has in thinking and
speaking about her own conscious psychological states—a privilege
that doesn't seem to be based on evidence, but where not being based
on evidence actually contributes to the strength of this privilege rather
than detracting from it. Similarly, the epistemological problem of
testimony may be best understood as a problem of explaining the
peculiar kind of privilege that an audience has in thinking and speaking
about knowledge it has gained at second hand, a privilege in virtue
which, as we've seen, the audience is entitled to defer epistemic
challenges back to the original speaker. For the sake of parallelism, we
might call this the problem of second-person authority or second-person
privilege.
This appeal to the notion of the second person is reminiscent of Reid's

claim that the social operations of mind are often expressed in
language through the use of grammatical markers of address. With
respect to the capacity for learning from testimony, Reid's thought
seems to be that the operation of this cognitive capacity requires two
individuals relating to one another as addresser to addressee. In this
respect, it is important to note that it seems to be a necessary condition
on an audience's acquiring an epistemic right of deferral that the
speaker has actually addressed the audience. Consider in this regard
what would happen if, instead of being told by Mary that George was at
the party, Alfred simply overheard Mary say that George was at the
party. We might be able to imagine cases like this in which Alfred would
still be able to defer Sylvia's challenge back to Mary, for example, cases
in which Alfred's “overhearing” of Mary's conversation occurred in a
public enough context such that the addressees of Mary's remark might
be taken to include everyone within earshot.25 Nevertheless, there (p.67)
is intuitively a kind of overhearing that does debar the person who

overhears from being able to defer challenges back to the speaker—
namely, the kind of overhearing that explicitly debars the overhearer
from being considered an addressee in even the widest possible sense.
We can call such a kind of overhearing mere overhearing. If Alfred
merely overhears Mary say that George was at the party by, for
example, eavesdropping on her telephone conversation over a wiretap,
then it seems pretty clear that it would be inappropriate in this

situation, even bizarre, for Alfred to defer to Mary Sylvia's challenge to
his claim to know that George was at the party. This isn't to say that
Alfred can't come to know that George was at the party based on
merely overhearing Mary's remark. The point is only that in doing so he
is coming to his own conclusion about things. He is assessing the
available evidence, evidence that includes Mary's statement that
George was at the party, and drawing his own conclusion, and this
means that he shouldn't defer challenges back to Mary. Testimonial
knowledge, knowledge involving an entitlement to defer challenges
back to the original speaker, can thus only be acquired when a speaker
(to at least some degree) actively addresses an audience.
The idea that the acquisition of properly testimonial knowledge

requires that an audience be an actual addressee of a speaker's
testimony is rejected by nearly all contemporary accounts of
testimony.26 Few accounts of the epistemology of testimony take note of
the phenomenon of the deferral of challenges, however, and address
does seem to be a necessary condition on deferral. If a hearer is not an
addressee of a speaker's testimony in even the widest possible sense,
like a detective listening in on a wiretap, then even though the mere
overhearer may still be in a position to acquire knowledge and justified
belief from the speaker's testimony, the mere overhearer is not entitled
to defer challenges. And if, as I have argued, properly testimonial
knowledge is distinctively characterized by the phenomenon of
deferral, then the mere overhearer is not in a position to acquire
properly testimonial knowledge.27
Being addressed by a speaker is, of course, only a necessary condition

on an audience's being able to acquire testimonial knowledge. There
are all kinds of cases in which a speaker can address an audience (with
a joke, an argument, an explanation, etc.) but where the audience is not
thereby put in a position to acquire testimonial knowledge. As we've
seen, testimonial knowledge is knowledge acquired from a speaker's
telling, and telling an audience that p involves far more than addressing
the audience. Nevertheless, it is important to see that, insofar as
address is a necessary condition on acquiring testimonial knowledge,
an audience seems to be put in the position of acquiring testimonial
knowledge based at least in part on the speaker's taking up a (p.68)
certain kind of attitude towards it—an attitude far more elaborate than
the bare notion of address but an other-directed attitude just the same.
Understanding the epistemology of testimony thus requires
understanding what exactly this attitude involves and how it can
produce in an audience an epistemic state with the peculiar features of

testimonial knowledge.
If one accepts this way of posing the epistemological problem of

testimony—if one accepts that the problem of the epistemology of
testimony is a problem of explaining the second-person privilege in
virtue of which a testimonial audience is epistemically entitled to defer
challenges—then I think that the solution is relatively simple. The best
explanation of the fact that an audience is entitled to defer certain
challenges to her testimonial beliefs back to the original speaker is
that, in testifying, a speaker is assuming an epistemic responsibility to
meet such challenges, and in accepting the speaker's testimony, the
audience is acknowledging the speaker's assumption of responsibility
by being disposed to defer such challenges. In this respect, testimonial
knowledge is genuinely the result of what Reid calls “a social operation
of mind.” It is the result of a complex cognitive capacity the exercise of
which requires the cooperative activity of both speaker and audience.28
The exercise of this capacity requires a speaker who, in addressing the
audience with her testimony, assumes partial epistemic responsibility
for the audience's belief, and it requires an audience who, in accepting
the speaker's testimony as such, acknowledges the speaker's
assumption of responsibility by being properly disposed to defer
challenges. If something goes wrong on either side of this reciprocal
requirement—if the speaker doesn't properly assume the responsibility
to meet challenges to the audience's belief, or if the audience doesn't
properly acknowledge the speaker's assumption of responsibility—then
the audience cannot acquire properly testimonial knowledge.29
That testifying involves such an assumption of responsibility on the part

of the speaker is born out by the fact that a speaker can choose to do
something less than testifying precisely in order to avoid assuming this
responsibility. We often use locutions such as “I believe that p” or “I
think that p” when we intend to communicate to an audience (p.69)
that we believe that p, perhaps intending the audience to take the fact
that we believe that p to be good evidence for the truth of p, but where
we are unwilling to assume the responsibilities involved in coming out
and telling the audience that p, in saying “p” or “I know that p.”30 Let's
somewhat artificially call this speech act of doing something less than
testifying the speech act of merely declaring a belief. In merely
declaring the belief that p, we may want the audience to know that we
believe that p and to take this into consideration in deciding whether it
should itself believe that p, but we do not intend the audience to take
our word for it.31 Instead, we intend the audience to come to its own
conclusion about things; we intend the audience to take our declaration
of belief as a consideration that tells in favor of p but one that the

audience must assess in the light of all the other available evidence.
And the reason we intend that the audience come to its own conclusion
about things is so that we don't incur the responsibility involved in
having to meet challenges that the audience may defer to us. If I merely
declare to the audience my belief that p—if I essentially say “I believe
that p, but make of it what you will”—then the audience isn't entitled to
defer challenges back to me (even if it turns out that I do in fact know
that p). Even if the audience can come to know that p based on my mere
declaration of belief, the audience isn't entitled to hold me partially
responsible for the justification of her knowledge that p. Such
knowledge is thus different from testimonial knowledge as here
defined; it is not knowledge that is secondhand in the demanding sense
of being justified by appeal to the authority of the speaker.
If a speaker does not genuinely assume partial responsibility for an

audience's testimonial belief by making her assertion an instance of
testimony, then the audience cannot acquire properly testimonial
knowledge. Equally, however, if the audience does not properly accept
the speaker's assumption of responsibility by being disposed to defer
relevant challenges to the content of her testimonial belief back to the
testimonial speaker, then the audience cannot acquire genuinely
testimonial knowledge. Even in cases in which testimony is genuinely
offered, we often do not accept it as such. We (p.70) instead treat the
speaker's testimony that p as an ordinary consideration that may or
may not count in favor of believing that p. In such cases, the speaker's
testifying that p is assessed in light of all of the other available
evidence, and then we draw our own conclusion: “So-and-so tells me
that p, and given all the other evidence, she's probably right.” We can
certainly acquire knowledge by treating a speaker's testimony in this
way, but such knowledge doesn't appear to be justified by appeal to the
authority of the speaker. Such knowledge doesn't involve taking the
speaker's word for it, and hence when we treat a speaker's testimony in
this way, we don't seem to be entitled to defer challenges. When
confronted with such a challenge, we are definitely required to
reconsider the evidence, but we are not entitled to maintain our belief
by deferring the challenge back to the testimonial speaker.32
We now have a much more detailed account of what is involved in (S).

(S) states that testimonial knowledge is secondhand in the demanding
sense of being justified by appeal to the authority of the speaker. We've
seen that though both reductionist and anti-reductionist accounts of
testimony should want to accept this claim, they both have trouble
making any detailed sense out of it. Reductionist accounts look too
juridical to be able to make clear sense of how testimonial knowledge

involves an appeal to the authority of the speaker, and anti-reductionist
accounts generally leave it mysterious how the authority of a speaker
actually plays a role in justifying knowledge and belief based on
testimony. The account I have offered avoids both of these difficulties.
Though it acknowledges that an audience must come to its own
conclusion about considerations of a speaker's trustworthiness in order
to be a candidate for testimonial knowledge, it holds that the audience
is not in the position of coming to its own conclusion about the content
of the speaker's testimony. Instead, an audience's testimonial
knowledge is the result of a complex social capacity, the exercise of
which involves the speaker's assuming an epistemic responsibility
towards the audience and the audience's acknowledging this
assumption on the part of the speaker.
(p.71) 2.4 An Argument from Secondhandness

This discussion of the deferral of challenges goes a long way towards
cashing out what is involved in (S). I have already argued that
reductionism about testimony looks to be in danger of making our
relation to testimony look too juridical to be able to account for (S).
Now that we have a more detailed account of what is actually involved
in (S), I want to argue that reductionism's commitment to (R) is actually
inconsistent with the secondhandness of testimonial knowledge as
expressed in (S).
(R) states that an audience's justification in knowing based on

testimony consists in the strength of an inference from a speaker's
saying that p, through independently available considerations
concerning the speaker's trustworthiness, to the conclusion that p. In
order for (R) to be consistent with the (S), it must be consistent with
the way in which testimonial knowledge involves an epistemic right of
deferral. However, the very idea that an audience's justification in
knowing based on testimony consists in the strength of an inference
makes it difficult to see how anyone other than the audience can
legitimately be held responsible if its testimonial knowledge is
challenged. If, as the reductionist has it, the justification of an
audience's testimonial knowledge consists in the strength of an
inference, then this justification should be something that the audience,
not the speaker, is ultimately responsible for. Whether it has done so
consciously or unconsciously, the audience has evaluated the relevant
premises and drawn the relevant conclusion. Therefore, it shouldn't be
entitled to defer to someone else responsibility for meeting epistemic
challenges to this conclusion. Reductionism about testimony, in its
commitment to (R), is thus inconsistent with the secondhandness of

testimonial knowledge as expressed in (S).
The conclusion of this argument from secondhandness is that an

audience's epistemic right of deferral and the relations of interpersonal
epistemic responsibility that this right involves are inconsistent with
the audience's justification being constituted by the strength of an
inference. In order to avoid this conclusion, the reductionist must either
(1) find some way to make the existence of an audience's epistemic
right of deferral consistent with her inferential conception of the
justification relevant to testimonial knowledge, or (2) deny that
testimonial knowledge is secondhand in the demanding sense outlined
in the previous two sections. I will consider the first strategy in this
section and the second strategy in section 2.5.
There are at least three ways in which the reductionist might attempt
to make the existence of an audience's epistemic right of deferral
consistent with her inferential conception of justification. First, the
reductionist might claim that an audience's deferral of a challenge back
to the original speaker is just the audience's way of displaying the
evidence upon which she has based her claim to know. Deferring a
challenge back to the original speaker is thus simply a matter of
deferring to the evidence. The (p.72) fact that an audience can defer to
the evidence in this way poses no threat at all to construing her
justification in terms of inference, so the existence of an epistemic right
of deferral is completely consistent with a reductionist account of the
justification appropriate to testimonial knowledge.
The problem, however, is that this doesn't look like an account of the
deferral of challenges at all. Such an account can make no sense of the
way in which the original speaker is epistemically responsible for
meeting challenges that her audience has deferred to her. If an
audience's deferral of a challenge simply amounts to displaying to the
challenger the evidence upon which its claim to know is based, then
this doesn't imply that anyone other than the audience is epistemically
responsible for meeting the challenge. As we've seen, however, the
original speaker does seem to be epistemically responsible for meeting
such challenges, and an audience's deferral of a challenge back to the
original speaker is a demand that she discharge this responsibility. An
audience's epistemic right of deferral thus cannot be understood simply
as a way of deferring to the evidence.33
Second, the reductionist might object that an audience's deferral of a

challenge back to the original speaker is, from the audience's
perspective, nothing more than a way for the audience to reassess the
available evidence. In deferring a challenge to what it has claimed to

know based on testimony, the audience is here holding the speaker
responsible for meeting the challenge, but from the audience's
perspective this is simply a way of confronting the original speaker with
the challenge in order to assess the strength of the evidence that the
original speaker is in a position to provide. The original speaker's
response to the challenge then provides a further premise from which
the audience is able to infer its own conclusion. If this is the case, then
an audience's right of deferral is fully consistent with its inferentially
coming to its own conclusion about what the speaker says.
As we've seen, when Alfred defers Sylvia's challenge to Mary, then her
response has a direct effect on his justification. On this proposed
reductionist account, however, her response affects his justification
only in virtue of providing additional evidence from which it is his
responsibility to infer the appropriate conclusion. The problem is that
when Mary responds to the challenge, it doesn't look like Alfred is now
in the position of thinking to himself, “Well, now I have further evidence
from which to reassess the strength of the argument available to the
conclusion that George was at the party.” The (p.73) problem isn't that
this is an overly intellectualist picture of Alfred's process of belief
formation. Rather, the problem is that Mary's response to the challenge
seems to reaffirm Alfred's knowledge rather than functioning as
additional evidence from which he must draw his own conclusion. When
Alfred defers the challenge to Mary, Mary can meet the challenge
simply by repeating her previous testimony—“Yeah, he was there.” The
additional evidence that merely repeating her testimony could provide
is quite minimal, and in the face of the kind of counterevidence that the
challenge represents, it is not clear how this additional evidence could
vindicate Alfred's justification in as swift and decisive a manner as
Mary's meeting the challenge in this way actually does. The
decisiveness for Alfred of Mary's response appears more akin to the
decisiveness of a person's returning to a room to check on a book she
thought she knew was there after a doubt has been raised about its
location. Seeing the book again doesn't provide more evidence that
needs to be assessed against the doubt that has been raised. Instead, it
simply reaffirms what was known. Similarly, Mary's meeting the
challenge by repeating her testimony simply reaffirms Alfred's
knowledge. An audience's deferral of challenges thus seems to play a
different epistemic role from that of the assessment of evidence for the
purposes of inference.
A third way in which the reductionist might seek to make the existence
of an epistemic right of deferral consistent with her inferential account
of justification is by claiming that the speaker's assumption of epistemic
responsibility towards the audience from which the audience's right of
deferral derives can itself function as evidence from which the audience
can infer the conclusion that what the speaker says is true. In testifying
that p, the speaker is assuming responsibility for the audience's
justification and thereby granting the audience the entitlement to defer
challenges back to her. The fact that the speaker is assuming
responsibility in this way can then function as a particularly significant
premise from which the audience can infer that p. We thus have an
account that understands the justification of an audience's knowledge
based on testimony in terms of inference while still providing a place
for the speaker's assumption of epistemic responsibility towards the
audience.
The problem is that on this account the speaker's assumption of

responsibility is just another consideration that figures argumentatively
into the audience's justification. As such, even though the speaker has
granted the audience the entitlement to hold her responsible if what
she tells turns out to be false, the audience seems to have shrugged
this entitlement off. Insofar as the audience has assembled and
assessed the argument the cogency of which constitutes the
justification of its knowledge, it shouldn't defer challenges to its
testimonial knowledge back to the speaker. Hence, if the assumption of
responsibility on the part of the speaker is to entitle the audience to
defer challenges back to the speaker, this assumption of responsibility
cannot simply figure inferentially into the audience's justification.
(p.74) Reductionism about testimony seeks to explain the epistemic

credentials of testimonial knowledge in terms of the cogency of an
argument from a speaker's saying that p, through independently
available considerations of the speaker's trustworthiness, to the
conclusion that p. I have argued that this is inconsistent with the
relations of authority and responsibility that characterize the
secondhandness of testimonial knowledge. Knowledge acquired by
testimony is secondhand in the sense that another person (the speaker)
is partially epistemically responsible for the audience's belief.
According to reductionism, epistemic responsibility for the audience's
belief must always lie with the audience, and so it is inconsistent with
the secondhandness of testimonial knowledge. If testimonial knowledge
is secondhand in the demanding sense captured by (S), then our
justification in knowing based on testimony cannot be understood on

the model of (R).
2.5 Skepticism About Knowing at Second Hand

There is yet another strategy the reductionist might employ in order to
resist the argument from secondhandness. I have argued that
reductionism's commitment to (R) makes it inconsistent with the
secondhandness of testimonial knowledge as expressed in (S). A
reductionist might very well accept this but retain a commitment to (R)
by simply rejecting (S), by denying that the audience's right of deferral
and the interpersonal relations of responsibility that this right involves
are of any proper epistemological significance. While it may be true
that we often defer challenges to our testimonial knowledge back to the
original speaker, and while it may be true that the original speaker is
often obligated to meet challenges thus deferred to her, perhaps this is
nothing more than a social convention that plays no role in
epistemically justifying an audience's testimonial belief. The deferral of
challenges and the relations of interpersonal responsibility that come
with it may thus be a legitimate social affair and may have legitimate
moral implications, but they are of no proper epistemological
significance whatsoever. If this is the case, then the justification of an
audience's testimonial knowledge can still consist in the strength of an
inference.
This is a difficult line of thought to combat. As we saw in chapter 1, the

idea that relations of interpersonal authority and responsibility can play
no legitimate epistemic role was a driving force behind the historical
development of philosophical conceptions of testimony. We have
inherited from the Enlightenment a strong tendency to think that
taking someone's word for something is just being gullible,
intellectually naϯve, or somehow epistemically irresponsible.
Nevertheless, it is important to recognize what is implied by rejecting
the epistemic significance of what I have identified as the
secondhandness of testimonial knowledge. If there were no
epistemologically significant right of deferral, if the deferral of
challenges was only a conventional or (p.75) moral issue, then the
audience would always be solely epistemically responsible for meeting
challenges to its testimonial knowledge. In our above case, Alfred would
be solely responsible for meeting Sylvia's challenge, and as is the case
for knowledge based on inference or perception, any failure to meet the
challenge himself would require that he give up his claim to know. If
Alfred attempted to defer the challenge to Mary, Sylvia would have the
right to refuse this; she would have the right to hold Alfred responsible
for the epistemic credentials of what he has claimed to know,

regardless of the fact that Mary told him so.
On the one hand, this just doesn't look like our ordinary epistemic
practice. We ordinarily feel entitled to defer challenges to what we
claim to know based on testimony back to the original speaker, and if
someone were to tell us that, at least epistemically speaking, this is
illegitimate, I think we would be unimpressed. Partially thought
through, then, rejecting the epistemic significance of the audience's
right of deferral leaves us with a picture of an epistemic practice that
doesn't look like our own.
More fully thought through, however, I doubt that such an account can
provide us with a picture of how we know things from other people at
all. Rejecting the interpersonal relations of authority and responsibility
that I have taken to characterize the secondhandness of testimonial
knowledge involves treating others as mere sources of information
rather than genuine authorities.34 It leaves the audience in an
epistemic position formally identical to that of someone who merely
overhears a speaker's assertion. Now, I have admitted that an audience
can gain knowledge of the content of a speaker's assertion by merely
overhearing her assertion if the audience has sufficient reason to
believe that there is an appropriate connection between what the
speaker says and the truth. I have simply claimed that such knowledge
doesn't involve an epistemic right of deferral and is thus inferential
rather than properly testimonial. The reductionist is now claiming that
all testimonial knowledge is like this, that all testimonial knowledge is
like knowledge based on mere overhearing. But if testimonial
knowledge can only be acquired in this way, then it becomes harder and
harder to see how what we are here picturing is a kind of knowledge
that is distinctively gained from other people. On this account, what
other people say is only epistemically significant insofar as it functions
as a source of inference, as just some more evidence from which we can
infer certain things about the world. This looks more like a denial that
there is such a thing as secondhand knowledge than a vindication of the
epistemic significance of the (p.76) words of others. A reductionism of
this form thus looks more like a form of skepticism—skepticism about
knowledge based on the authority of others. Reductionism about
testimony and skepticism about knowledge gained at second hand are
here two sides of the same coin.
Of course, one might think that skepticism about knowledge gained at

second hand isn't epistemologically unpalatable as long as we can
acquire knowledge based on inference from what people say. Still, I
think we can begin to see here that the real motivation for this kind of
reductionism about testimony arises prior to any explicitly

epistemological considerations concerning the place of testimony in
relation to other sources of knowledge. This kind of reductionism about
testimony is straightforwardly motivated by epistemic autonomy.
Absent a prior conviction that the authority of others cannot play a
legitimate epistemic role, absent a prior conviction that others can only
be sources of information and not genuine authorities, there is no
reason to think that the epistemic right of deferral and the relations of
interpersonal authority and responsibility that this right involves can
play no legitimate epistemic role.
I think the upshot of the argument from secondhandness is that we

must view the vast areas of knowledge that we have gained from
testimony as constitutively dependent on the relations of authority and
responsibility that we bear towards others. Knowing by testimony
involves ceding to the speaker partial responsibility for the justification
of our testimonial belief. As Elizabeth Anscombe puts it, it involves
trusting the speaker for the truth (1979: 151). Of course, there is still
much that we, as audience, must be responsible for if we are to be in a
position to gain knowledge in this way. We shouldn't trust just anyone.
We must be rationally responsive to considerations of the speaker's
trustworthiness if we are to so much as be candidates for testimonial
knowledge. The point is only that the justification of our testimonial
knowledge cannot be made out simply in terms of what we, as
audience, are responsible for. The exercise of our own judgment only
gets us to a position in which we are candidates for testimonial
knowledge, and from there we must trust the speaker.
Notes:
(1) It is much more natural to describe knowledge as something that
can be acquired at second hand than it is to describe belief in this way.
It is natural to say that I know something at second hand, but less
natural to say that I believe something at second hand. I suspect that
this has something to do with the way in which our ordinary concept of
knowledge, or at least some dimension of this concept, is much more
closely connected to the institution of testimony than is our ordinary
concept of belief. Though I won't try to defend this claim here, it may
be the case that whereas the concept of belief has its home in the
context of the explanation and prediction of behavior, the concept of
knowledge has its home in the context of the giving and taking of
testimony, a context that, as I argue in this book, is delineated in terms
of relations of authority and responsibility between persons.
(2) In fact, both of these situations involve multiple layers of testimony,

knowledge that is first acquired by testimony and then passed on to
someone else through testimony. I have introduced these multiple
layers to illustrate the ways in which an audience ordinarily justifies
and defends its testimonial knowledge. These ordinary epistemic
practices are what any account of the epistemology of testimony must
be in a position to explain (or explain away).
(3) Contemporary reductionists about testimony include Fricker (1987),

(1994), and (1995), Lyons (1997), and Shogenji (2006). I am here
eliding a distinction between global and local reductionism about
testimony. Global reductionism seeks to reduce our justification in
knowing based on testimony writ large to independent observation and
inference. Local reductionism seeks to reduce only our justification in
believing any particular piece of testimony to independently available
considerations of the speaker's trustworthiness, meaning such
independently available considerations can include information gained
from previous testimony. See Fricker (1995). My argument in this
chapter is pitched against the more plausible, local version of
reductionism, though it cuts against the global version as well. For a
recent defense of global reductionism, see Shogenji (2006).
(4) I am here taking considerations of a speaker's trustworthiness to

encompass all those considerations in virtue of which we judge a
speaker both competent concerning the subject matter at hand and
sincere in her assertion. I am here following Fricker's usage and
leaving aside the significant problem of spelling out in detail just how
these notions pertain to the epistemology of testimony. For some of the
issues here, see Moran (2004).
(5) Contemporary anti-reductionists about testimony include Coady

(1992), Burge (1993), (1997), McDowell (1998a), and Stevenson (1993).
(6) Anti-reductionists usually describe testimony as a “basic” or

“autonomous” source of knowledge rather than a sui generis one, and
the relevant sense of basicness or autonomy is often left quite unclear.
For an account of some of the possibilities here, see Audi (1997). In this
chapter I will argue that testimony is a sui generis source of knowledge
in that testimony is the result of an epistemic capacity that yields states
of knowledge with epistemic credentials unlike any other form of
knowledge. If this is right, then any anti-reductionist account that
simply portrays testimonial knowledge as the result of another reliable
belief-producing process must be inadequate.
(7) Some epistemologists have recently endorsed a kind of hybrid

reductionist/anti-reductionist position according to which the
justification of testimonial knowledge can consist either in the strength
of an inference or in simple comprehension. See Faulkner (2000),
Weiner (2003), and Graham (2006). On this kind of view, testimonial
knowledge is not a unitary epistemological category. Testimonial
knowledge amounts to an epistemological category that encompasses
items of knowledge with very different epistemic credentials. I think
such a view is methodologically undesirable. The epistemological
problem of testimony has always been a problem of making out what it
is that justifies knowledge and belief based on testimony. Anti-
reductionists have argued that, whatever in fact it amounts to, there is
a particular kind of justification that is peculiar to testimony.
Reductionists have argued that there is not, that testimonial knowledge
is simply a species of knowledge justified by inference. These would
appear to be mutually exclusive options. Either there is a kind of
justification peculiar to testimonial knowledge or there is not.
Moreover, if one holds that the term “testimonial knowledge” doesn't
refer to an epistemologically unified class of knowledge but rather is a
grab-bag term for any kind of knowledge that has an instance of
testimony somewhere in its causal ancestry, then testimonial knowledge
ceases to be an interesting epistemological topic. Rather, something
like comprehension-based knowledge (a subset of testimonial
knowledge) is now the epistemological category that is urgently in need
of explanation. I think that the most prudent epistemological position is
that the term “testimonial knowledge” should be used to mark out a
unitary epistemological category. On a reductionist account, the term
refers to a species of inferential knowledge, knowledge justified by the
strength of the kind of argument specified above. On an anti-
reductionist account, the term refers to that kind of knowledge that is
justified in whatever way is peculiar to testimony, say, by
comprehension. The anti-reductionist can admit, as the hybrid theorists
do, that we often treat testimony as evidence from which we simply
draw inferences, but she shouldn't take such knowledge to be properly
testimonial.
(8) As Faulkner claims, “testimony is mediated in the sense that the

intentions of another, and the justification possessed by another are
relevant to the audience's acquisition of knowledge. These
considerations have no parallel in either perception or memory” (2000:
581). This means that “we must support our acceptance of testimony
with reasons” (2000: 583) and that the need for these reasons is what
makes testimonial knowledge distinctive. See also Fricker (2006a).
(9) My account does have much in common with the work of some
philosophers who have been interested in the relationships between
people involved in communication. Writers who stress this aspect of
testimony include Austin (1979), Anscombe (1979), Cavell (1979: 389–
93), Ross (1986), Moran (2004) and (2005), Hinchman (2005), and
Faulkner (2007a) and (2007b). It should be clear to anyone familiar
with Moran's work on testimony that the account of the epistemology of
testimony that I develop in this chapter is heavily indebted to Moran. I
believe that this account comes most clearly into focus in the way in
which I have presented it here, and so I do not here present the
account as a development and refinement of Moran's view. I discuss
some of the differences between Moran's view and my own in chapter
3.
(10) Recall that the upshot of our discussion of the disagreement

between Hume and the Port Royal Logic concerning the rationality of
belief in miracles on the basis of testimony was that the Logic gives the
category of other persons a much more significant role in the
justification of belief than does Hume.
(11) For the distinction between “formal” and “natural” testimony that
is generally adopted in the subsequent literature, see chapter 1 of
Coady (1992). As we'll see, there is a real question whether “formal”
and “natural” testimony can be given a single epistemological
treatment. The problem is that in the court of law there seems to be
little room for taking anything on the authority of the speaker. Instead,
a jury is expressly charged with assessing and evaluating the testimony
of a witness against all the other available evidence and rendering a
verdict thereby. As I will argue, properly testimonial knowledge gained
from a speaker's “natural” tellings cannot be the result of adopting
such a juridical attitude. Acquiring testimonial knowledge requires that
an audience assess a speaker for trustworthiness, but it cannot involve
an audience's coming to her own conclusion about what the speaker
says.
(12) Lackey claims that Fricker and Audi endorse a view of testimony
according to which S testifies that p if and only if “S's statement that p
is an expression of S's thought that p” (2006a: 182). Given the above
quotations, this is a mischaracterization of both Fricker and Audi. Both
Fricker and Audi hold that testimony involves more than the mere
expression of a thought. The expression of the thought must be “aimed
at communication” or “intended to convey information,” which excludes
such things as passing remarks.
(13) Our account must also rule out cases in which, for example, an
audience comes to know that a speaker has a baritone voice on the
basis of her saying “I have a baritone voice” in an attempt to convey
this information to the audience but where the audience's knowledge is
based on or justified by the audience's perception of the quality of the
speaker's voice rather than the acceptance of her testimony. A similar
issue arises with regard to what we might call testimonial avowals, self-
ascriptions of conscious psychological states that also amount to
testimony. If, following Wittgenstein (1997), one thinks that there are
cases in which one can directly perceive a speaker's psychological
states in her avowal of those states, then in the case of testimonial
avowals it looks like one can acquire either firsthand or secondhand
knowledge of the speaker's states depending on whether one's
knowledge is based on her testimony or on perception of the state
expressed in her avowal. For more on the nature of testimonial avowals,
see McMyler (2011).
(14) Recall that in chapter 1 we saw that Locke distinguishes between

argument and testimony and then proceeds to claim that, while we
often do need to rely on another's testimony, it is her arguments that
are of real epistemic significance. In Locke's case, this downgrading of
the epistemic significance of testimony as compared to argument is a
direct result of his commitment to epistemic autonomy.
(15) This is a case in which knowledge and understanding appear to

come apart. I know that the conclusion of the proof is true, but
nevertheless I don't understand the conclusion in the mathematically
relevant sense. This might be reason to deny that I have “mathematical
knowledge” of the conclusion of the proof, where mathematical
knowledge requires exactly the kind of understanding that is here
absent. Recall that the classical conception of knowledge as episteme or
scientia takes all knowledge to require something like this kind of
understanding, and so it is no surprise that on this conception of
knowledge it is hard to see how genuine knowledge can be acquired
from testimony.
(16) I am here indebted to helpful comments from an anonymous

referee.
(17) For more on the complicated relationship between testimony,

inference, and implicature, see chapter 3.
(18) The issue of how to individuate speech acts is a vexed one. Austin
(1975) seems to individuate speech acts largely in virtue of their
employing different verbs. Later speech act theorists object to this,
choosing to individuate speech acts in terms of general theoretical

constraints on a theory of speech acts. See, for example, Searle (1969)
and (1979). I am not concerned here with the general theory of speech
acts. Instead, I am concerned with the epistemic significance particular
speech acts have for an audience, and so I will here individuate speech
acts in terms of their epistemic significance.
(19) For the nature and role of epistemic communities, see especially
Hardwig (1985) and (1991), Welbourne (1993), and Kusch (2002). I
won't have anything to say about these kinds of cases, but it is a good
and interesting question just how the account of the epistemology of
testimony that I develop here might be extended to meet them.
(20) For an almost exactly contradictory verdict, see Lackey (1999).

Lackey argues that “so far as possible we should offer a unitary account
of the knowledge we acquire from the reports of others . . . For
explaining how we acquire knowledge via testimony is explaining how
we acquire knowledge via the statements of others” (1999: 483). To
assume that explaining how we acquire knowledge via testimony is
explaining how we acquire knowledge via the statements of others is,
however, to beg the question at issue. As we've seen, the category of
statements isn't a unitary epistemic category. Statements that are
produced as the conclusion of argument have a very different epistemic
significance for an audience than statements that amount to tellings.
Lackey puts all of this in the same box as instances of testimony, and
due to this she is forced to give a very thin account of the nature of the
epistemology of testimony, one that is designed to capture a highest
common factor between all of these (epistemically very different)
categories of speech act. On her “dualist” account (2006b), all that is
required for a hearer to acquire testimonial knowledge is (roughly) that
(1) the speaker's words are reliably correlated with the truth and (2)
the hearer has reason to believe that this is so. (Lackey doesn't claim
that these are sufficient conditions, but she does seem to think that
they are pretty close to sufficient.) For Lackey, testimonial knowledge is
not knowledge based on the authority of a speaker. Instead, it is simply
knowledge that is reliably acquired from words. This extremely thin
account of the epistemology of testimony then leads to some very
counterintuitive consequences. On Lackey's account, it turns out that
the acquisition of testimonial knowledge doesn't even require that the
speaker be sincere. These consequences seem to me to be a reductio of
her view. For a general discussion of the kind of position that Lackey
endorses, see chapter 3.
(21) Brewer claims that on Fricker's local reductionist view,

“proceduralized caution attends every rational encounter with
testimony and not just testimonial encounters in the highly formalized
institutional setting of a courtroom. If she is right, legal rationality in
this domain is continuous with rationality more generally” (2006: 127).
(22) One might object that the effect Mary's response to the challenge
has on Alfred's epistemic standing isn't quite as direct as here
portrayed. According to this line of thought, Mary's response to the
deferred challenge is just another piece of evidence, though a rather
significant piece, that Alfred must weigh in deciding whether George
was actually at the party. I address this line of thought in section 2.3
below.
(23) Several philosophers have recognized the existence of something

like this epistemic right of deferral, but I think that they haven't
sufficiently appreciated the implications of this phenomenon for the
epistemology of testimony. Brandom (1983, 1994), for example, claims
that a speaker's assertion that p can license an audience's reassertion
of p in such a way that the audience is entitled to pass the epistemic
buck, to defer the justificatory responsibility for meeting challenges
back to the original speaker (1983: 642; 1994: 174–75). Importantly,
however, Brandom seems to hold not only that a speaker's assertion (or,
better, testimony) that p licenses the audience's reassertion (or, better,
belief) that p but also that it licenses the further assertion by the
speaker and the audience of propositions that can be inferred from p.
As he puts it, “asserting is the issuing of an inference license” (1983:
639). For Brandom, the true social significance of an assertion that p
lies in the way in which it licenses the assertion by the speaker and the
audience of further propositions that are inferable from p (1983: 647;
1994: 168). In section 2.4 below I argue that (leaving aside the case of
conversational implicatures, see section 3.5) an audience is always
solely epistemically responsible for the inferences that it draws from
what a speaker says. Hence, there is no genuinely epistemic sense in
which an audience's inference of further propositions from a
proposition asserted by a speaker can be licensed by the speaker. Talk
of epistemic licensing makes sense in the context of an audience's
coming to know that p on the basis of a speaker's testimony that p, and
hence the audience can here defer challenges to her knowledge that p
back to the original speaker. But an audience is completely
epistemically responsible for the inferences that it draws from what a
speaker has asserted, and therefore the audience cannot defer
challenges to its conclusions based on such inferences. Pace Brandom,
an audience's inferences are not something that can be interpersonally
epistemically licensed. Goldberg (2006) also discusses the phenomenon
of epistemic buck-passing, arguing that it is a distinctive feature of
testimonial belief. However, in contrast to what I argue in section 2.4,
he argues that the phenomenon is actually neutral with respect to the
debate between reductionism and anti-reductionism. Importantly,
Goldberg seems to understand the reductionism/anti-reductionism
debate in such a way that reductionism doesn't involve commitment to
the idea that testimonial justification is inferential. I disagree with this
way of construing the reductionism/anti-reductionism debate, but if
reductionism is understood in this way, then it may in fact be the case
that it is not inconsistent with the phenomenon of epistemic buck-
passing. However, this leaves wide open the question of how the
entitlement to pass the epistemic buck is a result of the way in which
testimonial belief is justified, as it seems it must be if it is a
distinguishing feature of testimonial belief. Goldberg endorses a

version of what in chapter 3 I call the “inheritance model” of epistemic
dependence. According to this model, a testimonial audience is in the
position of inheriting the justification that the speaker has for the belief
expressed in her testimony. This can then give us an account of the
nature of epistemic buck-passing. Buck-passing involves attempting to
point to or access the justification that the speaker has for the belief
expressed in her testimony. I discuss the differences between the
inheritance model of epistemic dependence and the model of epistemic
dependence that I argue for in this book, the “second-personal model,”
in chapter 3.
(24) See especially Moran (2001), Finkelstein (2003), and Bar-On

(2004).
(25) Here is a specific example: As a class is filing out of the room at

the end of a lecture, student A overhears professor P tell student B that
the due date for the final paper will actually be pushed back a week. If
A then goes and tells student C that the paper will be pushed back a
week, and if C challenges A by producing evidence that counts against
this (say, that P rarely changes such things), A then seems to have the
right to defer the challenge directly to P. The reason that A can defer
here without being the direct addressee has to do with the way in
which P, given her role as professor, is conventionally taken to be
responsible for addressing all the students in the class even when she is
only directly addressing one. For further discussion of this kind of case,
see McMyler (forthcoming).
(26) See, for example, Owens (2006), Fricker (2006b), and Lackey
(2008). A notable exception is Hinchman (2005).
(27) For further argument to this effect, see McMyler (forthcoming).
(28) Such an account of testimony should fit quite smoothly into a

socialized virtue-epistemological framework that construes states of
knowledge as resulting from the exercise of cognitive virtues,
competences, or abilities. The capacity for learning from testimony can
be construed as a cognitive virtue, competence, or ability that is
distinctively social in nature insofar as the exercise of the ability
requires cooperative activity between speaker and audience. Greco
(2007), Sosa (2007), and Goldberg (2009) have all recently suggested
that we may need to socialize our account of the cognitive virtues in
order to make sense of the epistemology of testimony.
(29) Of course, this reciprocal requirement only expresses a necessary

condition on the acquisition of testimonial knowledge. The point is that
this necessary condition is the key for understanding the distinctive
way in which testimonial knowledge, unlike any other form of
knowledge, is mediated by the authority of a speaker.
(30) Speech acts are not so neatly individuated by grammatical

locutions. In different contexts a speaker can testify that p by saying “I
know that p,” “I'm absolutely sure that p,” “I believe that p,” “I think
that p,” simply “p,” or even by nodding. Nevertheless, there clearly is a
distinction between giving one's word that p and doing something less
than giving one's word that p, and this is all that I need for my purposes
here.
(31) In this respect, the difference between the speech acts of testifying
and of merely declaring a belief is very much like the difference
between the speech acts of promising and of merely declaring an
intention. In both testifying and promising, a speaker intends the
audience to take her word for it, while in merely declaring a belief or
intention the speaker does not intend the audience to take her word for
it. This, I take it, is the point of the parallel J. L. Austin draws in “Other
Minds” between saying “I know” and saying “I promise,” a parallel
most properly construed as a parallel between the speech acts of
testifying and promising. For a defense of this reading of Austin's
parallel, see McMyler (2011).
(32) One might worry that my account of testimonial knowledge is here

becoming problematically constrained. Shouldn't an adequate account
of testimonial knowledge encompass the knowledge we gain from a
speaker's mere declaration of belief and the knowledge we gain from
treating a speaker's testimony as ordinary evidence? I am not
convinced that it should. The intuitive distinctions that I have drawn
between giving one's word and merely declaring a belief and between
taking someone's word and merely treating what someone says as
reliable evidence are of genuine significance for our ordinary epistemic
practice. We often dispute about whether a speaker has told us that p,
intending for us to take her word, or merely declared to us her belief
that p. If the former, then the question will be one of whether we should
accept the speaker as an authority on the matter. Should we believe
her? If the latter, then the question will be one of whether the speaker's
beliefs on the matter are likely to be reliable. Should we believe what
she says? It is also frequently a matter of concern for us whether an
audience has taken our word for something, thereby believing us, or
simply taken what we say as a reliable indicator of the truth, merely
believing what we say. The account of testimonial knowledge that I

have presented has the virtue of being able to account for these
intuitive distinctions.
(33) The reductionist might continue to insist that the deferral of

challenges is just such a deferral to the evidence. The reductionist
might simply take this to show that the original speaker is not
epistemically responsible for meeting challenges deferred to her,
though she may still be somehow morally or conventionally responsible.
This amounts to a denial that testimonial knowledge is secondhand in
the demanding sense of (S). This kind of objection accepts that (R) is
inconsistent with (S), but rejects (S). I address this kind of objection in
section 2.5.
(34) Craig (1990) distinguishes between “informants” and mere

“sources of information,” and he argues that the point of the concept of
knowledge is to identify informants. If informants are understood as
theoretical authorities, then the point of the concept of knowledge may
be to identify such authorities, persons in a position to transmit
knowledge by giving us their word. While Craig's work on the point of
the concept of knowledge has been influential, if controversial, his
distinction between informants and mere sources of information has
been largely neglected.
Three Models of Epistemic Dependence

Benjamin McMyler

Print ISBN-13: 9780199794331
DOI: 10.1093/acprof:oso/9780199794331.001.0001

Benjamin McMyler
DOI:10.1093/acprof:oso/9780199794331.003.0004

This chapter distinguishes between three competing models for
understanding the way in which, in believing something on the basis of
testimony, an audience is epistemically dependent on a speaker and her
testimony. According to the Evidential Model, the audience is
epistemically dependent on the speaker for providing the audience with
evidence; according to the Inheritance Model, the audience is
epistemically dependent on the speaker for expressing a belief the
justification of which can be inherited; and according to the Second-
Personal Model, the audience is epistemically dependent on the
speaker for assuming a kind of second-personal relationship towards
the audience. Chapter 2 defended a version of the Second-Personal
Model. This chapter examines what is at stake between the Second-
Personal Model and the Evidential and Inheritance Models.
Keywords: testimony, justification, sincerity, trust, the second person

The previous chapter developed and defended an essentially

interpersonal account of the epistemology of testimony. This involved
shifting our focus away from the narrow concern with justification
present in the debate between reductionists and anti-reductionists and
towards a broader appreciation of the interpersonal relations of
authority and responsibility that make learning from testimony a
distinctively social epistemic capacity. In this chapter I want to further
articulate the nature and significance of this shift of epistemological
focus by placing the interpersonal account of testimony developed in
the previous chapter in the context of an epistemological debate
decidedly different from that between reductionists and anti-
reductionists. In recent years a debate concerning the epistemology of
testimony has begun to emerge that largely cuts across the traditional
divide between reductionists and anti-reductionists. Unfortunately, the
contours of this new debate have yet to be sufficiently appreciated, and
so in this chapter I want to introduce this new debate and begin to
develop a general framework in which to understand what is at stake
between the disputing parties.
This new debate might be described as a debate concerning epistemic

dependence—more specifically, concerning how exactly it is that in
acquiring knowledge and justified belief based on testimony an
audience is epistemically dependent on a speaker and her testimony.
Three different models of epistemic dependence have been emerging in
the recent literature. According to what I will call the evidential model
of epistemic dependence, when an audience acquires knowledge or
justified belief based on a speaker's testimony, the audience is
epistemically dependent on the speaker for providing the audience with
evidence. In testifying that p, a speaker provides an audience with
ordinary evidence that (p.78) counts in favor of p, and the audience is
thus dependent on the speaker for making this evidence so available.
According to what I will call the inheritance model of epistemic
dependence, when an audience acquires knowledge or justified belief
based on a speaker's testimony, the audience is epistemically
dependent on the speaker for expressing a belief the justification of
which can be inherited by the audience. In testifying that p, a speaker
is in the position of passing along the justification that she has for her
own belief that p, and the audience's belief that p is thus dependent for
its justification on the justification inherited from the speaker's belief.
Finally, according to what I will call the second-personal model of
epistemic dependence, when an audience acquires knowledge or
justified belief based on a speaker's testimony, the audience is
epistemically dependent on the speaker for assuming a kind of second-
personal relationship towards the audience. In testifying that p, a
speaker is assuming a particular epistemic responsibility towards the

audience, and it is in virtue of this assumption of responsibility that the
audience's belief based on the speaker's testimony is justified.1
These three models can be further distinguished by noting that they

each appeal to a different metaphysical category in explaining what an
audience is ultimately epistemically dependent upon in knowing based
on a speaker's testimony. According to the evidential model, the
audience is ultimately epistemically dependent on a particular kind of
event, on a speaker's speech act—her statement, assertion, or telling. It
is features that properly belong to the event, particularly its reliability,
that do the distinctive work of justifying the audience's testimonial
belief. According to the inheritance model, the audience is ultimately
epistemically dependent on a particular kind of state, on a speaker's
belief. It is features that properly belong to the state, particularly its
justification, that do the distinctive work of justifying the audience's
belief. According to the second-personal model, the audience is
ultimately epistemically dependent on a person, on the speaker herself.
It is features that properly belong to the person, particularly the
speaker's relationship to the audience, that do the distinctive work of
justifying the audience's belief.
Now, adherents of all three models agree that in acquiring knowledge

and justified belief based on a speaker's testimony, an audience is
typically epistemically dependent on a speech act that is expressive of
the beliefs of a speaker. Where the models disagree, however, is over
what is doing the distinctive epistemological work here. The models
disagree about what ultimately does the work of epistemically
supporting the audience's testimonial belief. As in the debate between
reductionists and anti-reductionists, the issue of justification is here
central, but the question is not in the first instance one of whether or
(p.79) not the relevant justification is inferential. The question is rather
one of what it is that is doing the ultimate justificatory work, be it either
inferentially, by figuring as a premise in an argument, or
noninferentially, by, for example, being part of a properly reliable belief-
forming mechanism. The models disagree over whether an audience's
testimonial knowledge and belief is ultimately justified by a particular
kind of event (an act of a person that serves as evidence for some fact)
by a particular kind of state (a belief of a person the justification of
which can be inherited), or by a person herself (a speaker who stands
in a particular kind of relationship towards an audience).
Before we move on, I should say a bit more about the notion of
epistemic dependence itself. Taken alone, the notion of epistemic
dependence is horribly vague. Beliefs bear all sorts of relations of
dependence to other beliefs, to experiences, and to states of affairs.

Applied to the epistemology of testimony, however, the idea is
somewhat clearer. Intuitively, knowledge and belief based on testimony
appear to exhibit a kind of epistemic dependence that knowledge and
belief based on other epistemic sources do not. It is far from clear what
exactly this dependence involves—as we saw in chapter 1, this
dependence was traditionally taken to have something to do with the
notion of authority—but whatever the case, the fact of this dependence
is what makes testimony a distinctive and interesting topic for
epistemology. Testimony is standardly taken to be one of the typical
ways in which knowledge and justified belief can be acquired, but what
merits its inclusion alongside the sources of perception, memory, and
inference is that there seems to be something distinctive about it,
something aptly captured by talk of epistemic dependence. Any
particular model of the kind of epistemic dependence pertinent to
testimony should therefore be evaluated in terms of how well it
captures the distinctiveness of testimony as a source of knowledge and
justification.
Section 3.1 examines the evidential model. It argues that the model has
difficulty explaining how epistemic dependence on a speaker's
testimony is relevantly different from epistemic dependence on other
ordinary instruments and events. Section 3.2 argues that the
inheritance model fares better in this regard but that it also seems to
misconstrue what it is that does the work of epistemically supporting
an audience's testimonial belief. Sections 3.3 through 3.5 then examine
how the second-personal model fares in solving the problems facing the
evidential and inheritance models.
In chapter 2, I offered a positive argument for what I am here calling

the second-personal model of epistemic dependence. In this chapter my
aim is to illustrate what is important and distinctive about this model by
contrasting it with its main rivals. In so doing I will offer some
considerations that I believe tell against the evidential and inheritance
models and in favor of the second-personal model, but my goal in this
chapter will not be to provide a positive argument for the second-
personal model. My aim in this chapter is to illustrate the nature of
these three models of epistemic dependence and to clarify what is at
stake between them.
(p.80) 3.1 The Evidential Model

The evidential model is by far the most popular model of epistemic
dependence present in the contemporary literature. Most philosophers
writing on the epistemology of testimony, both reductionists and anti-
reductionists alike, endorse some form of the evidential model, and it
thus exhibits the most variety of any of the three models that we will
consider here. What all varieties of the evidential model have in
common, however, is the idea that an audience's testimonial knowledge
and belief is ultimately epistemically dependent on a particular kind of
communicative event that serves as evidence for a particular fact. The
ultimate epistemic significance of a speaker's testimony that p is as an
event that amounts to evidence counting in favor of p.
Varieties of the evidential model differ both with respect to their view
of the exact nature of the communicative event involved and with
respect to their view of the way in which this communicative event
serves to justify an audience's belief. On the one hand, some advocates
of the evidential model endorse a very broad conception of the
communicative event that justifies an audience's belief, while some
endorse a much narrower conception. On the other hand, reductionist
advocates of the evidential model think that testimonial knowledge and
belief is justified by the strength of an inference from the relevant
communicative event to the proposition believed, anti-reductionists
think that it is justified noninferentially, and hybrid reductionist/anti-
reductionists think that it can be justified either inferentially or
noninferentially. I will not discuss all of these variations in any detail
here. Since my concern is with the general contours of the evidential
model itself, I want to discuss what I take to be the most austere
example of the model available in the literature, that present in some
recent work by Jennifer Lackey. Few adherents of the evidential model
actually share Lackey's particular conception of the epistemology of
testimony. Nevertheless, her work serves to illuminate some of the core
tenets of the evidential model by, in effect, taking them to their logical
conclusion.
Lackey argues for what she calls the statement view of testimony.
According to the statement view:
The process of communicating via testimony does not involve a

speaker transmitting her belief to a hearer, along with the
epistemic properties it possesses. Instead, a speaker offers a
statement to a hearer, along with the epistemic properties it
possesses, and a hearer forms the corresponding belief on the
basis of understanding and accepting the statement in question.

Statements are not, therefore, merely vehicles for expressing
beliefs but, rather, they are the central bearers of epistemic
significance themselves. (2006c: 93)
Lackey develops her statement view in opposition to what she calls the
belief view of testimony. According to the belief view, “while statements
are necessary for the process (p.81) of communication, they are merely
vehicles for expressing beliefs—they enable us to make public what
would otherwise remain private. Strictly speaking, then, we do not
learn from another's words—we learn from another's beliefs” (2006c:
77).2 Lackey presents a pair of artfully constructed cases that are
supposed to amount to counterexamples to the belief view. One case is
supposed to show that, as long as a speaker's statement is reliable, the
reliability of the speaker's beliefs is unnecessary for an audience to
acquire testimonial knowledge. The other case is supposed to show that
the reliability of the speaker's beliefs is insufficient. Together, the cases
are designed to show that it is the reliability of the speaker's
statements that is of crucial epistemological significance, regardless of
the reliability of her beliefs.
I will only consider the necessity claim here. Lackey argues that as long
as a speaker's statement is suitably reliable, an audience can acquire
testimonial knowledge from the statement even though the speaker's
beliefs are themselves an utterly unreliable guide to the truth. In order
to do this she presents an example that she dubs CONSISTENT LIAR
(2006c: 82–83). In CONSISTENT LIAR, a girl named Bertha suffers a
head injury that makes her prone to tell lies about her perceptual
experiences concerning wild animals. Upon observing this, her parents
take her to a neurosurgeon, Dr. Jones, who discovers a lesion on her
brain that is the cause of Bertha's lying. Unable to repair the lesion, Dr.
Jones instead creates a new lesion on Bertha's brain that causes her
beliefs about wild animals based on her perceptual experiences to be
consistently aligned with her pattern of lying. So every time that Bertha
sees a deer, she believes that it is a horse, (p.82) but every time she
believes that it is a horse, she insincerely says that it is a deer. Bertha is
thus an utterly unreliable believer with respect to the presence of wild
animals, but she is nevertheless an extremely reliable testifier. Lackey
stipulates that Dr. Jones never tells anyone of the procedure he has
performed on Bertha and that Bertha comes to be regarded in her
community as extremely trustworthy even with respect to such things
as the presence of wild animals. Henry, Bertha's next-door neighbor,
can thus readily acquire knowledge of the presence of wild animals
based on Bertha's testimony even though Bertha doesn't actually

believe what she asserts.
Lackey claims that the upshot of CONSISTENT LIAR is that “the

statements of speakers are not only the basis, both causally and
epistemically, of the beliefs that hearers acquire via testimony, they are
also the bearers of epistemic significance” (2006c: 86, original
emphasis). Henry can acquire knowledge from Bertha's testimony even
though Bertha's beliefs themselves are utterly unreliable. In order to
have a unified epistemology of testimony, one that can encompass cases
like CONSISTENT LIAR, we thus need to see the justification of
testimonial beliefs even in ordinary, non-science-fiction cases as
deriving from the statements of speakers rather than from their beliefs.
The attempt to provide the materials for a unified theory of testimony is

an over-riding theme of much of Lackey's work. A unified theory of
testimony, for Lackey, is one that encompasses cases like that of
CONSISTENT LIAR, cases that Lackey thinks are obvious cases of
testimonial knowledge. However, one might question whether
testimonial knowledge is actually acquired in CONSISTENT LIAR and
thus whether the unity Lackey is seeking is a false one. An audience
might be able to acquire knowledge and justified belief from Bertha's
statements, but (1) it is questionable whether knowledge can be
acquired in the case as Lackey actually presents it, and (2), if we
modify the account to solve for this, it is questionable whether the kind
of knowledge that can be acquired from Bertha's statements is actually
knowledge based on testimony. I will address each of these points in
turn.
First, as Lackey presents the case, no one but Dr. Jones (and possibly
Bertha herself) knows that Bertha's statements are insincere. Henry
believes, and has good reason to believe, that Bertha's statements with
respect to the presence of wild animals are completely sincere, and
hence, as Lackey puts it, Henry has “trust in Bertha” (2006c: 83).
Henry's beliefs based on Bertha's testimony involve trusting Bertha,
and they therefore amount to standard cases of belief based on
testimony. However, insofar as Bertha's testimony is insincere, she
appears to be abusing Henry's trust. Henry's trust in Bertha is
predicated on the belief that Bertha is sincere such that, if Henry were
to learn that Bertha does not in fact believe what she is saying, he
would presumably cease to trust her. Henry's beliefs based on Bertha's
testimony thus appear to be suspect. Something has gone awry with the
process through which Henry has formed his beliefs such that, if he
were apprised of the way in which this process has gone awry, he would
give up his belief.
(p.83) Lackey thinks that nothing has gone awry with the process
through which Henry has formed his beliefs. This is because Bertha's
statements, though insincere, are completely reliable and deliberately
designed to be so. Due to this, it is no accident that Bertha's statements
are reliable guides to the truth. Lackey thus denies that CONSISTENT
LIAR amounts to a kind of Gettier case, a case in which Henry acquires
a justified true belief that fails to amount to knowledge due to the
presence of some form of accidentality in the way in which the belief
was formed. In fact, what Lackey has done is offer a slight variation on
a very standard kind of Gettier case, a case that we might call
UNLUCKY LIAR. In UNLUCKY LIAR a speaker, intending to deceive her
audience, insincerely tells an audience that p, but it just so happens
that, unfortunately for the speaker, p is in fact true. If the audience
believes the speaker's testimony and has every reason to believe that
the speaker is in fact trustworthy, then the audience can acquire a
justified true belief that p. However, due to the way in which this belief
was formed, it doesn't appear to amount to knowledge. UNLUCKY LIAR
thus amounts to a kind of Gettier case. The reason that UNLUCKY LIAR
amounts to a Gettier case is standardly taken to be that it is a mere
accident that the speaker's testimony, and hence the audience's belief
based on this testimony, is true. The audience was simply lucky, and the
liar unlucky. In effect, what Lackey does is invent a way of eliminating
this element of luck. CONSISTENT LIAR differs from UNLUCKY LIAR
only in that Lackey has added a way in which the speaker's insincere
statements don't just happen to be true.
In large part because of the role that Dr. Jones plays in Bertha's
condition, there is no relevant accidentality in the belief that
Henry forms on the basis of her testimony and, hence,
CONSISTENT LIAR cannot plausibly be regarded as a GETTIER-
type case. With respect to the truth of the belief in question, it is
neither an accident that Bertha reported that there was a deer on
the hiking trail nor that Henry came to hold this belief on the
basis of her testimony. For, because of her two brain lesions,
Bertha is such that nearly every time she sees a deer, she believes
it to be a horse, yet reports that it is a deer. Thus, Bertha's deer-
reports nearly always covary with her deer-sightings, despite
taking a slight detour through horse-believings. Furthermore,
given all of Henry's excellent inductive evidence on behalf of
Bertha's testimonial practices, there is also no relevant
accidentality in his coming to form the belief in question on the

basis of her report. (2006c: 84–85)
Given that CONSISTENT LIAR eliminates the accidentality involved in

UNLUCKY LIAR, Lackey thinks that CONSISTENT LIAR does not
amount to a Gettier case.
Still, it is questionable whether CONSISTENT LIAR solves for

everything that is actually problematic about UNLUCKY LIAR. The
problem with UNLUCKY LIAR (p.84) may be not only that the speaker's
testimony just happens to be true but also that it is insincere. It is
plausible that the audience's testimonial belief in UNLUCKY LIAR is
based on more than the mere reliability of the speaker's statement. The
audience's belief is justified by the speaker's testimony in such a way
that, even if the speaker's statement nonaccidentally co-varies with the
truth, the insincerity of the speaker's testimony is still enough to render
the process by which the audience's belief is formed problematic. In
this sense, even though CONSISTENT LIAR is clearly unlike the
standard barn-façade case, it might nevertheless be similar to other
kinds of Gettier cases, including Gettier's original cases where a
justified true belief fails to amount to knowledge by being based on a
false belief (Gettier 1963). In fact, if one is a reductionist about
testimony, one will think that Henry's belief based on Bertha's
testimony is justified by the strength of an inference from Bertha's
statement, through independently available considerations concerning
her trustworthiness, to the conclusion that what she says is true. If one
thinks that such an inference requires as a premise that the speaker is
sincere, then Henry's testimony looks to be based on an inference from
a false belief. Furthermore, even if one isn't a reductionist about
testimony, it is not generally clear that reliability in a belief-forming
process is sufficient to render beliefs that are the outcome of the
process immune to Gettier problems. Consider, for example, the
following Gettier case for reliabilism recently presented by Duncan
Pritchard:
Imagine, for example, that you find out what the temperature of
the room is by looking at the thermometer on the wall.
Furthermore, let us grant that this thermometer is very reliable in
this respect in that it will enable you to form accurate beliefs
about what the temperature is. Suppose, however, that
unbeknownst to you someone is playing a trick on you. The
thermometer is, in fact, broken and is fluctuating randomly.
Crucially, however, this isn't making the thermometer an
unreliable indicator of what the temperature in the room is for the
simple reason that someone is hidden in the room and adjusting
the temperature of the room to match whatever reading is on the
thermometer whenever she sees you look at the thermometer (we
won't concern ourselves with why). Accordingly, in this case you
are forming true beliefs about what the temperature of the room
via a method—looking at the thermometer—that is entirely
reliable, since every time you form a belief about what the
temperature in the room is by looking at the thermometer that
belief will be true. Intuitively, however, you don't know what the
temperature of the room is because the thermometer is broken
and you can't find out the temperature by looking at the
thermometer. (2006: 63)
This case is remarkably like Lackey's CONSISTENT LIAR in that it

involves a belief forming process that is intuitively faulty but that has
nevertheless been made to be reliable (p.85) by the machinations of
someone working behind the scenes. Despite the reliability of the
process through which your beliefs about the temperature were
formed, Pritchard holds that intuitively this is not a case of knowledge.
Similarly, despite the reliability of the process through which Henry's
testimonial beliefs were formed, it's plausible that these beliefs do not
amount to knowledge. Just as you can't acquire knowledge of the
temperature of the room by looking at a broken thermometer, so you
can't acquire testimonial knowledge by trusting a broken testifier. This
is all supported by the fact that, if Henry were to learn (1) that Bertha's
testimony is insincere but (2) that her statements nevertheless reliably
co-vary with the truth, it seems that at the very least he would not
continue to trust Bertha.3 He might very well go on to use Bertha's
statements as reliable guides to the truth, and he might very well
acquire knowledge thereby, but it is not at all clear that this involves
trusting Bertha. And if it is not clear that this involves trusting Bertha,
then it is not clear that this should be construed as a straightforward
instance of the acquisition of testimonial knowledge.4
This brings us to the second point. It is telling that Lackey does not
formulate CONSISTENT LIAR such that Henry is aware of Bertha's
condition. Lackey formulates the case such that Henry believes Bertha
to be sincere, and I have argued that this renders his belief based on
her testimony suspect. If we solve for this, however, and stipulate that
Henry does know about Bertha's condition, then this just makes it
questionable whether, in believing things on the basis of Bertha's
statements, Henry is trusting her in the way that, at least intuitively,
seems to be a distinguishing feature of the category of testimonial
knowledge. We can certainly treat the testimony of a speaker as a good
reason for belief, even a good reason for belief in the content of the
testimony, without trusting the speaker. For example, we might take the
speaker to be “double bluffing,” to be sincerely and truly telling us that
p but expecting that we will believe the opposite and thereby intending
to deceive us into believing that not-p (Anscombe 1979). If we see
through this, we may very well acquire knowledge that p thereby, but
surely it is counterintuitive to say that this involves trusting the
speaker. (p.86) The case of CONSISTENT LIAR seems even worse in
that Bertha's statements are not even sincere—they do not express her
actual beliefs. It thus seems at least equally implausible that, in seeing
through what has been done to Bertha and believing what she says on
this basis, one would be trusting Bertha. So if testimonial knowledge
and belief distinctively involves trusting others, then testimonial
knowledge cannot be acquired from Bertha and a properly unified
account of testimonial knowledge ought actually to exclude
CONSISTENT LIAR.
This second point will be an issue for any version of the evidential
model. Lackey's account is extremely austere in that it takes the
relevant event on which an audience is epistemically dependent to be a
speaker's statement, a speech act that requires only “the intention to
express communicable content” (2006c: 84), not the intention to
actually cause an audience to believe anything in particular.5 For
Lackey, a soliloquy intended to be said in private, insofar as it is
nevertheless intended to express truth-evaluable content that could be
used for the purposes of communication, still amounts to testimony.
Many philosophers writing on testimony would object to this conception
of the relevant event on which an audience is epistemically dependent
while still being attracted to the evidential model. Fricker, for example,
has characterized the event on which a testimonial audience is
epistemically dependent as “serious assertions aimed at
communication” (1994: 137), and more recently she has referred to
such events as “tellings” (2006b). Tellings are a much narrower
category of speech act than Lackey's statements. A soliloquy said in
private does not amount to a telling.6 Still, an advocate of the evidential

model takes a speaker's speech act, however the relevant speech act is
circumscribed, to be the ultimate source of justification for an
audience's testimonial belief, and this in itself makes it difficult to see
how such a belief distinctively involves trust. Insofar as the event is
taken to be the fundamental unit of epistemic significance, it is not
clear how epistemic dependence on such an event is relevantly
different from epistemic dependence on any other kind of ordinary
event, and insofar as this is not clear, it is not clear how beliefs based
on testimony exhibit a distinctive kind of epistemic dependence.
(p.87) A staunch advocate of the evidential model will happily embrace

this, claiming that there isn't anything of much epistemological
consequence that distinguishes beliefs based on testimony from beliefs
based on other kinds of ordinary evidence. This can then be held up as
a vindication of testimony as a perfectly legitimate, in virtue of being
perfectly ordinary, source of belief. It domesticates the epistemology of
testimony by playing down the differences between testimony and other
sources of inferential and noninferential belief. Importantly, however,
such domestication comes at the price of eliminating what is important
and interesting about the epistemology of testimony. Plausibly, beliefs
based on testimony are important and interesting precisely because
they exhibit a kind of epistemic dependence that beliefs based on other
kinds of ordinary evidence do not.
3.2 The Inheritance Model

Adherents of both the inheritance and the second-personal models
object to the evidential model on the grounds that it doesn't provide an
adequate account of the way in which the epistemic dependence
characteristic of testimonial knowledge and belief is distinctive. The
two models disagree, however, over just how it is that testimonial
dependence is so distinctive. According to the inheritance model,
testimonial belief is distinctive due to the way in which it inherits the
justification possessed by the speaker's belief. According to the second-
personal model, testimonial belief is distinctive due to the way in which
it is justified in virtue of the actual interpersonal relationship existing
between speaker and audience. I will consider the inheritance model in
this section and the second-personal model in section 3.3.
One of the clearest examples of the inheritance model is presented by

David Owens (2000).7 Instructively, Owens argues against the evidential
model by presenting an example that is actually quite similar to
Lackey's CONSISTENT LIAR. Owens asks us to consider a patient who
is hypnotized into believing that a particular woman was murdered.8
The woman was indeed murdered, and the hypnotist, knowing this fact,
deliberately induces this belief in the patient in an attempt to bring this
fact to light without having to tell anyone about it himself. The case is
similar to Lackey's in that, though the patient's statements are a
reliable guide to the truth, something has gone wrong with her beliefs.
Unlike Bertha, the patient's statements in Owens's example are in fact
sincere—they do express her genuine (p.88) beliefs—but due to the fact
that the patient's beliefs have been induced via hypnosis, her beliefs
are unjustified.9
Owens claims that, on the one hand, if I am not aware of the patient's
hypnotized condition, then I cannot acquire knowledge from her
assertion. “Here the patient's words are a reliable guide to the truth
but they do not give me knowledge of the murder” (2000: 166). On the
other hand, if I am aware of the patient's condition, then I am in a
position to acquire knowledge from the patient's assertion, but such
knowledge is not properly testimonial. In such a case, “I learn
something from his words, but nothing from him” (2000: 166). So in
contrast to Lackey, Owens presents this case as a counterexample to
the evidential model. Owens thinks that testimonial knowledge clearly
cannot be acquired in the case of the hypnotized patient, and he thinks
that the evidential model must therefore be mistaken.
The reason Owens thinks that testimonial knowledge cannot be

acquired in the case in which the audience is in fact aware of the
hypnotized patient's condition is that, in such a case, the audience's
believing the patient's testimony amounts to nothing more than the

audience's treating her assertion as a reliable indicator of the facts. A
speaker's words can certainly amount to reliable indicators of the facts,
and knowledge can certainly be acquired from reliable indicators, but
according to Owens, such knowledge is not properly testimonial. Owens
thus opposes the evidential model on the grounds that it doesn't
properly distinguish testimonial knowledge from knowledge based on
other kinds of ordinary instruments and events:
A linguistic assertion has a sort of meaning which the state of a

gauge or a cloud lacks. Statements express beliefs with a certain
content, and unlike instruments or clouds, those beliefs are
subject to reason. To treat someone as a source of testimony is to
treat him as a believer, as a person who adopts convictions for
reasons which are more or less appropriate to the content of
those convictions and then attempts to convey what he knows in
speech (2000: 166).10
(p.89) For Owens, properly testimonial knowledge is knowledge gained

from the speaker rather than from the speaker's words, and what it
means for testimonial knowledge to be gained from the speaker has to
do with the way in which acquiring testimonial knowledge requires
treating the speaker as a believer, as one who adopts beliefs on the
basis of reasons.
Building on work by Tyler Burge (1993), Owens's account of what it is

to properly acquire testimonial knowledge from a speaker is crucially
dependent on the idea that the justification of states of belief is
inheritable. If an audience believes that p on the basis of a speaker's
sincere expression of the belief that p, then, according to Owens, the
audience's belief inherits whatever justification there is for the
speaker's own belief. Owens holds that testimony, like memory, is
essentially a rationality preserving mechanism. Testimony contributes
to the justification of an audience's belief by serving as a causal
mechanism through which the audience's belief can inherit the
justification pertaining to the speaker's belief. Owens states that “to
believe something on the basis of testimony is to put yourself in the
power of others, not just in respect of whether you have knowledge but
equally in respect of the rationality of your belief” (2000: 171). A
testimonial audience is entitled to defer to the speaker for the
justification of its testimonial belief, and the speaker is thus, in some
sense, responsible for the justification of the audience's belief. It is
important to note, however, that what an audience is here dependent
upon when it is in the power of a speaker is the justification of the
speaker's own belief, and what the speaker is responsible for when it is
responsible for the audience's belief, is the justification of her own

belief. If the speaker's belief is not itself justified, then there is no
justification present for the audience's belief to inherit. The audience's
dependence on the speaker is thus ultimately a dependence on the
rationality of the speaker's belief. According to Owens, this is exactly
the problem with the hypnotized patient's testimony about the woman's
death: “To treat the patient's words as testimony about the death is to
defer to his reasons for belief on the matter. But the patient has no
reason to believe the woman was murdered. Therefore I have no reason
to believe the woman was murdered, and so I don't know that she was
murdered” (2000: 166). Testimony cannot transmit justification that
was never there in the first place.
When it comes to Lackey's CONSISTENT LIAR, Owens will have to say

something slightly different. Again, CONSISTENT LIAR differs from
Owens's case of the hypnotized patient in that Bertha's statements are
insincere. Due to this, the audience in CONSISTENT LIAR is even
worse off than the audience in Owens's case of the hypnotized patient.
In the case of the hypnotized patient, epistemic responsibility has
indeed been transferred from speaker to audience. The speaker's
sincere testimony has put the audience in the position to inherit the
justification of the speaker's belief. It is just that, insofar as the
speaker's belief is unjustified, there is no justification there to inherit.
In (p.90) CONSISTENT LIAR, however, the audience has not even been
put in the position to inherit the justification of the audience's belief.
Insofar as the speaker's testimony is insincere, epistemic responsibility
for the justification of the audience's belief has not been transferred
from the speaker to the audience. A speaker's insincere testimony
merely purports to transfer such responsibility, and hence no
justification can be inherited from such an insincere expression of
belief. Owens thus argues that both lying and cases of ordinary
misunderstanding should be classed together as “illusory transfers of
epistemic responsibility” (2006: 124).
For Owens, the role that the speaker plays in justifying an audience's
belief is one of publishing her belief, of putting her belief out into the
public domain so that the justification pertaining to the belief can be
picked up by a comprehending hearer.11 Strictly speaking, then, it is
not the speaker's testimony that justifies an audience's testimonial
belief but rather the justification pertaining to the speaker's own belief.
One can question the inheritance model on this very score. If I come to
believe that p on the basis of a speaker's testimony, and if someone then
asks me what justification I have for believing that p, I will likely say
“So-and-so told me that p.” If queried for my justification, my natural
response will be to cite the speaker's testimony. Intuitively, the

speaker's testimony, her act of coming out and telling me that p,
appears to be what it is that justifies my testimonial belief, regardless
of whatever exactly it is that justifies the speaker's own belief.
This is particularly striking in cases where the justification for the

speaker's belief is a priori. Imagine a case in which a mathematician
deductively proves a mathematical theorem and then tells me that the
theorem is true. Owens claims that in such a case I can acquire a belief
based on the very grounds on which the mathematician's belief is based
even though I have never gone through the proof of the theorem myself
(2006: 120). This, however, seems highly implausible. The
mathematician's belief in the truth of the theorem is justified a priori,
but it doesn't seem that my testimonial belief in the truth of the
theorem is thereby justified a priori as well. My belief appears to be
justified a posteriori by the mathematician's testimony, and hence if
queried for my justification I will cite the mathematician's testimony.
The mathematician's testimony appears to be what justifies my belief in
the truth of the theorem, not the proof of the theorem itself.
(p.91) According to the inheritance model, this can't be strictly

speaking correct. According to the inheritance model, what is really
going on when I cite a speaker's testimony is that I am deferring to,
pointing to, or attempting to access the speaker's own reasons for
belief on the matter. I am not citing the speaker's testimony as what
justifies my belief, but rather pointing to where the ultimate
justification of my belief can be located, a justification that the
speaker's testimony simply makes available or accessible to me.
Testimony does not strictly speaking justify an audience's belief but
rather is a vehicle for the expression of a state of belief with an
inheritable justification. In the case of the mathematician, the
mathematician's testimony simply makes available to me the
justification provided by the proof itself. The mathematician's testimony
itself does not justify my belief in the truth of the theorem.
An advocate of the evidential model will think that this is highly

implausible. According to the evidential model, what justifies an
audience's testimonial belief is not the same as what justifies the
speaker's own belief. The audience's belief is justified instead by the
speaker's testimony itself, by an event (a speech act) that serves as
evidence for some fact. An adherent of the second-personal model will
also object here, though along very different lines. According to the
second-personal model, a speaker's testimony is not a mere vehicle for
inheriting justification. It plays a much more direct role in the
justification of the audience's belief. However, the role that it plays is
not best captured on the model of ordinary evidence, at least if our

model of evidence is one according to which evidential considerations
are considerations that a subject is always in the position of coming to
her own conclusion about. On the second-personal model, a speaker's
coming out and telling an audience that p directly justifies an
audience's belief that p, but it does so in virtue of amounting to an
assumption of epistemic responsibility for the speaker's belief.
3.3 The Second-Personal Model

The account of the epistemology of testimony developed in chapter 2 is
an example of what I am here calling the second-personal model of
epistemic dependence. According to the second-personal model, cases
of testimonial knowledge are cases of knowledge that involve, as
Anscombe puts it, trusting a speaker for the truth (1979: 151). As
theorists of trust have noted, trusting another person involves more
than merely relying on the person. I can rely on a person without
trusting her. At the very least, trusting another involves, as Annette
Baier (1994) puts it, reliance on the goodwill of the other.12 Applied to
the case of testimony, this means that trusting another for the truth
involves more than merely relying on the other to reliably say true
things. At the very least, trusting (p.92) another for the truth involves
relying on the other to say true things out of the other's goodwill
towards one. In the absence of a speaker's goodwill towards an
audience, properly testimonial knowledge and belief—knowledge and
belief that involves trusting a speaker for the truth—cannot be
acquired.
According to the second-personal model, this distinction between trust

and mere reliance is crucial for understanding the epistemology of
testimony. Adherents of the second-personal model readily admit that
knowledge can be acquired from mere reliance on a speaker's
testimony—they readily admit that an audience can treat a speaker's
testimony as ordinary evidence—however they insist that this does not
involve trusting a speaker for the truth. To the extent that an audience
is treating a speaker's testimony as ordinary evidence, the audience
appears to be in the position of coming to its own conclusion about
things, but intuitively, not all cases of the acquisition of knowledge and
belief involve an audience's coming to her own conclusion about things.
In many cases an audience simply takes the speaker's word for it—the
audience accepts what the speaker says on the basis of her authority
and, so it seems, trusts her for the truth. These appear to be cases in
which the audience is not coming to her own conclusion about things,
and so adherents of the second-personal model think that such cases of
the acquisition of testimonial knowledge and belief are deserving of a

different epistemological treatment from those cases in which an
audience merely relies on the speaker to say true things.
Importantly, in denying that testimony amounts to ordinary evidence,

adherents of the second-personal model are denying that testimony
amounts to evidence only in a particular sense. They are denying that
testimony amounts to evidence only in the sense relevant to the
evidential model. They are not denying that testimony amounts to a
genuinely epistemic reason for belief. We must keep in mind that there
are a variety of ways in which the concept of evidence is frequently
employed. In ordinary language, the term ‘evidence’ is often used to
refer to something from which one can only draw inferences.13 In this
sense, the evidence left at a crime scene is something from which a
(p.93) detective must draw her own conclusions. As we saw in chapter
1, this is what Hacking calls “inductive evidence.” However,
philosophers often use the term ‘evidence’ in a much broader fashion
than the ordinary language notion of inductive evidence. In the
broadest philosophical sense, ‘evidence’ simply refers to any kind of
genuinely epistemic reason for belief, any kind of consideration that
genuinely counts in favor of the truth of, confirms, or probabilifies a
proposition.
Adherents of the second-personal model hold that a speaker's testimony

that p genuinely counts in favor of the truth of p. They thus accept that
testimony amounts to evidence in the broadest philosophical sense.
What they deny is that testimony counts in favor of the truth of a belief
either by amounting to inductive evidence or by amounting to
something analogous to the noninductive evidence provided by
perceptual representation. Both inductive and perceptual evidence are
considerations a subject is in the position of coming to her own
conclusion about, considerations that put a subject in a position to infer
or perceive things for herself. But according to the second-personal
model, testimony functions very differently. A speaker's testimony that
p is a consideration that genuinely counts in favor of the truth of p, but
the way in which it does so does not put a subject in a position in which
she is coming to her own conclusion about things.
According to the second-personal model, the real problem of the

epistemology of testimony is thus one of explaining how knowledge and
justified belief is acquired in cases in which an audience is not coming
to her own conclusion about things, cases in which the audience is
trusting the speaker for the truth. Like theorists of trust generally,
adherents of the second-personal model think that this narrow class of
genuinely trust-involving cases are cases that involve something like
reliance on the goodwill of the other. When it comes to epistemic cases

of trusting a speaker for the truth, this goodwill is cashed out in terms
of epistemic responsibility. The goodwill that a speaker must have
towards an audience in order for the audience to acquire genuinely
testimonial knowledge or belief is a matter of the speaker's assuming
an epistemic responsibility towards the audience. To testify is to
assume such an epistemic responsibility, and to trust a speaker for the
truth is to rely epistemically on this assumption of responsibility on the
part of the speaker.
In chapter 2, I argued that this assumption of epistemic responsibility

should be understood as an assumption of the responsibility to meet
certain epistemic challenges to the audience's testimonial belief. If an
audience comes to believe that p on the basis of a speaker's testimony,
and if a third party then challenges the audience's belief by producing
evidence that tells against p, the audience is entitled to defer the
challenge back to the original speaker. Again, the audience is not
entitled to defer all kinds of challenges back to the original speaker. If a
third party challenges the audience's testimonial belief by producing
evidence that tells against the trustworthiness of the (p.94) speaker,
then the audience is not entitled to defer the challenge. In this sense,
an audience is completely epistemically responsible for coming to its
own conclusion about the competence and sincerity of the speaker.
However, the audience is not completely epistemically responsible for
coming to its own conclusion about the content of the speaker's
testimony itself, and this is shown by the fact that the speaker is
entitled to defer challenges to this content back to the original speaker.
According to the second-personal model, a speaker's testimony

provides an audience with a kind of reason for belief that is
categorically different from the kind of reason for belief provided by
ordinary events and states of affairs and even by other kinds of
nontestimonial utterances and expressions of belief. The kind of reason
for belief provided by testimony is distinctively connected to the
responsibilities that a speaker assumes in making what she says an
instance of testimony, in addressing her audience with a claim that
entitles the audience to defer challenges back to her. Adopting some
recent terminology from Stephen Darwall, we might say that testimony
amounts to a second-personal reason for belief—testimony amounts to a
consideration that justifies a belief in virtue of relations of authority
and responsibility existing between an addresser and an addressee.14
3.4 Moran on Assurance

In recent years, the most influential advocate of the second-personal

model has been Richard Moran. Indeed, the account of the
epistemology of testimony that I offered in chapter 2 is heavily indebted
to Moran's work. Nevertheless, there is a difference of emphasis
between Moran's account and my own that merits attention. Whereas
the account I have presented focuses on the distinctive kind of reason
for belief that testimony provides for an audience, Moran is first and
foremost concerned with the speaker's first-person perspective in
offering her testimony to an audience.15 This isn't to say that Moran is
unconcerned with the audience's perspective or with the kind of reason
for belief that a speaker's testimony provides. Moran argues for an
account of testimony according to which testimony amounts to a
distinctively interpersonal kind of reason for belief. However, Moran
argues for this via consideration of the speaker's first-person
perspective. Moran argues from considerations concerning the
speaker's first-person perspective with respect to her testimony to the
conclusion that testimony (p.95) provides a distinctively interpersonal
kind of reason for belief. The interpersonal account of the epistemology
of testimony that I presented in chapter 2 does not rely on
considerations concerning the speaker's first-person perspective. In
this sense, though Moran and I offer very similar positive accounts of
the epistemology of testimony, we get there by different routes.
Moreover, I have some qualms about the route that Moran takes. While
the considerations he adduces concerning the speaker's first-person
perspective with respect to her testimony are consistent with the
account of the epistemology of testimony that Moran and I both
endorse, I don't think that they provide as decisive an argument for this
account as the argument developed in chapter 2.
Moran (2005) sets out to defend what he calls the assurance view of
testimony. According to the assurance view, a speaker's testimony that
p amounts to an assurance that p which is a consideration that counts
in favor of p in virtue of the speaker's assuming an epistemic
responsibility towards the audience. Moran opposes the assurance view
to what he calls the evidential view of testimony. According to the
evidential view, a speaker's testimony that p amounts to ordinary
evidence for p. Moran holds that both reductionists and anti-
reductionists about testimony typically construe testimony as a kind of
ordinary evidence and thus that they both miss what is distinctive about
the kind of reason for belief provided by a speaker's telling. Both
reductionists and anti-reductionists miss the way in which the kind of
reason for belief provided by testimony is distinctively interpersonal in
being dependent on the speaker's assuming the relevant
responsibilities involved in giving the audience an epistemic assurance.
In the terms that I have developed, this is to say that both reductionists
and anti-reductionists have trouble making sense of the way in which
testimonial knowledge is distinctively secondhand.
As we have seen, it is important to be clear about the precise notion of

evidence at work in this discussion. Moran's conception of evidence is
clearly broader than that of inductive evidence. Anti-reductionists
about testimony clearly think that testimony provides a noninferential
reason for belief. If Moran thinks anti-reductionists still end up
construing testimony as a kind of evidence, then he must be working
with a broader conception of evidence than that of inductive
evidence.16 However, Moran's conception (p.96) of evidence is
narrower than the broadest philosophical sense mentioned above.
Moran clearly holds that testimony amounts to a genuinely epistemic
reason for belief, and yet he thinks that this reason for belief is
categorically different from that provided by ordinary evidence.
On Moran's conception of evidence, if a consideration c is evidence for

some proposition p, then c counts in favor of the truth of p completely
independently of anyone's beliefs, desires, or intentions with respect to
c. In particular, the evidential status of c cannot be actively conferred
on it by a person's intending it to be a reason for belief. In this sense,
even if one takes a photograph with the intention of using it as evidence
in a court of law, the evidential status of the photograph is not
conferred on it by the photographer's intending it to be so used. The
photograph is evidence for the state of affairs it represents completely
independently of anyone's intending it to be used as evidence.17
Accordingly, if testimony amounts to evidence, then a speaker's
testimony that p must count in favor of the truth of p completely
independently of the speaker's (or anyone else's) intending it to count
in favor of the truth of p. The evidential status of a speaker's testimony
must be something that cannot be actively conferred on it by the
speaker.
Moran notes that the intentions of the speaker clearly do contribute to

the way in which testimony amounts to a reason for belief. Testimony
amounts to a reason for belief largely in virtue of the communicative
intentions expressed by the speaker in making her utterance an
instance of testimony. This is precisely what distinguishes testimony
from other speech acts like explaining, demonstrating, or arguing. To
illustrate this point, Moran appeals to the Gricean distinction between
natural and nonnatural meaning. According to Moran, a speaker's
testimony that p counts in favor of the truth of p in a way similar to that
in which, according to Grice (1989), an utterance nonnaturally means
that p. On Grice's account of nonnatural meaning, an utterance
nonnaturally means that p in virtue of the speaker's intending the

audience to believe that p on the basis of the audience's recognition of
the speaker's intention. This is different from natural meaning. Smoke
naturally means fire in virtue of being a natural sign or indication of
fire, something that doesn't depend on anyone's intentions. In contrast
to natural meaning, nonnatural meaning is actually conferred on an
utterance by the speaker's intentions. Moran transports this account of
nonnatural meaning into an epistemological register. For Moran, a
speaker's testimony that p counts in favor of the truth of p in virtue of
the speaker's intending it to do so and in virtue of the audience's
recognition of this very intention on the part of the speaker.
(p.97) Moran thinks that the evidential view of testimony is not in a

position to explain how the intentions of the speaker play this role in
constituting what the speaker says as a reason for belief. On the face of
it, however, the evidential view can provide such an explanation. From
the audience's point of view, the speaker's intending the audience to
believe that p can simply serve as additional evidence that counts in
favor of the truth of p. The speaker's openly intending the audience to
believe that p in making her utterance an instance of testimony can
serve as an additional consideration, beyond the speaker's utterance
itself, that increases the overall case for p. Moran sometimes talks as if
such an account is straightforwardly incoherent, as if intentionally
produced evidence will necessarily look like bad or doctored evidence,
but this would only seem to be the case if the relevant intentions are
deemed to be deceptive. If the intentions of the speaker are not deemed
to be deceptive, then they would seem to be ordinary states of affairs
that can serve for the audience as perfectly good evidence for believing
that p.
Moran's ultimate objection to this kind of account concerns the

speaker's first-person perspective with respect to her testimony. Moran
claims that the speaker cannot consistently relate to her communicative
intentions in testifying as just some more evidence that increases the
case for p. From a first-person perspective, the speaker's own
intentions are typically very unlike ordinary states of affairs with which
she is simply confronted. The speaker has the freedom to determine the
content of her intentions—the speaker's intentions are, typically, up to
her—and this has implications for the way in which, from a first-person
perspective, the speaker can conceive of her intentions as contributing
to her testimony's amounting to a reason for belief. From the speaker's
perspective, she is able to actively confer epistemic import on what she
says by making her utterance an instance of testimony. The speaker is
thus able to freely determine that what she says amounts to a
testimonial reason for belief, and according to Moran, this is

inconsistent with the speaker's taking an evidential stance towards her
own testimony.
Moran thinks that this is something that cannot be accounted for on the
evidential view. The evidential view cannot make sense of the way in
which the speaker is in a position to freely determine the epistemic
import of what she says.
The [evidential and assurance] views, then, oppose each other

most directly over this issue of the role of the speaker's freedom,
and the hearer's dependence on it. On the evidential view,
dependence on the freedom of the other person just saddles us
with an additional set of risks; now we have to worry not only
about misleading (natural) evidence but deliberate distortion as
well. On the assurance view, dependence on someone's freely
assuming responsibility for the truth of P, presenting himself as a
kind of guarantor, provides me with a characteristic reason to
believe, different in kind from anything provided by evidence
alone. (2005: 7)
(p.98) Let's assume that Moran is right that from a first-person

perspective a speaker cannot consistently take an evidential stance
towards her own testimony. The speaker must instead see herself as
freely conferring epistemic import on what she says. What implications
does this have for the way in which the audience must relate to the
speaker's testimony? Does this mean that the audience cannot (or
ought not) take an evidential stance towards a speaker's testimony?
One might think that the fact that the speaker cannot relate to her own
testimony evidentially is simply a constraint imposed by the first-person
perspective and thus that it has no implications for the audience's
perspective on the speaker's testimony. Even if the speaker must see
herself as freely conferring epistemic import on what she says, from the
audience's perspective what the speaker says amounts to ordinary
evidence like any other. Moran considers such a response:
It might still be asked, however, whether it doesn't still all come

down to evidential relations in the end. The following
reconstruction may be offered. Yes, the speaker freely assumes
responsibility for the truth of what he asserts. But now this very
act of assurance is a fact, which the audience confronts as
evidence (of some degree of strength) for the truth of what has
been asserted. Speech is acknowledged to be importantly

different from other (indicatively) expressive behavior, but the
audience's relation to it, as a reason to believe something, can
only be evidential. (2005: 23).
Fricker (2006b) has recently endorsed something like this view.

Impressed by Moran's account of the speech act of telling, Fricker now
holds that the paradigmatic kind of speech act relevant to the
epistemology of testimony is the speech act of telling, a speech act in
which a speaker actively confers epistemic import on what she says.
However, she thinks that this doesn't have any implications for the
audience's epistemic relation to the speaker's testimony. An audience is
still in the position of treating what the speaker says as evidence (for
Fricker, inductive evidence).18
Moran admits that a speaker's testimony, just like anything else a

person says, can be treated as ordinary evidence. However, he claims
that since the speaker cannot consistently relate to her own testimony
as evidence, if the audience were to treat what the speaker says as
ordinary evidence, then the speaker and audience would be in
disharmony. (p.99) The speaker would be presenting the audience with
what she considers to be a nonevidential assurance, and the audience
would be refusing to accept this assurance from the speaker. Due to
this, Moran claims that there must be a norm of correspondence
between the reason offered and the reason accepted:
For the act of telling to complete itself there must be a norm of

correspondence between the reason being presented by the
speaker and the reason accepted by his audience. This is the
nexus that is aimed at in the self-reflexive aspect of the Gricean
formula, wherein the speaker asks that the very reason he is
thereby presenting be the reason that the audience thereby
accepts (i.e., through recognizing that very intention). Telling
aims at being believed, which proceeds, via the speaker's overt
assumption of responsibility, by joining together the particular
belief proposed for acceptance, the kind of reason being
presented for it, and the reason accepted by the audience. An
evidential stance, by contrast, de-couples all of these from each
other, to be reassembled as the observer thinks best. But such a
stance is contrary to the speaker's perspective on his action,
insofar as it pictures his presentation of himself as meaning, in
effect, that as far as reason-giving force goes, the audience is on
his own; as if the meaning of his utterance were “Now I have
spoken; make of it what you will” rather than “Take it from

me.” (2005: 26)
Moran here claims that a speaker's testimony does not function for an
audience as a consideration that the audience must come to its own
conclusion about. Though an audience can treat what a speaker says as
something that it must come to its own conclusion about, this is not to
trust the speaker for the truth. If the audience trusts the speaker for
the truth about p, then the audience is not coming to its own conclusion
about p, even if it must still come to its own conclusion about the
trustworthiness of the speaker. However, Moran seeks to support this
by appeal to a norm of correspondence between reason offered and
reason accepted, and I don't think that this supplies the support that he
needs. Why think that an audience's treating the speaker's testimony as
evidence would violate the norm of correspondence between reason
presented and reason accepted? Even if it is true that the speaker
cannot consistently take a purely evidential stance towards her own
testimony, even if she must see it as an epistemic assurance, why think
that the audience's accepting the speaker's assurance requires
anything other than taking her testimony and her communicative
intentions openly expressed therein to be good evidence for belief?
The evidential view of testimony has, in effect, a particular view of what

it is to accept a speaker's testimony—to accept a speaker's testimony is
to take the speaker's testimony to be good evidence for belief. This can
also be construed as a particular view about what it is to accept an
assurance. On the evidential view, to accept an assurance (p.100) is to
take the assurance to be good evidence for believing what is assured.
Moran thinks such acceptance is disharmonious, that accepting an
assurance ought to involve something other than taking it to be good
evidence for believing what is assured—but this is itself something that
needs to be demonstrated. On the evidential view, there is nothing else
that accepting an assurance could possibly be. Just because the speaker
cannot take an evidential stance toward her testimony does not
necessarily mean that the audience's taking such a stance would be out
of tune with the speaker. This might simply be part of the difference
between the first-person point-of-view of the speaker and the non-first-
person point-of-view of the audience.
If Moran's argument from the nature of the speaker's perspective

towards her own testimony to the assurance view is to succeed, he
must find a way of showing that the audience's treating the speaker's
testimony as evidence is actually out of tune with the speaker's
presentation of her testimony as an assurance. I think that there is at
least one point at which Moran suggests the materials with which to do
this.
The disharmony between speaker and audience entailed by the

evidential view comes out in the consideration of two possible
responses to receiving a promise. If someone promises to mail a
letter for me, one thing I might do is accept his promise, placing
myself in his hands and taking myself to now have sufficient
reason to believe that he will mail the letter. If it turns out he
doesn't mail the letter, either through carelessness or because he
never really intended to, then I will feel aggrieved and let down.
This is the ordinary expectation and liability to disappointment. I
might, however, opt for another kind of response altogether. Here
I don't accept the promise; I simply don't go in for that sort of
thing, as I may not accept promises from a small child or (for
different reasons) from someone I despise, but in another way I
do take seriously the fact that he made one to me. In this spirit I
may reason: “He is unlikely to make a promise he won't fulfill,
since that would discredit him as a future promisor, and there are
great and obvious advantages in remaining someone whose
promises are accepted. Therefore, the fact that he made this
promise to me makes it probable that he will in fact mail the
letter. So I believe he will.” If, on this second scenario, I later
discover that he did not mail the letter after all, my reaction will
be different. I will be disappointed, of course, and I will be
surprised that he would discredit himself in this way. But I can't
confront him with my complaint or my resentment because I
never accepted the promise in the first place (2005: 24).
Moran here claims that an audience's treating a speaker's promise as

good evidence that the speaker will do as promised is not in fact to
accept the promise. This is because accepting a promise has a
particular form. Accepting a promise involves the audience's putting
itself in the speaker's hands such that the audience acquires a right of
complaint if the promise should go unfulfilled. If the audience treats the
speaker's promise as evidence, (p.101) even as very good evidence,
then the audience does not acquire this right of complaint. I think this
is correct. Treating a speaker's promise as mere evidence that she will
do as promised is not to accept the promise, and this is demonstrated
by the fact that so treating a speaker's promise does not generate the
right of complaint normally generated by acceptance of a promise.
Accepting a speaker's promise to ϕ generates for the audience a right of
complaint if the speaker should fail to ϕ, and this particular right of
complaint is not generated if the audience simply takes the speaker's

promise to be good evidence for believing that the speaker will ϕ.
But what implications does this have for testimony? Moran clearly holds
that as goes for the acceptance of promises, so goes for the acceptance
of testimony. This is something that needs to be demonstrated,
however, especially since someone committed to the evidential view
might very well think that accepting testimony is in fact very different
from accepting a promise. I think it can be demonstrated that treating a
speaker's testimony as good evidence for the proposition testified to is
not yet to accept the speaker's testimony, but in order to do so we need
to show that accepting a speaker's testimony, like accepting a speaker's
promise, takes a particular form. We need to show that accepting a
speaker's testimony generates for an audience a right of complaint that
is relevantly analogous to the right of complaint generated by accepting
a promise.
In chapter 2, I argued that an audience has the right to defer certain

epistemic challenges to her testimonial beliefs back to the original
testifier and that this is a distinguishing feature of testimonial belief.
This epistemic right of deferral is only generated when an audience
accepts a speaker's testimony as such, thereby acknowledging the
speaker's assumption of responsibility towards the audience. If the
audience does not accept the speaker's testimony as such—if the
audience instead treats the speaker's testimony as ordinary evidence
from which the audience comes to its own conclusion—then the
audience does not acquire an epistemic right of deferral. In this sense,
the epistemic right of deferral does appear to be relevantly analogous
to the right of complaint generated by acceptance of a speaker's
promise. It is an entitlement generated only when the speaker's speech
act is accepted in a particular way.
Note, however, that this appeal to the epistemic right of deferral is

much more specific than Moran's appeal to a generic right of
complaint. As I have argued, the epistemic right of deferral is
distinctively generated by a speaker's testimony. It is not generated by
a speaker's arguing or by a speaker's mere expression of belief. A
speaker's arguing or mere expression of belief may, however, generate
a generic right of complaint. Misleading an audience through argument
or through other nontestimonial forms of assertion may violate all sorts
of moral, social, and conventional norms, and an audience has a right to
complain when this occurs. If we are to distinguish the reason for belief
provided by testimony from the reason for belief provided by other
forms of assertion, then we must appeal to something much more

specific than a generic right of complaint.
(p.102) Moreover, if we are going to establish that testimony amounts

to a distinctive kind of reason for belief in virtue of the fact that it
generates a distinctive right of complaint, we must appeal to a right of
complaint that has a clear epistemological significance, a right of
complaint that is not simply a moral or conventional issue but that is in
fact connected to the way in which testimony serves to epistemically
support an audience's belief. In arguing against Moran's assurance
view, Lackey claims that “being entitled to the reactions in question
lacks any epistemological significance and hence fails to establish that
there is an epistemologically relevant difference between justification
or warrant from telling and justification or warrant from mere
asserting” (2008: 235). Note that this charge cannot be so easily
leveled against the view of the epistemology of testimony that I
developed in chapter 2. The epistemic right of deferral certainly
appears to be epistemologically relevant, and the difference between
telling and mere asserting or mere declaration of belief does matter to
the epistemic right of deferral.
In arguing for his assurance view, Moran doesn't specifically appeal to

the audience's epistemic right of deferral. I think that this leaves him
unable to conclusively establish that accepting a speaker's testimony
must take a distinctive form and thus leaves his assurance view
insufficiently supported. In this sense, defending the assurance view of
testimony requires paying more attention to the audience's perspective.
Establishing that a speaker's testimony amounts to a sui generis,
essentially interpersonal kind of reason for belief requires establishing
that an audience's acceptance of a speaker's testimony takes a
distinctively interpersonal form—a form that, as I have argued, is
exhibited by the ordinary practice of epistemic buck-passing.
3.5 Intermediate Cases and the Return of Epistemic Autonomy

We can now return to the general contours of the debate between the
second-personal, evidential, and inheritance models. In order to
understand exactly what is at stake between the three models, it will be
helpful to consider a concrete example. Imagine that I am a police
detective using a wiretap to listen in on a telephone conversation
between two criminals. I hear criminal A tell criminal B that a shipment
of narcotics is arriving on Wednesday. Given everything else that I know
about these criminals and about their relationship, I believe what
criminal A says—I believe that a shipment of narcotics is arriving on
Wednesday. If it is indeed true that a shipment of narcotics is arriving
on Wednesday, then it looks like I can here acquire knowledge of this
fact on the basis of the criminal's testimony. Is this a case of testimonial
knowledge? On the evidential and inheritance models, it looks like it is.
On the evidential model, the criminal's speech act amounts to good
evidence for believing that a shipment of narcotics is arriving on
Wednesday. On the inheritance model, the criminal's testimony amounts
(p.103) to an expression of belief the justification of which can be
inherited by a comprehending hearer. On the second-personal model,
however, things are not so clear. Knowledge can definitely be acquired
in this case, but on the second-personal model, it isn't clear that such
knowledge ought to be construed as properly testimonial. This is
because it isn't clear that this knowledge involves trusting a speaker for
the truth. On the second-personal model, trusting the speaker for the
truth involves more than relying on the speaker. It involves relying on
the goodwill of the speaker, and it isn't clear that this case of mere
overhearing involves anything of the kind. This is demonstrated by the
fact that, if someone were to challenge my belief that a shipment of
narcotics is arriving on Wednesday by producing evidence that tells
against this, I would not feel entitled to defer the challenge to the
criminal. I might cite the criminal's testimony as evidence for my belief,
but I would not feel entitled to hold the criminal responsible for
meeting the epistemic challenge to my belief. In this sense, it looks like
this is a case in which I have come to my own conclusion about things.
Even though the criminal's testimony clearly figures into the
justification of my belief, the way in which it figures is different from
cases in which I take a speaker's word for something, from cases in
which I trust a speaker for the truth. According to the second-personal
model, then, this case looks like it should be excluded from the
category of properly testimonial knowledge.19
A similar line of thought figures into the way in which the adherent of
the second-personal model will respond to Lackey's CONSISTENT
LIAR. According to the second-personal model, testimonial knowledge
cannot be acquired from insincere testimony. Insofar as the capacity for

learning from testimony involves trusting the speaker for the truth, the
acquisition of testimonial knowledge is crucially dependent on the
goodwill of the speaker. But this goodwill is exactly what is missing in
the case of insincere testimony. Insincere testimony is testimony in
which a speaker merely purports to assume the epistemic
responsibilities involved in testifying, and according to the second-
personal model testimonial knowledge cannot be acquired through the
merely purported assumption of such responsibilities. In this sense,
testimonial knowledge is crucially dependent on the sincerity of the
speaker. If the audience knows the speaker to be insincere, then the
audience may still treat the speaker as a reliable indicator, just as the
detective above treats the criminal as a reliable indicator of the arrival
of the narcotics shipment. But according to the second-personal model,
this is (p.104) not a way of acquiring properly testimonial knowledge.
Properly testimonial knowledge, knowledge that involves trusting a
speaker for the truth, simply cannot be acquired from insincere
testimony.
An adherent of the second-personal model will have to say something

slightly different about Owens's case of the hypnotized patient. Owens's
hypnotized patient is not insincere; she is sincerely expressing her
genuine, though unjustified, beliefs. She is not therefore merely
purporting to assume the relevant epistemic responsibilities involved in
addressing an audience with her testimony. She is genuinely assuming
those responsibilities. However, the responsibilities she is assuming are
ones that she cannot, in the end, discharge.20 If the relevant
responsibility a speaker assumes in making her utterance an instance
of testimony is a responsibility to meet, on the audience's behalf,
certain challenges to the audience's belief that the audience is entitled
to defer back to her, it doesn't look like the hypnotized patient is in a
position to do this. Though the patient has genuinely assumed the
responsibility to meet such challenges, if such a challenge were
actually deferred to her, there is no way in which she could even begin
to go about meeting it. Note here that she couldn't even defer it herself,
as she could if she was a link in a testimonial chain. If the doctor had
told her that the woman was murdered instead of merely hypnotizing
her, then she could defer the challenge back to the doctor, but the
doctor's project of hypnosis was precisely designed so as not to allow
this.
So the second-personal model can explain Lackey's and Owens's cases

in a way that is different from the explanation provided by the
inheritance model. Both the second-personal and inheritance models
hold that the problem with CONSISTENT LIAR is that epistemic

responsibility has not been transferred from Bertha to Henry in the way
required for Henry to acquire properly testimonial knowledge, but
while the inheritance model holds that this is because Bertha hasn't
genuinely expressed her belief, the second-personal model holds that
this is because Bertha hasn't assumed the epistemic responsibilities
involved in sincerely testifying. Similarly, both the second-personal and
inheritance models hold that the problem with the case of the
hypnotized patient is that even though epistemic responsibility has been
appropriately transferred, this has not succeeded in appropriately
justifying the audience's belief; but while the inheritance model holds
that this is because there is no justification present for the speaker's
belief to inherit, the second-personal model holds that this is because
the speaker has assumed an epistemic responsibility that she is not in a
position to discharge.
(p.105) It might seem that the differences here between the second-
personal and inheritance models are quite small. In fact, however, they
are quite large. According to the second-personal model, an audience's
testimonial beliefs are justified by the speaker's assumption of an
epistemic responsibility towards the audience. An audience's belief is
justified in virtue of the actual person-to-person relation involved in a
speaker's addressing her testimony to the audience. On the inheritance
model, these kinds of addressive person-to-person relations are not
necessary for the acquisition of testimonial knowledge. Owens claims
that “the kind of epistemic responsibilities at stake in testimony are not
duties owed to anyone; testimony can be presented quite
unintentionally to an audience who thereby learn that it is true because
they are entitled to depend on the speaker for justification” (2006:
117). He then presents the following example:
Think of a secret diary containing one's most intimate thoughts. It

is not just that the author of this diary has no intention of
communicating these thoughts to others by writing them down;
he has every intention of not communicating them and is careful
to keep his diary secret. Such a diary may be filled with assertions
but assertions which lack an intended audience and are not meant
to assure anyone of anything . . . .
Suppose that, unbeknownst to the speaker, I read his secret diary

(or bug his private monologues). Here the diarist may have
specifically intended to keep his diary away from me. Still, can't I
learn from the author's diary in just the way I learn from his
conversation, or his published works? True, he may not have told
me anything, but his diary has and to believe his diary is to
believe him, to take his word for it . . . Though this diarist didn't
intend to put himself under an obligation to us, we can learn from
what he says just as we could had he been speaking to us directly.
(2006: 117–118)
For Owens, we can acquire properly testimonial knowledge from

assertions that are not addressed to us and even from assertions that
are not addressed to anyone.21
On the second-personal model, this is far from clear. We can agree that
the secretive diarist doesn't intend to address her utterances to any
particular other, but are her utterances not addressive at all? In
different circumstances, the addressees of an utterance can include
individuals, groups of varying sizes, and sometimes even (p.106) the
public in general. Moreover, the very medium of the written word
seems to acquire a good deal of its significance from the way in which it
serves to concretize addressive relations over time and space. It thus
wouldn't be too much of a stretch to think that part of the point of
keeping a secretive, private diary is to address one's speech to a
distant, perhaps wholly imaginative person or community. There might
thus be at least some element of address present even in the
statements of a secret journal.22
Even if one doesn't find this very plausible, however, why should we
think that the statements of such a secret diary are in fact clear cases
of testimony? If we do find such a diary and acquire sufficient evidence
to believe that the author was indeed making sincere statements with
respect to matters concerning which she was competent, would we
then trust the speaker? Would we then, as Owens claims, take her word
for things? We could certainly treat the author's statements as good
evidence for belief, but both the inheritance and the second-personal
models agree that this is insufficient for acquiring a properly
testimonial belief.
Due to this, I don't think that an advocate of the second-personal model

should find Owens' alleged counterexample very compelling. The
second-personal model certainly takes its cue from the addressive
relations involved in the kind of face-to-face exchange in which one
person comes out and tells something to another, but an advocate of the
model should admit that these addressive relations can shade off to the
point where sometimes it just isn't very clear whether an audience is
treating a speaker's statements as testimony or as mere evidence.
There are intermediate cases. But the advocate of the second-personal
model will think that this is just as it should be. It is often unclear
whether we are trusting a speaker for the truth or merely treating what
she says as good inductive evidence for belief, and an adequate
epistemology of testimony should be in a position to account for this.
An adherent of the second-personal model is in a position to claim that

there are at least three dimensions along which our ordinary practice of
giving and receiving testimony can shade off and become unclear. First,
in particular situations it can be unclear whether a speaker is
straightforwardly offering testimony. Whether a speaker's speech act
amounts to testimony or to something less than this—a passing remark,
a mere declaration of belief, et cetera—is something that we often
dispute and that no doubt depends on a whole range of contextual
factors. There is no simple test for determining whether what a speaker
says calls for the audience (p.107) to trust her for the truth. Instead,
this is something that is typically rendered contextually salient.23
Second, even in cases in which testimony is clearly offered, it is often

unclear just how far the epistemologically relevant content of the
speaker's testimony extends. Consider in this respect the case of
conversational implicatures. Standardly, for something to count as a
conversational implicature, it must be capable of being modeled on the
form of an explicit inference. We must be able to understand the
implicated content as the conclusion of an argument that begins from
the conventional meaning of what is said and proceeds through
whatever relevant conversational maxims one takes to be in play. One
might then think that an audience's epistemic relation to implicated
content is just like an audience's relation to the conclusion of any
argument that it constructs for itself.24 But even if implicatures must be
capable of being modeled on the form of an explicit inference, this
doesn't necessarily mean that an audience's epistemic relation to
implicated content is always inferential. Grice himself holds that
conversational implicatures are rationally required by the maxims of

conversation (1989: 30–31). If an implicature is a genuinely
conversational implicature, then not only must the implicature be
capable of being modeled on the form of an explicit inference, but the
audience's drawing of the implicature must be rationally required in
order to make sense of the speaker's utterance within the
conversational context. There thus appears to be some sense in which
the audience's drawing of the implicature is nonoptional. This makes
conversational implicatures unlike ordinary run-of-the-mill inferences.
An audience may draw all sorts of inferences from what a speaker says,
but the drawing of most of these inferences is not required to so much
as make sense of the speaker's assertion.
As Jonathan Adler notes: “Many inferences are not required (but

optional) to avoid violation. Those are cases in which we could say that
the inference is suggested, allowed or risked. The implication is not
mandated, since the assumptions violated (p.108) are not contextually
mutual to speaker and hearer, even if widely held. (A test is that
cancellation without reinterpretation of the assertion generates at least
some puzzlement only for genuine implicatures)” (1997: 445). The
audience's drawing of a genuinely conversational implicature is
nonoptional at least in the sense that, within the context of utterance, if
the implicature is not drawn then the utterance is rendered puzzling.
The drawing of genuinely conversational implicatures is necessary for
understanding speaker meaning.
Adler's distinction between genuine conversational implicatures and

inferences that are merely “suggested, allowed, or risked” is still a
distinction within the class of inferences. One might therefore think
that whether an inference drawn from a speaker's assertion is
“optional” or not, the audience is still in the epistemic position of
coming to its own conclusion about the content of the inference. If a
speaker's assertion that p genuinely conversationally implicates that q,
then even though the audience's drawing of q is necessary for
understanding speaker meaning, the audience is still solely responsible
for the justification of its belief that q.25 However, several philosophers
have claimed that a speaker can sometimes assume just as much
responsibility for the content of her conversational implicatures as for
the literal content of what she has said. Stanley Cavell, for example,
writes: “Intimate understanding is understanding which is implicit. Nor
could everything we say (mean to communicate), in normal
communication, be said explicitly—otherwise the only threat to
communication would be acoustical. We are, therefore, exactly as
responsible for the specific implications of our utterances as we are for

their explicit factual claims” (1976: 12).26
Cavell claims that we are exactly as responsible for the specific

implications of our utterances as for our explicit claims, but not all
implicatures are equally specific. As we've seen, some inferences may
only be suggested by a speaker's assertion—some inferences may not
be required by the maxims of conversation in order to so much as make
sense of the utterance in the conversational context—and so some
inferences may not rise to the level of genuine conversational
implicatures. It also seems plausible that some conversational
implicatures may be more required in a context than others. Dan
Sperber and Deirdre Wilson thus hold that there is a continuum of
implicature that stretches (p.109) from implicatures with regard to
which the speaker assumes as much responsibility as she does for the
proposition literally expressed to implicatures with regard to which the
speaker assumes virtually no responsibility at all (1991: 384–385, 1995:
198–199). Some conversational implicatures are required by the
context of utterance to such a degree that the speaker assumes as
much responsibility for them as for the literal content of her assertion,
but in contexts in which the hearer must choose between many possible
alternative implications, all of which are consistent with the principle of
relevance, the hearer is thereby more responsible for the implicatures
that she draws.
Importantly, Sperber and Wilson claim that what varies along this
continuum of implicature is “the amount of foreknowledge the speaker
must be taken to have had of the way the utterance would be
processed, and with it the degree of responsibility he must take for the
particular conclusions derived,” and they gloss this by saying that “a
proposition can be expressed by a speaker with a stronger or weaker
guarantee of truth, and . . . this guarantee of truth may be more or less
trusted by the speaker” (1991: 385). Here they seem to be saying that
what varies along their continuum is the extent to which the speaker
has offered, and the audience is in the position to accept, an assurance
or guarantee of the truth of the implicated proposition. This casts the
issue in a decidedly testimonial register. Putting things in the terms
that we have developed, what varies along this continuum of
implicature is the extent to which the speaker can be seen as calling for
the audience to trust her for the truth concerning the implicated
proposition. What varies is the extent to which the implicated
proposition is included in the epistemologically relevant content of the
speaker's testimony. The greater the extent to which the implicated
proposition is required to understand speaker meaning, the greater the
extent to which the speaker is assuming responsibility for the truth of

the implicated proposition, meaning the greater the extent to which the
speaker is responsible for meeting challenges to the audience's belief in
the implicated proposition that may be deferred back to her.27
Finally, even in cases in which testimony is clearly offered and in which

the epistemologically relevant content of the speaker's testimony is
clearly defined, it is often unclear whether the testimony is clearly
accepted as such by the audience. It seems that an audience can take
what a speaker says more or less on her authority, trust the speaker for
the truth to a greater or lesser degree, and this will be exhibited in the
audience's being disposed to defer back to the speaker a broader or
narrower range of challenges. In this sense, though there are clear
cases in which an audience trusts a speaker for the truth and clear
cases in which an audience does not (though it may still treat what the
(p.110) speaker says as good evidence for belief), there will be a broad
continuum in between. Within this continuum, the greater the range of
epistemic challenges the audience is disposed to defer, the greater the
extent to which the audience is trusting the speaker for the truth. In
some cases, perhaps cases like Owens's secretive diarist, an audience
may be willing to trust a speaker for the truth to a very minimal degree
beyond which the audience will treat the speaker's testimony as
ordinary inductive evidence. But again, there is no simple procedure for
determining the extent to which this is so. The extent to which an
audience is rationally entitled to trust a speaker for the truth, and
hence the range of epistemic challenges the audience is rationally
entitled to defer, is again something that is typically rendered
contextually salient.
I have been arguing that Owens's case of the secretive diarist doesn't
cut against the second-personal model as decisively as he seems to
think. I suspect, however, that Owens has deeper reasons for rejecting
the second-personal model, reasons that he in fact shares with
advocates of the evidential model. Many philosophers will think that the
very idea that testimonial knowledge and belief can be justified in
virtue of a speaker's assumption of epistemic responsibilities towards
an audience is simply nonsensical. Relations of authority and
responsibility between persons are simply the wrong kind of thing to
play a genuinely justificatory role in the processes of belief formation
and knowledge acquisition. Such considerations may be of legitimate
moral or social concern, but they have no genuinely epistemic
significance. I suspect that something like this idea is behind Owens's
claim that “the kind of epistemic responsibilities at stake in testimony
are not duties owed to anyone” (2006: 117, original emphasis).
Relations of responsibility a speaker bears to an audience are not the

kind of thing that can play a role in epistemically supporting an
audience's belief. An advocate of the evidential model will surely agree.
However, Owens isn't willing to go so far as to hold that testimony
amounts to ordinary evidence. He thinks that there is an important
sense in which epistemic responsibility for an audience's testimonial
belief is shared out between speaker and audience. However, since he
thinks that this epistemic responsibility simply can't take the form of
responsibility to, he thinks the responsibility that a speaker in fact
bears for the justification of an audience's testimonial belief is simply
responsibility for the justification of the speaker's own belief, a
justification that can then be inherited by a comprehending hearer.
Hopefully it is now becoming clear what is at stake between these three

competing models of epistemic dependence. The evidential model holds
out hope for a very broad and domesticated account of the
epistemology of testimony. This is both its virtue and its vice. It holds
out hope for providing a unified account of the nature of knowledge and
justified belief based on statements, but it does so at the expense of
losing the intuitive sense in which testimonial knowledge appears to
exhibit a distinctive kind of epistemic dependence. The inheritance
model seeks to provide a narrower account of (p.111) these cases of
distinctively testimonial dependence, excluding from the category of
knowledge and belief based on testimony those cases in which a
speaker's statements are treated as mere evidence, but it doesn't go so
far as to hold that the epistemology of testimony is irreducibly
interpersonal. Instead, it holds that testimony is an essentially
preservative mechanism much like memory, a mechanism that allows
an audience to inherit the rationality of a speaker's belief.28 This also
amounts to a kind of domestication of the epistemology of testimony. On
the inheritance model, testimony is just as unproblematic a source of
knowledge and justification as memory. Testimony is unproblematic in
virtue of being merely preservative. An advocate of the second-personal
model, however, will think that even this much domestication is too
much. Trusting a speaker for the truth is not like recalling information
through memory. Other people are not simply sources from which we
can inherit justification. Instead, when we trust a speaker for the truth,
the actual relationship that the speaker assumes towards us does some
of the work of epistemically supporting our testimonial beliefs. Our
testimonial beliefs are at least partially justified in virtue of our
relationship towards a person, not by the justification of the speaker's
belief. Hence, when asked for the justification of our belief, we simply
cite the speaker: “So-and-so told me so; and I believe her.” This,
however, is to cast testimony as a strangely sui generis kind of reason
for belief—an essentially second-personal reason for belief, as I have

called it—and one might think that this goes too far. This makes the
epistemology of testimony look just too strange. It doesn't fit with our
well-worn ways of thinking about knowledge and justification.
As we saw in chapter 2, one response to the argument from

secondhandness involves simply rejecting the epistemic significance of
what I have identified as the epistemic right of deferral. We can see
essentially the same issue arising again here. Adherents of both the
evidential and inheritance models think that the kinds of interpersonal
relations of authority and responsibility appealed to by adherents of the
second-personal model may be of legitimate moral, social, or
conventional concern but that such considerations simply cannot play a
genuinely justificatory role in the acquisition of knowledge and belief.
In chapter 2 I argued, first, that this seems to be a mischaracterization
of our ordinary epistemic practice, but also, second, that this seems to
evince an as yet un-argued-for skepticism about the possibility that
interpersonal relations might play a genuinely justificatory role in the
processes of belief formation and knowledge acquisition. In the
remaining two chapters I want to try to further defend the second-
personal model by working from two different but related directions to
lessen the appeal of such skepticism.
(p.112) One domain in which such skepticism can arise concerns the
second-personal model's appeal to the notion of trust. Trust, one might
think, is a matter for ethics, not for epistemology. Trust is something we
do in the absence of good reasons for belief—indeed, something we
may choose to do in the teeth of good evidence to the contrary. This
way of thinking is actually encouraged by much of the philosophical
literature on trust. Theorists of trust often argue that trust is
importantly different from belief and that there is an inherent tension
between trust and theoretical rationality. If this is so, then the second-
personal model's appeal to trusting a speaker for the truth looks like an
appeal to a kind of psychological attitude that falls outside the purview
of epistemology proper. In chapter 4 I will argue that this is not so.
Drawing on the interpersonal account of testimony developed thus far,
we can develop a compelling account of the general nature of trust that
can account for the important interpersonal dimension of trust while
nevertheless construing trust as a species of belief. On such an account,
there is no tension between trust and theoretical rationality, and
trusting a speaker for the truth can be understood as a robustly
cognitive, robustly rational attitude.
Construing trust as a genuinely cognitive attitude requires admitting

that interpersonal relations of authority and responsibility between
persons can play an irreducible role in epistemically supporting belief.
A second and related form that the aforementioned skepticism might
take involves denying precisely this. The skeptic might admit that such
interpersonal relations are of legitimate moral, social, or conventional
concern, but deny that they have any genuinely epistemic significance.
Interpersonal relations of authority and responsibility between persons
are simply the wrong kind of thing to play a role in epistemically
supporting belief. This amounts to a denial of the possibility of what I
have called second-personal reasons for belief. According to this line of
thought, second-personal relations might play a genuine role in
practical rationality, in reasoning that issues in action, but they cannot
play an analogous role in theoretical rationality, in reasoning that
issues in belief. In chapter 5, I will argue that this is not the case. There
is just as much reason to think that interpersonal relations of authority
and responsibility have an irreducible epistemic significance as there is
to think that they have an irreducible practical significance. This means
that if one is willing to accept that authority relations can play a
distinctive and irreducible role in practical rationality, then one should
also accept that they can play a distinctive and irreducible role in
reasoning about what to believe.
Notes:
(1) Note that these are very rough and ready characterizations of each
model. It is the task of the entirety of this chapter to spell out each of
these models in more detail and to clarify what is at stake between
them.
(2) Lackey's distinction between the statement view and the belief view
might appear similar to my distinction between the evidential model
and the inheritance model. However, her characterization of the
statement view is actually narrower than my evidential model and her
characterization of the belief view is broader than my inheritance
model. In effect, what Lackey's statement view does is combine the
evidential model with an extremely broad characterization of the
relevant communicative event from which testimonial knowledge and
belief is to be gained, namely statements that may or may not be
expressive of the beliefs of a speaker. Many adherents of the evidential
model hold that testimonial knowledge and belief can only be acquired
from assertions that are actually expressive of the beliefs of a speaker.
Lackey classes these positions as instances of the belief view, but it is
important to note that such positions do not hold that a speaker's
assertions are vehicles for the expression of beliefs the justification of
which can be inherited. Similar to Lackey's own view, such positions

hold that testimonial knowledge and belief is justified by the evidence
provided by the speaker's testimony. One might wonder, as Lackey in
effect does, why a speaker's assertions must be actually expressive of
her beliefs in order to amount to such evidence, but this seems to be a
less dramatic dispute than that between the evidential and inheritance
models as I have characterized them. On my characterization of the
inheritance model, the central claim is that an audience's testimonial
belief is justified by whatever it is that justifies the speaker's belief, a
justification that the audience's belief inherits from the speaker. Far
fewer philosophers hold this view than hold that sincerity is necessary
for the acquisition of testimonial knowledge. In this sense, whereas
Lackey's distinction between the statement and belief views classes
positions like those of Fricker and Coady under the belief view, my
distinction between the evidential and inheritance models classes them
both under the evidential model.
(3) Note the parallel to Prichard's case. If you were to learn (1) that the
thermometer is broken but (2) that someone is working behind the
scenes to make it the case that the thermometer reliably indicates the
correct temperature, it seems that at the very least you would not
continue to simply believe the thermometer. You might continue to
believe what the thermometer indicates, but only if you trust the person
working behind the scenes.
(4) Recall that Lackey formulates CONSISTENT LIAR such that Henry
has “trust in Bertha” (2006c: 83). It is noteworthy that adherents of the
evidential model standardly characterize ordinary cases of the
acquisition of testimonial knowledge and belief as cases of trusting a
speaker. This puts them in an odd dialectical position due to the fact
that, as we'll see, they typically deny that the acquisition of testimonial
knowledge and belief has anything to do with trust. Fricker (2006b) is
an exception. Fricker argues that the proper object of explanation for
the epistemology of testimony is knowledge acquired from trust in
testimony, but she nevertheless seems to think that trust in testimony
can be adequately understood on the evidential model.
(5) See also Lackey (2006a).
(6) Fricker claims, “If I overhear you exclaiming to yourself that P, I

cannot, or cannot so easily, pick up the same sort of entitlement to
believe P. Without an intended audience there is no distinction between
an exclamation which involves vouching for truth—an assertion, and a
mere expression of thought without any such commitment” (2006b:
598). For Fricker, unlike Lackey, the justification relevant to testimony
can only be acquired from a speech act that is addressed to an

audience. Fricker does not hold, however, that the speech act must be
addressed to the audience who forms the testimonial belief. An
audience can merely overhear a telling addressed to someone else and
still acquire the relevant testimonial entitlement. The necessary
addressive features are conceived as features of the act of telling itself,
conceived as an event, rather than as part of an interpersonal
transaction between speaker and testimonial believer, as is
characteristic of the second-personal model. See section 3.3.
(7) Other adherents of the inheritance model include Burge (1993),

Goldberg (2006), and Schmitt (2006).
(8) Coady (1992: 45–46) presents a similar case, though Coady

stipulates from the start that the audience is aware of what has been
done to the speaker. Coady claims that if we are aware of what has
been done to the speaker, then we may treat what she says as good
evidence for belief, but we will not thereby trust the speaker. This
contrasts with cases of genuinely testimonial belief. “When we believe
testimony we believe what is said because we trust the witness” (1992:
46).
(9) Owens's example might be modified to eliminate even this

difference. We might stipulate that the patient actually believes
(incorrectly) that this particular woman wasn't murdered and that she
is hypnotized such that every time she intends to state that the woman
was not murdered she instead states the opposite.
(10) See also Holton (1994: 74). Lackey is entirely unmoved by this kind
of objection:
But what is the epistemic problem with treating speakers as mere

truth-gauges? To be sure, there may be psychological problems
with a mother treating her children merely as truth-gauges, moral
problems with a husband treating his wife merely as a truth
gauge, and pragmatic problems with a quarterback treating his
teammates merely this way. Given that the central epistemological
goal is the acquisition of true beliefs and the avoidance of false
ones, however, the epistemic objection here—if indeed there is
one—appears to be entirely misguided. (2008: 249)
Of course, there is no epistemic problem with treating speakers as mere

truth-gauges. We often do treat speakers as such, and so treating
speakers can be a way of acquiring knowledge. The point is simply that
we do not always treat speakers in this way. It is a matter of fact that
we do not always treat speakers as mere truth gauges, and as I argued
in the previous chapter, this is born out by our ordinary practice of
deferring challenges. In cases in which we take a speaker's word for
things, we are entitled to defer epistemic challenges back to the
speaker, something we are not entitled to do with respect to other kinds
of ordinary instruments and events.
(11) Note that an advocate of the inheritance model must provide an

account of why it is that only the speech act of testimony (or perhaps
assertion) and not that of arguing or explaining serves to render the
speaker's justification inheritable. Owens (2006) argues that it is a
feature of the intentional expression of belief that it is constrained by
the speaker's view of the rationality of the belief expressed in such a
way that the audience's belief can inherit the justificatory status of the
speaker's belief. But if this is a general feature of the intentional
expression of belief, then it would seem to pertain just as well to the
cases of assertion as the conclusion of argument and assertion as the
mere declaration of belief as to the case of outright telling.
(12) For more on the general nature of interpersonal trust, see chapter
4.
(13) This narrow conception of evidence is at work in the following

passage from Austin:
The situation in which I would properly be said to have evidence

for the statement that some animal is a pig is that, for example, in
which the beast itself is not actually in view, but I can see plenty
of pig-like marks on the ground outside its retreat. If I find a few
buckets of pig-food, that's a bit more evidence, and the noises and
the smell may provide better evidence still. But if the animal then
emerges and stands there plainly in view, there is no longer any
question of collecting evidence; it's coming into view doesn't
provide me with more evidence that it's a pig, I can now just see
that it is, the question is settled. And of course I might, in
different circumstances, have just seen this in the first place, and
not had to bother with collecting evidence at all. (1962: 115)
In a footnote to this passage, Austin goes on to note that philosophers

often use the term ‘evidence’ in a much broader fashion: “I will have, it
will be said, the ‘evidence of my own eyes’. But the point of this trope is
exactly that it does not illustrate the ordinary use of ‘evidence’—that I
don't have evidence in the ordinary sense” (1962: 116).
(14) Darwall (2006b) actually denies that there are any genuinely
second-personal reasons for belief. For Darwall, second-personal
reasons are limited to the realm of the practical. I discuss Darwall's
account of second-personal reasons for action and his rejection of
second-personal reasons for belief in chapter 5.
(15) In this sense, Moran's work on testimony grows quite naturally out
of his prior work on the first-person. See Moran (2001).
(16) In his discussions of the way in which typical anti-reductionist

accounts are consistent with the evidential view, Moran sometimes
talks as if anti-reductionists hold that testimony amounts to
observational evidence from which an audience must make inferences.
See, for example, Moran (2005: 4–5). This is a mischaracterization of
the anti-reductionist position, but I think it is inconsequential for
Moran's purposes. Moran's considered view is clearly that even if we
have some kind of a priori entitlement to take testimony to be a direct,
noninferential reason for belief, analogous to perceptual
representation, this is still consistent with treating testimony as a kind
of evidence. Whereas reductionists treat testimony as a kind of
inductive evidence, anti-reductionists treat testimony as akin to the
kind of noninferential evidence provided by perceptual representation.
(17) It is debatable whether the representational capacity of a

photograph can be understood completely independently of the
intentions of the photographer, as Moran, following Walton (1984),
suggests. See Snyder (1993). It is therefore questionable whether the

evidential status of a photograph can be understood completely
independently of the photographer's intentions.
(18) Fricker (2006b) thus admits that the speech act of telling is one
that is constituted by certain kinds of addressive relations, but she
seeks to decouple these addressive relations from what it is that
justifies an audience's belief based on the telling. The addressive
relations involved in making an utterance an instance of testimony have
nothing to do with the justification of a hearer's testimonial belief, and
so addressees and mere overhearers are in an equivalent epistemic
position with respect to a speaker's testimony. If they have reason to
believe that the speaker is trustworthy, then they are in the position to
infer that what the speaker says is true.
(19) Moran seems to think that mere overhearers of a speaker's

testimony can acquire genuinely testimonial knowledge, though he
claims that they cannot acquire the “right of complaint” typically
associated with testimonial belief (2005: 22). This seems to me to be a
result of his not having squarely in view the nature and significance of
the epistemic right of deferral. If, as I argued in the previous section,
we must appeal to the epistemic right of deferral in order to establish
that accepting a speaker's testimony takes a distinctive interpersonal
form, then it is hard to see how a belief could be justified in this
distinctively interpersonal way without generating an entitlement to
defer challenges.
(20) Moran's response to Coady's case of the hypnotized patient is

actually somewhat different. Moran claims that in Coady's case the
hypnotized subject is not presenting himself as properly responsible for
what he says (2005: 18). This seems to be because Moran is thinking of
the case as one in which the subject's utterance takes place in a
hypnotic trance and so isn't a genuine instance of testimony. This is
different from Owens's case in which only the patient's belief was
induced through hypnosis, not her assertion. The assertion in Owens's
case thus appears to amount to a genuine case of sincere testimony.
(21) Recall that in her recent work, Fricker (2007), an advocate of the
evidential model, holds that tellings must at least be addressed to
someone, though not necessarily to the person acquiring knowledge
from the testimony.
(22) At the beginning of George Orwell's 1984, for example, Winston

Smith begins keeping a secret diary, and he wonders explicitly about to
whom his secret diary is being addressed (1949: 26–27).
(23) There may even be some nonverbal but nevertheless openly

communicative expressions of psychological states that actually call for
an audience to trust a speaker for the truth in the way characteristic of
testimony. Imagine a case of a person's storming out of a room and
slamming a door in a huff. Such an expression of anger might be openly
intended to communicate to an audience that the person is angry and
thus, epistemically speaking, it might come very close to a case of
testimony. In expressing her anger in this particular way the person
might very well intend to communicate to her audience that she is
angry in a way that calls for her audience to trust her for the truth.
Now, certainly, the audience will be in a clear position to simply observe
that the person is angry by witnessing her explosion, but it might
nevertheless be the case that the person intends her audience to
believe her that she is angry rather than to simply observe her anger in
her expression. Overtly communicative expressions like this one seem
to amount to more than the person's behaving expressively in front of
an audience. The person is not simply performing. She is doing
something closer to telling her audience that she is angry.
(24) Burge, for example, claims that our entitlement to believe the
content of conversational implicatures is different from our entitlement
to believe the literal content of what is said, including conventional
implicatures (1993: 483n).
(25) A thought like this is likely at work behind the idea that one can
avoid the moral wrong involved in lying by instead exploiting
conversational implicatures. Peter Geach, for example, appeals to the
following story of St. Athanasius: “If we read the lives of the Saints, we
see how they managed to avoid lying in crises. St. Athanasius was
rowing on a river when the persecutors came rowing in the opposite
direction: ‘Where is the traitor Athanasius?’ ‘Not far away’, the Saint
gaily replied, and rowed past them unsuspected” (1977: 114). If one
thinks that an audience is always in the position of coming to its own
conclusion about implicated content, then one might think that a
speaker is not morally responsible for what an audience concludes
concerning this content.
(26) Thanks to Nat Hansen for drawing my attention to this passage.
(27) Note that if one can assume just as much responsibility for the
implicatures of one's utterance as for the content literally expressed,
then one may not always be able to avoid the moral wrong involved in
lying by instead exploiting conversational implicatures. See note 25.
(28) For a recent attempt to understand the interpersonal nature of the

epistemology of testimony by construing the process of testimonial
belief-formation on analogy with memory, see Goldberg (2010).
Trusting A Person

Benjamin McMyler

Print ISBN-13: 9780199794331
DOI: 10.1093/acprof:oso/9780199794331.001.0001
Trusting A Person
Benjamin McMyler
DOI:10.1093/acprof:oso/9780199794331.003.0005

One might object to the interpersonal account of the epistemology of
testimony developed in chapters 2 and 3 by arguing that relations of
interpersonal trust are a matter for ethics, not epistemology. Indeed,
theorists of trust often claim that there is an inherent tension between
interpersonal trust and theoretical rationality. This chapter argues that
this is not the case. Interpersonal trust can be aptly construed as a
form of belief justified in virtue of the very same kind of interpersonal
relations that serve to justify testimonial belief.
Keywords: trust, rationality, belief, propositional attitudes, the second person, bipolar
normativity
Trust is an important and distinctive kind of psychological state. On the

one hand, trust appears to be a cognitive attitude similar to belief, an
attitude that involves taking something to be true with the aim of
getting it right. When I trust a person to do something, I take it that it
is true that the person will do it, and I generally take myself to have
good reasons for so taking it. On the other hand, trust appears to
involve a kind of interpersonal dependence on the person trusted in a
way that ordinary belief does not. In virtue of this, if a person that I
trust fails to do what I trust her to do, then barring mitigating
circumstances, she lets me down in a way that she does not if she fails
Trusting A Person
to do what I merely believe she will do. The task for the theorist of trust
is to make out how these can both be true at once, how trusting a
person can be an attitude that is both genuinely aimed at getting things
right and irreducibly dependent on the person trusted.
Theoretical accounts of trust typically emphasize one of these aspects

of trust while downplaying the other. Social scientists and political
theorists typically portray trust as a very ordinary kind of cognitive
attitude, an attitude of accepting something as true that is
paradigmatically manifested in a disposition to cooperate with others in
situations of iterated exchange.1 We might call this cognitivism about
trust. Unsurprisingly, the problem (p.114) with cognitivist accounts of
trust, as well as with empirical studies meant to model and measure
trust, is that it is often difficult to see how such accounts are accounts
of trust rather than of mere belief, reliance, or rational cooperation.2
Whatever exactly trust involves, it seems to involve something more
than a rational disposition to cooperate with or rely on another. This is
something that is typically emphasized by philosophers interested in
trust, but philosophers often take this as reason to believe that trust is
not a robustly cognitive attitude.3 Philosophers often take trust to be an
affective attitude that can occur in the absence of or needn't be fully
responsive to evidential considerations. Karen Jones thus claims that
trust is analogous to blinkered vision: “it shields from view a whole
range of interpretations about the motives of another and restricts the
inferences we will make about the likely actions of another. Trusting
thus opens one up to harm, for it gives rise to selective interpretation,
which means that one can be fooled, that the truth might lie, as it were,
outside one's gaze” (1996: 12). In this way, philosophers often take the
normative constraints on trust to be very different from the normative
constraints on belief. We might call this noncognitivism about trust.
However, noncognitive accounts of trust seem to be in tension with our
ordinary practice of trusting others. We typically trust the people we
have most reason to trust, and hence our ordinary practice of trusting
others seems to track evidence of trustworthiness quite closely. In this
respect, the normative constraints on trust appear to be very much like
the normative constraints on belief.
This very brief sketch of some of the literature on trust is extremely

abstract, and it doesn't do justice to the sophisticated ways in which
particular theorists of trust attempt to accommodate these competing
intuitions. Nevertheless, I hope that what I have said here helps to
illustrate what is at stake in theoretical debates about the nature of
trust. The more we view trust as a cognitive attitude based on evidence,
the more difficult it becomes to accommodate the intuitive sense in
Trusting A Person
which trust involves a (p.115) kind of interpersonal dependence on the

person trusted; and the more we view trust as a noncognitive, affective
attitude, the more difficult it becomes to make sense of the way in
which trust seemingly is and ought to be responsive to evidence.
I think that the key to resolving these issues and to providing a fully
adequate account of trust will involve admitting that interpersonal
relations can play an irreducible role in epistemically supporting
certain kinds of belief. In this chapter I will argue that trusting a person
to ϕ always involves believing that the person will ϕ. Unlike cases of
believing that the person will ϕ that do not involve trust, however,
trusting a person to ϕ involves believing that the person will ϕ in a way
that is epistemically supported in virtue of the interpersonal
relationship existing between truster and trusted. The belief that the
person will ϕ is justified in virtue of the interpersonal relationship
between truster and trusted in much the same way that an audience's
testimonial belief is justified in virtue of the interpersonal relationship
between speaker and audience. In the case of testimonial belief, we
have a speaker performing an explicit act of testifying that amounts to
an overt assumption of responsibility for the audience's belief. In the
case of nontestimonial trusting belief, we often lack such an overt
assumption of such responsibility, but nevertheless the truster's belief
that the trusted will ϕ is justified in much the same way in virtue of a
standing interpersonal relationship between truster and trusted. Even
though there isn't an explicit act of testimony that serves as the
truster's reason for belief, there is nevertheless an implicit second-
personal reason for belief generated by the standing interpersonal
relationship. Like testimonial belief, interpersonal trust is thus a form
of belief in which epistemic responsibility for the belief is shared out
between truster and trusted.
Sections 4.1 and 4.2 examine some of the grammar of ordinary trust
ascriptions and propose that trust ought to be understood as a member
of an interesting class of psychological states that I call second-
personal attitudes. Sections 4.3 and 4.4 then show how this proposal
can help to make sense of the nature of trust. My hope is that
construing trust as a second-personal attitude will help us to make
sense of the way in which trust is interpersonal without having to claim
that it is noncognitive and hence without having to claim that there is
an inherent tension between trust and theoretical rationality.
4.1 The Grammar of Trust

Consider the following sentence:
Trusting A Person
(1) I trust Mary to pick up the kids today.
According to standard propositional attitude psychology, such a self-

ascription of trust can be understood as a self-ascription of a
psychological attitude that is “directed to” (p.116) or “about” a
particular content. (1) ascribes to me the attitude of trust with respect
to a particular content, the content being that which follows the
attitudinal verb in the sentence. This seems fine as far as it goes.
However, standard propositional attitude psychology takes the content
towards which the attitude is directed to be expressible in the form of a
proposition embedded in a that-clause. In this sense, the attitude
ascription in (1) should be analyzed in the form of (2).
(2) I trust that Mary will pick up the kids today.
Certainly, (2) is a grammatically well-formed sentence, as is (1). Both

(1) and (2) are instances of grammatical forms of trust ascriptions that
are common in ordinary English. According to standard propositional
attitude psychology, however, the propositions expressed by (1) and (2)
should be the same. (1) should be analyzable into (2) without a change
in meaning. This doesn't seem to hold in this case, however. The
intuitive truth conditions of (1) and (2) appear to be different, and this
is evident from the fact that we can imagine cases in which the
proposition expressed in (2) is true while the proposition expressed in
(1) is false.
Imagine that I am a police detective and that I have been surveilling

Mary for several weeks. I come to know her routine, the days on which
she picks up her kids from school, the days on which she has her
personal assistant do it, and so on. One day a fellow detective asks me
if I have any idea what Mary will be doing today after work, and I
respond, “I trust that Mary will pick up the kids today.” This use of the
trust-that locution seems to make sense. It might sound a bit old
fashioned, but there is nothing illicit about its use in circumstances like
this one. In different circumstances, the trust-that locution can be used
to convey something like a veiled threat, as in a boss's saying to an
employee “I trust that you won't be late this time,” but this is not how it
is being used here. In the present case, it seems that the trust-that
locution is simply being used to convey the fact that I believe (with a
greater or lesser degree of confidence) that Mary will pick up the kids
today.
So we can imagine that it is true in this case that I trust that Mary will
pick up the kids today. But is it true that I trust her to pick up the kids
today? It seems that in this case the proposition expressed by (1) might
Trusting A Person
very well be false. I might believe Mary to be a crook—perhaps this is

why I am surveilling her—and so I might not trust her at all.
Nevertheless I still might trust that she will do various things, in the
sense that I believe that she will do them. The intuitive truth conditions
of (1) and (2) thus appear to be different. We can imagine cases in
which (2) is true while (1) is false. There may be many cases in which
both (1) and (2) are true, and it may be the case that whenever (1) is
true (2) is true, but if there are cases in which (2) is true while (1) is
false, then (1) and (2) must express different propositions.
(p.117) Note that the way in which (1) and (2) express different
propositions has nothing to do with contextual variation. (1) and (2) will
certainly express different propositions in different contexts due at
least to the presence of the indexicals ‘I’ and ‘today.’ What I have
claimed, however, is that (1) and (2) express different propositions in
the same context where the referents of all the contextually variable
elements of the sentences turn out to be the same. In a single context,
it can be true that I trust that Mary will pick up the kids today while it
is false that I trust her to pick up the kids today.
If (1) and (2) express different propositions, then it appears that the
attitude ascribed in (1) cannot be straightforwardly analyzed into the
form of an attitude towards a proposition embedded in a that-clause.
Doing so actually yields a different proposition, a proposition with
intuitively different truth conditions. So what is the problem? Why is it
that ascriptions of trust in the form of (1) cannot be analyzed into
ascriptions of trust in the form of (2)?
It is instructive to note that there are significant syntactic and semantic

differences between the sentences in (1) and (2) due to the fact that
‘Mary’ functions as a syntactic and semantic argument of the verb
‘trust’ in (1) but not in (2). The syntax of the sentence in (1) contains an
argument place, here filled by the word ‘Mary,’ that is absent from the
sentence in (2). This is shown by the fact that the transformation from
(1) to (2) results in a change in meaning while the transformation from
(1) to (3) does not.
(3) I trust Mary that she will pick up the kids today.
In (3) the syntactic and semantic role played by ‘Mary’ in (1) is

preserved, even while the infinitive clause is replaced by a sentence.
The intuitive truth conditions of the propositions expressed by (1) and
(3) are thus, I submit, the same. Indeed, one might think that sentences
of the form of (1) are all fully analyzable into the form of (3). Sentences
of the form of (1) can thus be understood to be ascriptions of
Trusting A Person
propositional attitudes, of attitudes directed towards propositions, but

they are directed towards their propositional contents only via being
directed towards a person. The point is that the syntactic and semantic
argument position occupied by ‘Mary’ cannot be eliminated without a
change in meaning.
There are at least two further linguistic considerations that support the
view that sentences like (1) take a syntactic and semantic argument
that is missing from sentences like (2).4 First, the position occupied by
‘Mary’ in sentences like (1) cannot be occupied by “expletive” or
semantically vacuous arguments. Compare (4a-b) and (5a-b):5 (p.118)
(4)
a. *I trust there to be a resolution tomorrow.
b. I trust that there will be a resolution tomorrow.
(5)
a. *I trust it to be clear what my preference is.
b. I trust that it is clear what my preference is.
The (a) examples are ungrammatical because the syntax of these

sentences associates ‘there’ and “it” with semantically contentful roles.
However, in these cases ‘there’ and “it” are semantically vacuous; they
are placeholders used to fill syntactic positions without a genuine
semantic role. The ungrammaticality of the (a) examples as opposed to
the (b) examples thus reveals that there is a significant semantic role
being played by the position occupied by ‘Mary’ in (1) that is absent
from (2).
Second, sentences like (1) exhibit active-passive nonsynonymy in a way

that sentences like (2) do not. Compare (6a-b) to (7a-b):
(6)
a. I trust Mary to pick up the kids.
b. I trust the kids to be picked up by Mary.
(7)
a. I trust that Mary will pick up the kids.
b. I trust that the kids will be picked up by Mary.
(7a-b) mean the same thing while (6a-b) clearly do not. It is a feature of
passivization that passivization does not of itself affect the linking of
semantic roles to syntactic positions. The transformation from (7a) to
(7b) thus does not affect the meaning of the sentence. However, the
transformation from (6a) to (6b) does affect the meaning of the
sentence, indicating that there is a linking of semantic role to syntactic
Trusting A Person
position in sentences of the form of (6a-b) that is lacking from

sentences of the form of (7a-b). Passivization across this linkage results
in a change of meaning. In (6a) it is Mary that I trust, while in (6b) it is,
strangely, the kids.
I think that this is all strong evidence for the view that sentences like
(1) contain a syntactic and semantic argument position that is lacking
from sentences like (2). Let's call the syntactic and semantic role
played by ‘Mary’ in (1) the locus of trust role. It is the fact that ‘Mary’
takes a locus of trust role in (1) that is lacking in (2) that makes the
intuitive truth conditions of the propositions expressed by (1) and (2)
different. Moreover, in sentences like (1), ‘Mary’ is playing a dual
semantic role. Not only does ‘Mary’ occupy the locus of trust role, but
‘Mary’ is also a semantic argument of the embedded proposition
following the verb. This can be made explicit by analyzing sentences of
the form of (1) into sentences of the form of (3). (3) makes perspicuous
the dual semantic role played by ‘Mary’ in (1).
(p.119) Theorists of trust haven't sufficiently appreciated the syntactic

and semantic difference between trust ascriptions of the form of (1) and
trust ascriptions of the form of (2).6 There is, however, a common
philosophical distinction that goes some distance towards capturing
this difference. Philosophers often distinguish between propositional
attitudes and objectual attitudes, between mental states that are
“about” or “directed upon” a proposition and mental states that are
“about” or “directed upon” an object. Whereas the state of believing
that Mary will pick up the kids can be understood as an attitude of
belief that is directed upon the proposition “Mary will pick up the kids,”
the state of resenting Mary can be understood as an attitude of
resentment that is directed upon Mary herself. In (2) the attitude of
trust is clearly directed upon the proposition “Mary will pick up the
kids today.” In (1) and (3), however, the attitude of trust appears to be
directed on both a person, Mary, and the proposition “Mary will pick up
the kids.” In fact, the attitude of trust appears to be directed on the
proposition “Mary will pick up the kids” via being directed on Mary
herself. The way in which the attitude is directed on its propositional
content is somehow mediated by its direction on Mary herself.
For our purposes, it is important to note that in this respect the attitude
of trusting a person to ϕ is similar to the attitude of believing a person
that p. Trust ascriptions of the form of (3) exhibit a similar syntactic
and semantic structure to belief ascriptions of the form of (8):
Trusting A Person
(8) I believe you that it's cold there.7
Like (3), the verb is here followed by a personal object and then by a
proposition where the personal object appears somehow to mediate the
way in which the attitude is directed on its propositional content.
Eliminating the syntactic and semantic role played by ‘you’ in this
sentence clearly changes the intuitive truth conditions of the sentence.
(8) is clearly different than (9).
(9) I believe that it's cold there.
(p.120) Certainly, I can believe that it's cold there without believing you
that it's cold there.
As it turns out, there are a whole range of attitudinal verbs that are
very much like ‘trust’ in this respect, including, notably, verbs that refer
to what Peter Strawson (1974) calls “personal reactive attitudes.”8
(10) I resent him for doing that.

(11) I love her for saying that.
(12) I hate him for doing that.
(13) I forgive you for saying that.
I submit that in all of these cases we have an attitudinal verb followed

by both a personal and (what can be analyzed as) a propositional object
where the personal object somehow mediates the way in which the
attitude is directed on its propositional content.9
I think it would be a surprise if the grammatical commonalities

exhibited by this range of verbs was a mere coincidence and didn't
reflect some common underlying feature of the mental states or
attitudes referred to by the verbs. I therefore take these grammatical
considerations to be good defeasible evidence for the hypothesis that
the range of mental states or attitudes referred to by these attitudinal
verbs have a common feature, a feature having to do with the way in
which the attitudes are directed towards the persons that figure as the
personal objects of the verbs.
This is supported by the difference in the intuitive truth conditions

between sentences like (1) and (2). What distinguishes (1) from (2) is
that the attitude in (1) is directed to Mary in a way that the attitude in
(2) is not. Trusting Mary to pick up the kids is different from trusting
that she will pick up the kids due to the peculiar way in which, in
trusting Mary, the attitude of trust is directed to her. As it turns out, the
range of attitudinal verbs that we have identified allow for distinctions
in the truth values of the sentences in which they appear very much
Trusting A Person
like that between (1) and (2), distinctions that render their analysis into
standard propositional attitude form problematic. The intuitive truth
conditions of the sentences (10–13) in the left-hand column below
appear to be different from those of the corresponding sentences (14–
16) in the same row in the right-hand column.
(10) I resent him for doing that. (14) I resent that he did that.
(11) I love her for saying that. (15) I love that she said that.
(12) I hate him for doing that. (16) I hate that he did that.
(13) I forgive you for saying that.
(p.121) Syntactically, the difference between the sentences in the left-

hand column and the sentences in the right-hand column is that the
personal object that appears in the sentences in the left-hand column is
eliminated in the corresponding sentences in the right-hand column
and replaced by a that-clause. Doing this, however, changes the
intuitive truth conditions of the sentences. We can imagine cases in
which the sentence in the right-hand column is true while the sentence
in the same row in the left-hand column is false. Arguably, (14) can be
true while (10) is false; (15) can be true while (11) is false; and (16) can
be true while (12) is false. In all of these case, the sentences in the
right-hand column that eliminate the personal object from the
corresponding sentences in the left-hand column also seem to eliminate
the peculiar directedness to a person that characterizes the attitudes
ascribed by the sentences in the left-hand column. The attitudes
referred to in the right-hand column are all about people but without
being directed to them in the peculiar way that characterizes the
attitudes referred to in the left-hand column. In this respect, it is
interesting that (13) doesn't even admit of a formulation that eliminates
the personal object in favor of a that-clause. It seems that the attitude
of forgiveness can't take another person as an object without being
directed to the person in the peculiar way characteristic of the personal
object position of the verb.10
Trusting A Person
4.2 Second-Personal Attitudes

If the intuitive truth conditions of the sentences in the left-hand column
above are different from the intuitive truth conditions of the
corresponding sentences in the right-hand column, then the sentences
express different propositions. I have suggested that the difference
between the propositions in the right and the left-hand columns is that,
due to the presence of the personal object in the sentences in the left-
hand column, the attitudes referred to by the propositions in the left-
hand column are directed to people in some peculiar way that needs to
be spelled out. This directedness to a person is not simply a matter of
people figuring into the content of the attitude. People figure into the
contents of the attitudes in the right-hand column as well. Rather, it is
the particular way that people figure into the contents of the attitudes
in the left-hand column that renders these attitudes distinctive. The
personal objects of the attitudes in (p.122) the left-hand column
actually serve to somehow mediate the way in which the attitude is
directed on its propositional content.
I want to claim that the difference between the attitudes in the right-
and left-hand columns is that the attitudes in the left-hand column are
directed to people second-personally while the attitudes in the right-
hand column are directed to people only third-personally. The people
who are referred to by the personal objects in the sentences in the left-
hand column are being picked out second-personally, while the people
who are referred to in the that-clauses of the sentences in the right-
hand column are being picked out third-personally. We can thus call the
attitudes that appear in the left-hand column second-personal attitudes
and the attitudes that appear in the right-hand column third-personal
attitudes.
This might seem like a strange use of the distinction between the
grammatical second and third-person. The second-person pronoun,
‘you,’ is typically used to refer to the addressees of an utterance, and
none of the above second-personal attitudes require the addressing of
an utterance. Nevertheless, I think it is illuminating to think of these
attitudes as implicitly addressive. I have said that the attitudes in the
left-hand column are directed to people in a way that the attitudes in
the right-hand column are not. I now want to claim that the particular
way in which these attitudes are directed to people is by being
implicitly addressed to them. Like the cases in which an utterance is
explicitly addressed to a person, these second-personal attitudes
implicitly call for a kind of reciprocal responsiveness from their
Trusting A Person
addressee and thereby presume upon an interpersonal relationship

between the subject of the attitude and the personal object.
Admittedly, making out just how these second-personal attitudes are

implicitly addressed to another is a difficult task. Stephen Darwall
(2006b) has recently argued that Strawsonian reactive attitudes are
second-personal in something like this respect. Strawson (1974) seeks
to distinguish a range of attitudes that essentially involve participation
in a relationship with others from those that involve taking up a purely
objective attitude towards others. He thinks of the former as attitudes
that are essentially reactions to the good or ill will of others, and he
terms these “reactive attitudes.” He then distinguishes between
“participant” or “personal” reactive attitudes like gratitude,
resentment, forgiveness, love, and hurt feelings, and “impersonal”
reactive attitudes like moral indignation, obligation, guilt, and remorse.
The former are attitudes of the particular individuals participating in
the interpersonal transaction itself, while the later are what Strawson
takes to be impersonal or generalized analogues of these attitudes that
are felt from the impartial, moral point of view. Darwall thinks that all
of this can be cashed out in terms of the second person.
I claim that reactive attitudes are always implicitly second-

personal and that they therefore invariably carry presuppositions
of second-personal address about the (p.123) competence and
authority of the individuals who are their targets, as well as about
those who have them. Personal reactive attitudes are felt as if
from the second-person standpoint of a relevant transagent, and
impersonal reactive attitudes are felt as from the standpoint of
members of the moral community. (2006b: 67)
According to Darwall, the sense in which reactive attitudes essentially

involve participation in a human relationship is by being implicitly
addressive of another in a way that presupposes the other's
competence to be appropriately moved by such address and her
authority to reciprocally address such attitudes herself. Personal
reactive attitudes do this from the perspective of the agents involved in
the interpersonal transaction itself, while their impersonal analogues
do so from the perspective of the moral community.
I have admitted that it might seem like a stretch to view the range of
attitudes that I have identified as involving a kind of implicit second-
personal address. Certainly, it is even more of a stretch to view
impersonal attitudes like moral indignation, guilt, and remorse as
essentially second-personal. Both Strawson and Darwall are concerned
to draw out the reactive or second-personal nature of these moral
Trusting A Person
attitudes, but this is a far more ambitious project than I want to take on
here. I will thus limit my attention to the personal reactive attitudes, to
attitudes like gratitude, resentment, love, and forgiveness.
Let's consider resentment. Darwall writes: “Resentment is felt as if in

response to a violation of a legitimate claim or expectation, and not
simply as directed toward the violator, but as implicitly addressing her.
This is what makes resentment “reactive” rather than “objective.” It is
a form of “holding responsible,” an address of the other as a person
with the capacity and standing to be addressed in this way and
charged” (2006b: 72). Darwall here construes the addressive aspect of
resentment as a matter of the way in which resentment involves
holding a person responsible and thereby assumes that the person is in
a position to appropriately respond to her being so held. The attitude of
resentment is called for only in situations in which the person resented
is capable of appreciating the resentment that others may feel towards
her, capable of seeing herself as one person among many and as
bearing relations of responsibility towards the others with whom she is
in community. This interpersonal structure is then built into the very
attitude of resentment itself. Resenting S for ϕ-ing is an attitude that
involves characterizing S in a second-personal way, as a person
standing in interpersonal relations of responsibility with the subject of
the attitude. In this sense, attitudes that are directed to others second-
personally involve locating those others in a kind of interpersonal
logical space, a logical space articulated in terms of relations of
responsibility obtaining between the subject and the persons that are
referred to by the personal object.
(p.124) Michael Thompson has suggested something similar. Thompson

(2004) argues that there is an important difference between judgments
like “X wronged Y in doing A” and “X did wrong in doing A.” The former
judgment, but not the latter, exhibits what Thompson calls bipolar
normativity: “In all such [bipolar normative] judging, whatever the
determinate form, I may be said to view a pair of distinct agents as
joined and opposed in a formally distinctive type of practical nexus.
They are like for me the opposing poles of an electrical apparatus: in
filling one of these forms with concrete content, I represent an arc of
normative current as passing between the agent-poles, and as taking a
certain path” (2004: 335). Thompson thinks that there is a whole range
of judgments that exhibit a bipolar normative structure in which the
subject and the person that figures as the object are related to one
another as opposing poles. One pole is always responsible to the other,
and one pole is always owed something by the other. In this sense,
these bipolar normative judgments are different from merely monadic
Trusting A Person
judgments like “X did wrong in doing A.” Whenever X wrongs Y in doing

A, X does wrong in doing A, but it is not the case that, whenever X does
wrong in doing A, X wrongs someone. To wrong someone is to do
something above and beyond simply doing wrong, something that
cannot simply be identified with doing wrong. To judge that X wronged
Y in doing A is to locate X and Y in a particular bipolar interpersonal
nexus and to judge that, in doing A, X has violated the interpersonal
demands that characterize this nexus. Thompson thinks that persons
can enter into many different kinds of bipolar normative relationships,
from those of etiquette and the rules of games to those of justice and
morality. Bipolar normative thought can get a hold in all of these
various domains, though with respect to each particular domain it must
be, as he puts it, “shifted into a particular gear” (2004: 342).
On the face of it this distinction between bipolar normative and merely

monadic judgments looks to be a distinction with respect to what the
judgments are about. Bipolar normative judgments look to assert of
people that they stand in a certain interpersonal relationship, while
merely monadic judgments do not. Thompson wants to argue for
something much stronger than this, however. He writes: “Thought
takes a distinctive turn here, a turn which cannot simply be reduced to
its taking a certain body of concrete relations, practical relations, as its
theme—and still less by making reference to a special class of objects:
namely, agents. Such thought has, among other things, a novel and
particular relation to what it is about” (2004: 337). Thompson holds, I
take it, that the thinking of such bipolar normative thoughts involves
the adoption of a particular “posture of mind” in which the thinker of
the bipolar thoughts views herself as related to those she makes the
judgment about in a bipolar normative way. In this sense, we should
speak of a special bipolar form of thought and not just a special bipolar
form of fact (2004: 348).
(p.125) Most interestingly for our purposes, Thompson goes so far as to

float the idea that a whole range of psychological attitudes might
exhibit this kind of bipolar normativity:
“The bipolarity that interests us might also be found in the

representation of certain states of directed feeling that you might
bear toward Sylvia, or Sylvia toward you. Judgments employing
the concepts of grievance, grudge bearing, and resentment would
be clear examples. A philosophical comprehension of these
concepts of feeling presupposes a grasp of this formal feature of
the judgments in which they are exercised” (2004: 348n).
Trusting A Person
If this is right, then bipolar normativity seems in the end to

characterize a sui generis kind of mental directedness. It characterizes
the way in which attitudes like resentment are directed to the persons
who figure as their objects, and hence judgments about resentment,
such as “X resents Y for doing A,” amount to bipolar normative
judgments. In this sense, the interpersonal logical space in which
second personal attitudes work to locate the subject and personal
object of the attitude is a space that is articulated in terms of bipolar
normativity. It is a space that is articulated in terms of one pole's being
owed something from the other and in terms of the other pole's being
responsible to the former.
Thompson's distinction between bipolar relational judgments like “X

wronged Y in doing A” and merely monadic judgments like “X did wrong
in doing A” should strike one as very similar to the distinction between
the two different forms of attitude ascription in our right and the left-
hand columns. I have said that the attitudes that appear in the left-hand
column are directed to the people who figure as personal objects
second-personally, while the attitudes that appear in the right-hand
column are directed to people only third-personally. We can now say, I
think, that the propositions in the left-hand column are bipolar
normative judgments, while the propositions in the right-hand column
are merely monadic judgments. The attitudes ascribed by the
propositions in the left-hand column are thus distinguished from the
attitudes ascribed by the propositions in the right-hand column by the
fact that they are directed to others in a way that involves locating
them in a particular bipolar interpersonal nexus.
Note, however, that while Darwall and Thompson hold that particular
psychological attitudes like resentment and forgiveness exhibit a
second-personal, bipolar structure, our left-hand column is not
distinguished from our right-hand column by involving the ascription of
completely different attitudes. Our left- and right-hand columns involve
the ascription of the same attitudes (or at least attitudes with the same
name) only with a different content. Hence ascriptions of resentment
appear on both sides, and so the attitude of resentment appears to
admit of both second-personal (bipolar) and merely third-personal
(monadic) directedness to persons. I think this helps to show that the
feature of mindedness that Darwall and Thompson have identified is in
(p.126) fact a distinctive kind of mental directedness. Hence, particular
psychological attitudes can partake of both second and third-personal
mental directedness. Resentment, for example, can be an attitude that
is directed to a person in the second-personal way that Darwall and
Trusting A Person
Thompson have identified, but it can also be directed to people merely

third-personally.
Still, one may have been worrying for some time that my distinction
between the attitudes in the right and the left-hand columns is strained.
Though there may be a grammatical distinction between “I resent him
for doing that” and “I resent that he did that,” one might think that in
the end there is really no difference in meaning between these two
sentences. There is no substantive distinction between resenting a
person for doing something and resenting that a person did something,
and hence these two sentences both express the same proposition.
Indeed, philosophical discussions of the reactive attitudes don't
typically admit that there may be cases in which they are not really
reactive at all, cases in which they are directed to people only third-
personally. Strawson certainly doesn't admit that attitudes like
resentment may take nonreactive forms. If it is not fully reactive, then
it is not really resentment, no matter what the grammatical differences
are between particular sentences that ascribe the attitude.
I am sympathetic to this worry. It is fine with me if, in the end, one

wants to assert that there is no difference in intuitive truth conditions
between “I resent him for doing that” and “I resent that he did that.”
This will ultimately turn on a much more detailed account of the
particular attitude of resentment than I am in a position to give here.
Even if the distinction between the attitudes ascribed in our right- and
left-hand columns breaks down in particular cases, however, I don't
think that this is an objection to the general distinction that the two
columns are meant to illustrate. As we've seen, some attitudes may only
be able to appear in one column. This is clearly the case for
forgiveness. There isn't even a form of the verb ‘forgive’ in which it can
take a that-clause as complement. The present objection concerning
resentment simply states that, even though the verb ‘resent’ can
grammatically take a that-clause as complement, the attitude so
ascribed is really like forgiveness; the attitude so ascribed can only
take the fully reactive form more clearly represented by ascriptions in
which the verb takes a personal object. Even if this is true, however, it
doesn't seem to be the case for all of the attitudes that appear in our
columns. There is, as I have argued, a genuine distinction between
trusting that Mary will pick up the kids today and trusting Mary to pick
up the kids today. In this sense, it seems to be a virtue of our two
columns that they make perspicuous how some particular attitudes
seem to fall on only one side while other attitudes seem to fall on both
sides. Forgiveness (and perhaps some other reactive attitudes) seem to
fall squarely on the second-personal side, while attitudes like wanting
and desiring (despite the fact that, in English, they can sometimes take
Trusting A Person
a (p.127) personal object) seem to fall squarely on the third-personal

11
side. Other attitudes, including the attitude that I am concerned with
here, the attitude of trust, seem to fall on both sides; they seem to be
able to be directed to people both second-personally and third-
personally. We can always argue about where a particular attitude does
indeed fall, but this will be an argument about the particular attitude
involved. It will be an argument concerning the kind of mental
directedness that pertains to the particular attitude in question.
4.3 Trust as Second-Personal

I have argued that the sentences “I trust Mary to pick up the kids
today” and “I trust that Mary will pick up the kids today” express
different propositions. The latter can be true while the former is false,
suggesting that trusting a person is somehow different from trusting
that a certain state of affairs does or will obtain. One might have been
suspicious about my drawing such a conclusion from what might seem
to be rather flimsy grammatical considerations, and so I have tried to
demonstrate that there is a systematicity to the grammar, that a whole
range of attitudinal verbs admit of this kind of grammatical distinction.
Still, I have claimed that the existence of this grammatical distinction
doesn't guarantee that the attitude ascribed by the verb can be directed
to people both second and third-personally, so I now want to look at the
attitude of trust in more detail.
I have suggested that the difference between trusting Mary to pick up

the kids today and trusting that Mary will pick up the kids today has
something to do with the way in which the former attitude is directed to
Mary in a way that the latter attitude is not. We can now say that the
attitude of trusting Mary to pick up the kids is directed to Mary second-
personally while the attitude of trusting that Mary will pick up the kids
is directed to Mary only third-personally. The attitude of trusting Mary
involves locating oneself and Mary in an interpersonal logical space
characterized by bipolar normative relations. If I trust Mary, then (1) I
am thereby recognizing her as having a certain second-personal
competence and authority; (2) there is a sense in which she is
responsible to me for doing what I trust her to do; and (3) if she does
not do what I trust her to do, I will thereby be entitled to adopt second-
personal attitudes like resentment back towards her. None of this is
true of the attitude of trusting (p.128) that Mary will pick up the kids
today. If I merely trust that Mary will pick up the kids today, then I am
not thereby recognizing Mary as having a certain second-personal
competence and authority; there is no sense in which she is responsible
to me for doing what I trust that she will do; and if she does not do
Trusting A Person
what I trust that she will do, I will not thereby be entitled to adopt
second-personal attitudes like resentment back towards her.
This may seem to introduce an undesirable bifurcation into our account

of trust. Theorists of trust all readily admit that the concept of trust is
employed in a wide variety of ways in ordinary life. In addition to other
individuals, we often say that we trust groups of people, governmental,
societal, and religious institutions, and even inanimate objects. I might
be said to trust that a rickety-looking footbridge will bear my weight
(and perhaps I could even be said to trust the footbridge to bear my
weight). Given this variety in usage, theorists of trust rightly limit their
attention to cases of interpersonal trust, to cases in which one person
trusts another to do something, the paradigmatic cases of trusting that
seem to be central to the concept of trust. Still, I think it is important to
see that there are cases of trust ascriptions that take another person as
their object but that are not yet fully interpersonal in the sense that
typically concerns philosophers. On the one hand, this helps us to be
clearer about the precise way in which the attitude of trusting a person
is essentially interpersonal. It is interpersonal in that it exhibits a
particular kind of mental directedness that can't be reduced to the kind
of third-personal mental directedness characteristic of standard
propositional attitude psychology. But on the other hand, it also helps
us to see that many ordinary trust ascriptions pretty clearly involve the
ascription to a subject of a straightforwardly cognitive attitude. If I
trust that this footbridge will bear my weight, then pretty clearly this
simply involves my believing (with a greater or lesser degree of
confidence) that the footbridge will bear my weight. It doesn't look like
we have to appeal to anything noncognitive in order to explain this
ascription of the attitude of trust. In this sense, I think that an adequate
account of the fully interpersonal attitude of trusting a person shouldn't
make it look like a mere coincidence that ordinary trust ascriptions can
take this straightforwardly cognitive form. Even if we want to insist
that the concept of trust has its home in the context of interpersonal
relations and hence that we can't fully understand the concept without
understanding its interpersonal instantiations, we need to see how the
interpersonal cases are at least related to the straightforwardly
cognitive ones.
This said, the second-personal characterization that I have offered here

of the fully interpersonal attitude of trusting a person seems to fit
philosophical conceptions of trust quite nicely. As we saw in chapter 3,
it is common in the philosophical literature to distinguish trust from
mere reliance. Annette Baier claims that reliance on another is
Trusting A Person
insufficient for trust; that we can rely on others without yet trusting
them. (p.129)
We may have no choice but to rely on the local shop for food, even
after some of the food on its shelves has been found to have been
poisoned with intent. We can still rely where we no longer trust.
What is the difference between trusting others and merely relying
on them? It seems to be reliance on their goodwill toward one, as
distinct from their dependable habits, or only on their dependably
exhibited fear, anger, or other motives compatible with ill will
toward one, or on motives not directed on one at all. Once we
have ceased to trust our fellows, we may rely on their fear of the
newly appointed security guards in shops to deter them from
injecting poison into the food on the shelves. We may rely on the
shopkeeper's concern for his profits to motivate him to take
effective precautions against poisoners and also trust him to want
his customers not to be harmed by his products, at least as long
as this want can be satisfied without frustrating his wish to
increase his profits. Trust is often mixed with other species of
reliance on persons. Trust which is reliance on another's goodwill,
perhaps minimal goodwill, contrasts with the forms of reliance on
others' reactions and attitudes which are shown by the comedian,
the advertiser, the blackmailer, the kidnapper-extortioner, and the
terrorist, who all depend on particular attitudes and reactions of
others for the success of their actions. We all depend on others'
psychology in countless ways, but this is not yet to trust them . . .
When I trust another, I depend on her goodwill toward me. (1994:
98–99)
For Baier, trust is a particular kind of reliance on another; it is reliance

on another to do something out of the other's goodwill towards one. In
this sense, trust must be distinguished from reliance on another that is
based on a judgment that the other will do something from a motivation
other than goodwill towards one. We may rely on the shopkeeper to
take effective precautions against poisoners out of a concern for her
profits, but Baier thinks this is not yet to trust her. Trusting the
shopkeeper requires taking her action to be motivated by goodwill
towards me, not by a concern with her own profits. I needn't see the
shopkeeper as bearing goodwill towards me in particular, but in order
to trust her I must at least take her action to be motivated by a general
goodwill towards her potential customers.
Trusting A Person
Baier points out that since the attitude of trust involves reliance on the
goodwill of another, the attitude incurs certain risks not associated with
other forms of reliance.
Where one depends on another's goodwill, one is necessarily

vulnerable to the limits of that goodwill. One leaves others an
opportunity to harm one when one trusts, and also shows one's
confidence that they will not take it. Reasonable trust will require
good grounds for such confidence in another's goodwill, or at
least the absence of good grounds for accepting another's ill will
or indifference. Trust, then, (p.130) on this first approximation, is
accepted vulnerability to another's possible but not expected ill
will (or lack of goodwill) toward one. (1994: 99)
Note that being vulnerable to the limits of another's goodwill is

different from simply being vulnerable to the other's not doing what
one relies on her to do. Relying on the shopkeeper to take appropriate
precautions against poisoners out of a concern for her profits does
leave me vulnerable to the risk that she might not in fact do so. It might
turn out that she is not competent to know what the appropriate
precautions are, and it might turn out that she thinks that taking the
appropriate precautions wouldn't be profitable. Still, if I am simply
relying on the shopkeeper, I am not vulnerable to her goodwill. I may be
disappointed by her not taking the appropriate precautions, but I am
not let down by her. If I actually trust the shopkeeper, if I rely on her
goodwill towards me, then I will not simply feel disappointed by her not
doing what I trust her to do. I will feel betrayed. The peculiar risks
associated with trust as opposed to mere reliance thus appear to be
essentially interpersonal in nature. The attitude of trust is directed on
another person and her goodwill towards one in a way that makes
failures of trust a personal affront to the truster.
Expanding on this point, Richard Holton argues that trust is to be

distinguished from mere reliance due to the way in which it involves
adopting what he calls, following Strawson, a “participant stance”
towards others.
I think that the difference between trust and reliance is that trust
involves something like a participant stance towards the person
you are trusting. When you trust someone to do something, you
rely on them to do it, and you regard that reliance in a certain
way: you have a readiness to feel betrayal should it be
disappointed, and gratitude should it be upheld. In short, you take
Trusting A Person
a stance of trust towards the person on whom you rely. It is the

stance that makes the difference between reliance and trust.
When the car breaks down we might be angry; but when a friend
lets us down we feel betrayed. (1994: 67).
For Holton, trust is very much like a Strawsonian reactive attitude. It is

an attitude that we adopt in the context of interpersonal relationships
with others, an attitude that only makes sense in the context of persons
relating to one another as persons. I have claimed that the attitude of
trust is not always directed to people second-personally and hence that
I can trust that Mary will pick up the kids without adopting any kind of
participant stance towards her. Still, Holton's characterization of the
interpersonal attitude of trusting a person is very congenial to the
second-personal account that I have proposed.
However, in characterizing the difference between trust and mere

reliance as involving the adoption of a “stance,” Holton is led to deny
that trusting a person is a genuinely (p.131) cognitive attitude. Holton
claims that the adoption of the participant stance is an essentially
“practical” affair: “Trusting someone does not involve relying on them
and having some belief about them: a belief, perhaps, that they are
trustworthy. What it involves is relying on them to do something, and
investing that reliance with a certain attitude. This is to take a practical
stance” (1994: 67). Holton doesn't offer much more by way of cashing
out how exactly this stance ought to be construed as practical, but
presumably it should be characterized in terms of dispositions to
respond to failures of trust with feelings of betrayal or resentment. In
any case, the point is clearly that the kind of reliance involved in
trusting a person ought to be construed as reliance that is adopted
from some kind of practical perspective and not as reliance based on
evidence. Holton isn't alone in this regard. Darwall also seems to think
that the kind of second-personal mental directedness that he is
concerned with is an essentially practical affair. If this is right, and if
the attitude of trusting a person is an attitude characterized by this
distinctive kind of mental directedness, then it looks like trust cannot
be a robustly cognitive attitude.
Trusting A Person
4.4 Trust as Cognitive

I want to argue that the attitude of trusting a person is both second-
personal and fully cognitive. Trusting a person is an attitude that
involves the truster's locating the trusted in an interpersonal logical
space, but it is also an attitude that essentially involves taking
something to be true with the aim of getting it right. In this sense, trust
is very much like belief. In fact, it might as well be construed as a
species of belief. The difference between trust and other species of
belief, I will argue, has to do with the way in which the attitude is
epistemically supported or justified. The kind of belief involved in the
second-personal attitude of interpersonal trust is justified in virtue of
the interpersonal relationship existing between truster and trusted. In
trusting Mary to pick up the kids today, I believe that Mary will pick up
the kids today. What distinguishes my trusting Mary to pick up the kids
today from my trusting that Mary will pick up the kids today is that, in
the case of the former, my interpersonal relationship with Mary plays
an irreducible role in epistemically supporting my belief that she will
pick up the kids today. Genuine interpersonal trust is belief that is
justified by an at least implicit second-personal reason.
What exactly is the relationship between trust and belief? Philosophers

often assume that trust can be a ground or basis for belief but that (1)
trust is not itself a form of belief and (2) the ground or basis for belief
that trust provides is a nonevidential one. While my trusting Mary may
cause me to believe certain things about her, my trust is not itself a
form of belief and does not epistemically support or justify my beliefs.
As Judith Baker writes: (p.132)
Belief aims at truth—that is, we want our beliefs to be true. So a

very natural and compelling picture of rationality with regard to
belief would be the idea that we should only accept those beliefs
which are likely to be true. Hence, what may be relevant, or,
strictly speaking, reasons, for believing will be those supporting
facts, items, which give evidence for the truth of what is believed.
But trust seems to involve beliefs which are not accepted on the
basis of evidence and beliefs which in some cases may be highly
resistant to evidence that runs counter to them. (1987: 1)
Baker here seems to accept that trust may involve beliefs. Trusting
someone may involve believing certain things about her, for instance,
that she will do certain things. Still, Baker claims that these beliefs are
not based on and can be highly recalcitrant to evidence. For Baker,
trust thus amounts to a nonevidential basis for belief, a kind of evidence
substitute. There is something very intuitive about this. We often think
of trusting as something we must do in the absence of sufficient
Trusting A Person
reasons for belief, as something we are forced to do simply because we

don't have access to enough evidence. Trust is thus conceived as a
necessary but epistemically regrettable basis for belief—“I guess I'll
just have to trust you.”
While there is something right about this idea, I think that the way in
which it is here cashed out is incorrect. Consider in this regard the
following two examples of ordinary cases of interpersonal trust.
Imagine that I tell you that a certain brand of food has just been
recalled, a brand that you happen to purchase regularly. You wonder if
what I have said is true. As far as you know, this brand has never been
recalled before. You ask me, “Are you sure?” “Trust me,” I say, and let's
suppose that you do. You throw out the allegedly recalled brand and
buy a different one. Now imagine that you need me to mail a letter for
you. You tell me that it is very important that the letter is mailed today,
and you ask me if I am sure that I will be able to do this. “Trust me,” I
say, and let's suppose, again, that you do. You plan on the letter's being
mailed today, and you go on to make arrangements that depend on this.
There are two important things to note here. First, in these two cases
your trusting me doesn't seem to merely result in your believing what I
say but to positively require your believing what I say. Imagine what
would happen if you informed me that you didn't actually believe that
the brand of food had been recalled or that the letter would be mailed
today, but that you would rely on my goodwill nevertheless. I might
quite naturally respond, “What, don't you trust me?” If you don't
believe that the brand of food has been recalled, and if you don't
believe that the letter will be mailed today, then no matter whether you
go on to act as if these things are the case, I can justly accuse you of
not trusting me. What I am calling for in asking for your trust in these
cases is your belief in what I say, not merely your practical reliance on
my goodwill. Full-fledged (p.133) trust in these cases positively
requires belief, and to the extent that you don't believe what I say, then
to that extent you don't trust me.12
Second, both of these cases involve simple trusting. That is, they
involve belief in what I say in the absence of other positive reasons for
belief. You believe that the brand of food has been recalled and that the
letter will be made today despite a lack of other positive evidence for
these beliefs. In this respect, you simply trust me. We might imagine
you saying, “I guess I'll just have to trust you.” Clearly, however, this
does not mean that your beliefs in these two cases are not accepted on
the basis of evidence. If we construe evidence broadly to encompass
anything that genuinely counts in favor of the truth of a proposition,
then clearly your beliefs are based on evidence, namely my telling you
Trusting A Person
that the brand of food has been recalled and that I will mail the letter.
Your simply trusting me in these cases does not involve belief in the
absence of evidence but rather belief based on a particular kind of
evidence in the absence of other kinds of evidence. You simply trust me
in that you don't have any other impersonal evidence for believing that
the brand of food has been recalled and that the letter will be mailed
today, but you nevertheless have genuine testimonial evidence for
believing both of these things.
So your trusting me in these two cases seems to positively require your

believing what I say on the basis of a genuinely evidential
consideration, namely my testimony. This suggests that trust is both
intimately connected to belief and not aptly construed as a kind of
evidence substitute. The belief in what I say that is positively required
for your trust in these cases is a belief that is genuinely based on
evidence. In fact, my calling for your trust in these cases seems to
involve nothing more than my calling for you to believe what I say on
the basis of my testimony. If you believe me about what I say, then you
will thereby have fulfilled my call for your trust. Of course, believing me
about what I say will typically require your being disposed to act in
certain ways, to rely on me practically, but this is a feature of belief
generally—we hesitate to ascribe beliefs to others in the absence of
relevant dispositions to act in accordance with these beliefs. The point
is that simply being disposed to act as if the brand of food has been
recalled or as if the letter will be mailed today is insufficient for
trusting me. Trusting me involves actually believing these things, and
believing them on the basis of a particular kind of evidence, namely my
testimony.
The fact that the required belief in these cases is based on a particular
kind of evidence, rather than no evidence at all, appears to be what
makes it the case that these are cases of trust in the first place. You
trust me in these cases precisely in virtue of the fact that you believe
me about what I say. These are cases of trust precisely because they
(p.134) involve belief on the basis of a particular kind of evidence.
Granted, the relevant evidence in these cases is very different from the
kind of impersonal evidence provided by our own observation of
ordinary objects and states of affairs. The evidence in these cases is a
kind of evidence that leaves you dependent on my goodwill towards you
in a way that necessarily incurs certain risks of betrayal not associated
with dependence on other kinds of ordinary evidence. But nevertheless,
your trusting me in these cases appears to be a straightforward matter
of your believing what I say on the basis of a particular kind of
Trusting A Person
genuinely epistemic reason for belief, what I have called a second-

personal reason for belief.
The interpersonal account of testimony that I have offered thus allows

us to make sense of the way in which simple trusting in these cases is
characterized by both belief based on evidence and by an irreducible
interpersonal dependence on the person trusted. In the first of these
examples, your trusting me that this particular brand of food has been
recalled involves your believing that this particular brand of food has
been recalled on the basis of my telling you that it has. My telling you
that this particular brand of food has been recalled is a reason for you
to believe that it has been recalled in virtue of the way in which my
telling you this involves my entering into an interpersonal relationship
with you in which I assume partial epistemic responsibility for your
belief. Similarly, in the second of the above examples, your trusting me
to mail the letter involves your believing that I will mail the letter on
the basis of my testimony, where my testimony provides you with the
particular kind of reason for belief that it does in virtue of the
interpersonal relationship between us. The second-personal nature of
the attitude of trusting a person is here cashed out in such a way that it
is fully consistent with the attitude being a form of belief that is fully
subject to the norms of theoretical rationality. Trust is a second-
personal attitude insofar as the way in which it is directed on its
propositional content is mediated by its directedness on a person, but
the nature of this mediation is a matter of the way in which the attitude
is epistemically supported. The person occupying the locus of trust role
is here epistemically mediating the truster's relation to the facts.
Recall that outright ascriptions of belief can themselves take the

grammatical form of ascriptions of what I have called second-personal
attitudes. When I believe you that it's cold there, for example, the way
in which my attitude is directed on its propositional content is mediated
by its directedness on a person, but the nature of this mediation is a
matter of the way in which the person is playing a role in epistemically
supporting my belief. This strongly suggests that robustly cognitive
attitudes can partake of genuinely second-personal mental directedness
in a way that is fully consistent with theoretical rationality. There is
here no tension between the second-personal nature of the attitude and
theoretical rationality, only the existence of a particular kind of reason
for belief.
(p.135) The two cases of trust that I have been discussing admit of this
very same kind of explanation because they are both themselves cases
of believing someone that p. Insofar as these cases of trust are cases of
believing someone that p, it shouldn't come as a surprise that they
Trusting A Person
involve belief on the basis of evidence. But not all cases of trusting a
person are cases of believing someone that p. Cases of believing
someone that p are cases of properly testimonial belief, but not all cases
of trusting a person are cases of properly testimonial belief. I can trust
someone to do something even though she hasn't explicitly told me that
she will do it. For example, if I have an ongoing relationship with Mary,
I might trust her to pick up the kids today even though she hasn't
explicitly told me that she will pick up the kids today. Here the reason
for belief present in the cases of testimonial belief is conspicuously
lacking. So aren't these cases of simple nontestimonial trust cases that
needn't involve belief based on evidence?
I think that the account of trust that I have offered in the case of
testimonial belief can be plausibly extended to cover cases of
nontestimonial trust, though what I will offer here is little more than a
suggestion. First of all, it is important to note that a great many cases
in which we trust someone to do something are cases in which the
person has more or less explicitly told us that she will do it. A great
many cases of ordinary interpersonal trust are thus cases of testimonial
belief, cases in which believing the person is positively required for full-
fledged trust and in which the person's testimony gives us a distinctive
kind of reason for belief. The idea that interpersonal trust generally
functions as a kind of evidence substitute thus seems to involve a
serious discounting of the scope and epistemic significance of
testimony. Nevertheless, there clearly are a great many cases of
interpersonal trust that are not cases of testimonial belief. The extent of
interpersonal trust is wider than that of testimonial belief. A general
theory of the nature of interpersonal trust should thus be able to
account for both testimonial and nontestimonial cases of trust. It should
be able to account for what is common between them in virtue of which
they all amount to cases of interpersonal trust. One way to do this
would be to attempt to understand the interpersonal trust involved in
cases of testimonial belief in a way that is not essentially cognitive. I
have presented reasons for thinking that this isn't very plausible—the
trust called for in cases of testimonial belief positively requires belief
that is epistemically supported by the speaker's testimony. Another way
to do this would be to try to understand interpersonal trust generally as
a species of belief that is epistemically supported in a distinctively
second-personal way. This is what I propose.
I want to suggest that cases of nontestimonial interpersonal trust are

very much like cases in which I believe something on the basis of
outright testimony. When I trust Mary to pick up the kids today even
though she hasn't explicitly told me that she will do this, I believe that
she will pick up the kids today, and my belief that she will pick up the
Trusting A Person
kids today is justified in virtue of the ongoing relationship between us.

In such a (p.136) case the relationship is ongoing and so doesn't need
to be enacted by the offering of a specific piece of testimony, but
nevertheless my belief is still epistemically supported in virtue of this
relationship. Moreover, the epistemic support that my ongoing
relationship with Mary provides serves to parcel out epistemic
responsibility for my belief between Mary and myself in much the same
way that characterizes properly testimonial belief. If a third-party were
to challenge my belief by presenting me with evidence that counts
against Mary's picking up the kids today—for example, that someone
else has claimed that Mary is leaving town after lunch—I would be
entitled to maintain my belief without meeting the challenge myself by
instead deferring it back to Mary. I might respond that, given our
ongoing relationship, Mary wouldn't leave town without telling me and
informing me about the arrangements she has made for picking up the
kids. This would not be the case if, like the detective, I merely trusted
that Mary will pick up the kids. If I merely trusted that Mary will pick
up the kids, perhaps because I have been keeping her under
surveillance, then I would not be entitled to defer to Mary epistemic
challenges to my belief that she will pick up the kids. In this respect,
trusting S to ϕ is differentiated from trusting that S will ϕ by the fact
that all cases of trusting S to ϕ are, in effect, cases of trusting S for the
truth—they are cases of trusting S that p (in this case, trusting S that S
will ϕ). Interpersonal trust is a species of belief that is epistemically
supported in virtue of the very same kind of interpersonal relations that
serve to make outright testimony a distinctive kind of reason for belief.
I have the following general picture in mind. Any case of trusting S to ϕ

is a case of believing that S will ϕ. Just as trusting that S will ϕ pretty
clearly involves believing that S will ϕ, so too trusting S to ϕ involves
believing that S will ϕ. However, trusting S to ϕ is distinguished from
trusting that S will ϕ due to the fact that trusting S to ϕ involves
believing that S will ϕ on the basis of a particular kind of reason. If the
case is a case of testimonial belief, then this reason will be S's
testimony, a distinctively second-personal reason that serves to justify
the audience's belief in virtue of interpersonal relations of authority
and responsibility between speaker and audience. If the case is not a
case of testimonial belief, then there will be no such explicit second-
personal reason for belief. Nevertheless, I think that we should
conceive of such cases as involving a kind of implicit second-personal
reason. We should conceive of nontestimonial cases of trusting S to ϕ as
cases of believing that S will ϕ on the basis of an implicit second-
Trusting A Person
personal reason that justifies the belief in virtue of the ongoing

interpersonal relationship existing between truster and trusted.13
We also need to see the attitude of trusting S to ϕ as itself responsive to

evidence. Considerations of S's trustworthiness amount to evidence
that I must be responsive to (p.137) in order to be in a position for my
relationship with S to provide a good reason for me to believe that S will
ϕ. In this sense, we need to distinguish the kind of reason for belief
involved in trusting from the kind of reason for trusting provided by
considerations of a person's trustworthiness.14 When we deliberate
about whether to trust someone, we do so by weighing the evidence of
a person's trustworthiness. This evidence is assessed with respect to
the degree to which it counts in favor of our adopting the attitude of
trusting someone. It is assessed with respect to the degree to which it
counts in favor of adopting the belief that the person will ϕ on the basis
of a particular second-personal reason. In this respect, evidence of a
person's trustworthiness functions with respect to trusting a person
very much like evidence of the normalcy of lighting conditions and of
the proper functioning of the sense organs functions with respect to
perceptual belief. Evidence of the normalcy of lighting conditions and of
the proper functioning of the sense organs counts in favor of adopting
beliefs on the basis of perceptual evidence, but it doesn't directly count
in favor of the beliefs so adopted. In normal cases, it is not the reason
on which the belief is based. Similarly, evidence of the trustworthiness
of a person counts in favor of adopting beliefs on the basis of the
truster's interpersonal relationship with the trusted, but it doesn't
directly count in favor of the beliefs so adopted. Evidence of Mary's
trustworthiness counts in favor of trusting Mary to pick up the kids
today, but it doesn't directly count in favor of believing that Mary will
pick up the kids today. Just as if Mary had explicitly told me that she
would pick up the kids today, my reason for believing that she will pick
up the kids today is a second-personal reason that is generated by the
interpersonal relationship between us.
I think that distinguishing in this way between the considerations that

justify the attitude of trusting a person to ϕ and the considerations that
justify the belief involved in the attitude of trusting a person to ϕ is
crucial for understanding the interpersonal nature of trust.15 If we see
the belief that the person will ϕ involved in trusting a person (p.138) to
ϕ as justified simply by considerations of the person's trustworthiness,
then it becomes hard to see how there is anything irreducibly
interpersonal about the attitude of trust. Trust looks straightforwardly
cognitive, but it doesn't look sufficiently different from nontrusting
belief. But if we see the belief that the person will ϕ involved in trusting
Trusting A Person
a person to ϕ as justified in virtue of the interpersonal relationship

existing between truster and trusted, and if we see considerations of
the trusted's trustworthiness as a background that puts the truster in a
position to take this interpersonal relationship to generate an at least
implicit second-personal reason for belief, then we can retain a robust
sense in which trust is an irreducibly interpersonal attitude. In
deliberating about whether to trust S to ϕ, I am coming to my own
conclusion about whether the at least implicit second-personal reason
generated by my relationship with S is good reason for believing that S
will ϕ, but when I do in fact trust S to ϕ on this basis, I am not coming to
my own conclusion about whether S will ϕ. In taking my interpersonal
relationship with S to generate a second-personal reason that is a good
reason for believing that S will ϕ, I am giving over partial responsibility
for my belief to S. In this sense, even though I am completely
responsible for coming to my own conclusion about whether S is
trustworthy, even though I am completely responsible for coming to my
own conclusion about whether to trust S to ϕ, when I do trust S to ϕ my
belief that S will ϕ is justified by an irreducibly second-personal
consideration.
So here is my proposed solution to the problem of how trust can be

both cognitive and irreducibly interpersonal. Trust is cognitive in the
sense that trusting S to ϕ involves believing that S will ϕ on the basis of
a genuinely epistemic reason. But trust is interpersonal in the sense
that the reason that justifies the belief is irreducibly second-personal.
As a result, in trusting S to ϕ, I am not fully epistemically responsible
for my belief that S will ϕ. It is in this sense that, in trusting S to ϕ, S is
responsible to me. S is responsible to me in virtue of being partially
epistemically responsible for the justification of my belief that S will ϕ
such that, if a third-party challenges my belief that S will ϕ by
producing evidence that tells against S's ϕ-ing, I am entitled to defer
the challenge back to S.
The account of trust offered here helps to do away with the idea that
there is an irresolvable tension between trust and theoretical
rationality. To see how this is so, consider the following example,
presented by Baker, that is meant to illustrate the supposedly
nonevidential character of beliefs based on trust.
Trusting A Person
Suppose I trust a friend who has been accused of wrongdoing,

with an impressive amount of evidence brought against her.
Typically, I am faced with a novel situation, where there is no
prior set of tests or testing situations that she has come through
with flying colours. Suppose she is accused of selling secrets to a
foreign government. (p.139) It is unlikely that I have ever seen
her approached by foreign agents, offered vast sums of money or
other inducements to betray her government, indeed, unlikely
that I have witnessed any situations offering great temptations.
If I trust her in such a situation, I do not merely stand by her,

acting in ways that support her, either materially or emotionally. I
believe she is innocent. I do not, however, come to believe she is
innocent, despite the evidence, by weighing or balancing present
evidence against her past record. First, by hypothesis, there is
precious little relevant past record. Second, what others regard as
evidence against her isn't considered by me as evidence at all. It
is not that I close my ears to what people say, or refuse to look at,
or repress, the facts. I believe that there is an explanation for the
alleged evidence, for the accusation, which will clear it all up.
In advance of hearing the case, I am prepared to believe that

there is such an explanation. I am biased in favor of my friend, in
favor of her innocence. To put it another way, I am committed to
her being innocent. Moreover, as the case grows, as evidence
mounts, I do not have corresponding mounting doubts. Although
there may come a time when I cease to believe in my friend, there
are no limits which can be set in advance, on epistemic grounds,
which would determine the point at which it is irrational to
continue to trust her. (1987: 3)
The first thing to note about this case is that, as Baker presents it, it is
not clear that this is a case of trust at all. Much of what Baker says here
seems to concern what is involved in being loyal to one's friend.16
According to Baker, the loyal friend will believe in her friend's
innocence despite the evidence.17 Perhaps this is true. But even if it is,
it doesn't have any clear implications for trust. Being a loyal friend may
require believing certain facts about the friend's character, but this
isn't yet to trust the friend with respect to these facts. Still, we can
easily reformulate Baker's case so as to make it a clear issue of trust.
Baker is concerned with the status of my belief that my friend is
innocent, so let's simply add that my friend has told me that she is in
fact innocent and that my belief that she is innocent is based on my
Trusting A Person
trusting her testimony. Now, does what Baker says about my trust in
this case ring true?
Baker asserts that my belief that my friend is innocent is not based on

weighing or balancing present evidence against her past record, but it
is not at all clear that this is the case. Pace Baker, there seems to be a
quite substantive past record of my friend's trustworthiness. She is my
friend after all, and I presumably have a lot of experience with her
competence and sincerity on particular matters. Granted, I may have
never (p.140) seen her approached by foreign agents, but nevertheless
my prior experience with her still gives me a quite substantive basis
upon which to deliberate about whether to trust her in this particular
case. And again, pace Baker, this deliberation will surely take into
account the evidence against her. I needn't take the evidence against
her to be decisive. I may in the end judge that my friend's testimony, a
distinctively interpersonal reason for belief, is stronger evidence than
the evidence against her, and I may therefore judge that the balance of
evidence tells in favor of my friend's innocence. This can even be true
in cases of nontestimonial interpersonal trust. Even in a case in which
the friend has not explicitly told me that she is innocent, the ongoing
interpersonal relationship between us might generate an implicit
second-personal reason for belief that I take to outweigh the evidence
against her. Baker interprets my belief in my friend's innocence as bias,
but this isn't necessarily the case. Of course, I might misjudge the
balance of evidence that I have by taking the implicit second-personal
reason generated by my ongoing relationship with my friend to
outweigh the evidence against her when in fact it does not, and this
might be caused by my emotional attachment to my friend. In such a
case, my belief would be biased and epistemically unjustified—it would
be caused and sustained by considerations that don't genuinely count in
favor of the truth of the belief. However, not all cases of interpersonal
trust in the face of counter evidence need be interpreted in this way. If
the account I have proposed is correct, then our relationships to others
can generate genuinely epistemic reasons for belief, reasons that in
particular cases might outweigh substantial evidence to the contrary.18
Significantly, not only can our relationships to others generate reasons

for belief that might outweigh substantial evidence to the contrary, but
they can also serve in important ways to over-determine our reasons for
belief. Even when an audience has all sorts of nontestimonial reasons
for believing that p, a speaker's testimony can serve to provide an
audience with an additional reason for belief that p, an additional
reason that, in virtue of its second-personal character, has a distinctive
human significance. This might go some distance towards explaining,
Trusting A Person
for example, the significance of victim testimony in the context of such

things as truth and reconciliation committees. Even in cases in which
the crimes perpetrated against a victim or a class of victims is well
documented and beyond reasonable doubt, there can nevertheless be a
point to having victims tell their stories. Even though a victim's
testimony might over-determine our reasons for belief concerning the
crimes perpetrated against her, her testimony nevertheless provides a
particular kind of reason for belief that can be particularly significant in
this context. Trusting the victim for the truth can be one way of
attempting to (p.141) reconstitute and repair the very basic person-to-
person relations that have been transgressed by the crimes in question,
and so the particular kind of reason for belief provided by a victim's
testimony might serve to further the aims of reconciliation even in
cases in which the content of the victim's testimony is already known
and acknowledged.
Construing trust as a second-personal yet cognitive attitude thus has

the resources for explaining a wide variety of phenomena concerning
trust. It fits nicely the linguistic evidence concerning the syntax and
semantics of ordinary trust ascriptions; it has the resources to account
for most of what motivates some philosophers to think that trust is
fundamentally different from belief; and it is in a position to explain the
way in which trust is an attitude aimed at getting things right.
Moreover, the viability of such a view of trust should help to alleviate
the worry that, in appealing to the notion of trust, the interpersonal
account of testimony developed in chapters 2 and 3 is appealing to a
psychological attitude that falls outside the purview of epistemology
proper. If the account provisionally developed here is on the right track,
then interpersonal trust is not simply a matter for ethics.
Nevertheless, one might remain suspicious of this account of trust for

one major reason. I have here proposed that trusting a person to do
something involves believing that the person will do it on the basis of
an at least implicit second-personal reason, a reason that justifies the
belief in virtue of the interpersonal relationship existing between
truster and trusted. If one is attracted to epistemic autonomy, then one
will simply deny this. According to epistemic autonomy, relations of
authority and responsibility between persons cannot play a genuinely
justificatory role in the processes of belief formation and knowledge
acquisition. Epistemic autonomy may thus play a major role in
motivating the thought that trust can have only a nonevidential
influence on belief. In the final chapter of this book, I will return one
last time to the issue of epistemic autonomy. I will argue that if we
accept that interpersonal relations can play a genuinely justificatory
Trusting A Person
role in practical reasoning, as many philosophers do, then we have just

as much reason to think that such relations can play an irreducible role
in reasoning about what to believe.
Notes:
(1) See, for example, Hardin (2002) and (2006). Hardin's encapsulated
interest model of trust is a quite sophisticated attempt to develop a
cognitivist account of trust that captures many of the relevant senses in
which trust is interpersonal. According to the encapsulated interest
model, A trusts B just in case A believes that B's interests “encapsulate”
A's interests, where B's interests encapsulate A's interests just in case it
is in B's interest to act on A's interests qua A's interests or for the
reason that they are A's interests. In explaining the relevant sense of
encapsulation, Hardin says that the reason for A's trusting B must be
that A sees it as being in B's interest to maintain a relationship between
them: “The encapsulated interest account is a rational expectations
account in which the expectations depend on the reasons for believing
that the trusted will fulfill the trust. The typical reason is that the
relationship is ongoing in some sense and that the trusted would like it
to continue” (2006: 31). This brings Hardin's account close to my own.
On the account of trust that I offer in this chapter, A trusts B to ϕ just in
case A believes that B will ϕ on the basis of an at least implicit second-
personal reason that justifies A's belief in virtue of interpersonal
relations of authority and responsibility existing between A and B.
(2) Along these lines, Becker writes, “cognitive accounts of trust appear
to eliminate what they say they describe. That peculiarity, I think, is
what drives (some) theologians, poets, narrative artists, and moral
philosophers toward noncognitive trust” (1996: 47).
(3) See, for example, Baker (1987), Becker (1996), Jones (1996), Holton
(1994), and Hertzberg (1988). I should say that I am sympathetic to the
spirit of noncognitive or affective accounts of trust. I will here argue
that trust is a very distinctive kind of belief based on a very distinctive
kind of reason, and in so doing I hope to capture most of what
motivates noncognitive or affective accounts of trust. However, I will
deny that trust occurs in the absence of or is not fully responsive to
evidential considerations. I think philosophical accounts of trust are led
to deny this only in virtue of employing an overly restrictive conception
of evidence.
(4) I am here indebted to comments from Josef Stern and Chris

Kennedy.
(5) In what follows, I use the symbol * to indicate ungrammaticality.

Trusting A Person
(6) Faulkner (2007a) makes a distinction between “predictive” and

“affective” trust that comes close to capturing the distinction I am
concerned with here, though, as I also argue, I don't think the attitude
of trust ascribed in (1) is best construed as an affective attitude. Trust
is not aptly construed as an attitude that, as Faulkner claims, “need not
be based on evidence and can demonstrate a willful insensitivity to the
evidence” (2007a: 876).
(7) In her discussion of what is involved in believing a person,

Anscombe (1979) remarks that she has sometimes encountered
resistance to the idea that “believing S that p” is a proper grammatical
form. Consider the following example, however. I am in Chicago, and I
am chatting on the phone with a friend who lives in New York. We are
both complaining about how cold it has been in our respective locales,
and we start arguing about whether it has been colder in New York or
in Chicago. In the midst of our argument, I say, “I believe you that it's
cold there, but it's not colder there than it is here.”
(8) Among the personal reactive attitudes, Strawson also lists gratitude
and hurt feelings. Gratitude seems to admit of a similar form in which
the verb takes a personal object, though the personal object must be
proceeded by ‘to,’ as in “I'm grateful to her for saying that.” In this
respect, it seems similar to the grammar associated with pride, as in
“I'm proud of her for passing the test.”
(9) One might think that a sentence like “I love Mary” is akin to a
sentence like “I love eggs.” “I love Mary” thus should not be
understood as elliptical for a sentence stating that Mary is loved for
something. Love would then be a straightforward objectual attitude.
(10) Perhaps there are some instances of sentences in which the verb
‘forgive’ doesn't take a personal object, as in “I forgive her saying that,”
but we certainly don't say things like “I forgive that she said that.”
(11) In English, ascriptions of wanting typically take the grammatical

form of the attitude ascriptions in our left-hand column. We typically
say: “I want him to go to the store.” Nevertheless, it seems that such
want-ascriptions can indeed be straightforwardly analyzed into the
form of our right-hand column: “I want that he will go to the store.” The
intuitive truth conditions of these two want-ascriptions appear to be the
same, and hence philosophers typically characterize the attitude of
wanting in terms of our right-hand column.
Trusting A Person
(12) For a good discussion of the ways in which trust positively requires
belief and some of the intuitions that lead us to think the contrary, see
Hieronymi (2008).
(13) I am here assuming that interpersonal relations can play this role
of generating distinctively second-personal reasons for belief. For
further defense of this claim, see chapter 5.
(14) Hardin (2002 and 2006) glosses over this distinction when he
claims that trust is a matter of knowledge or belief concerning the
trustworthiness of the trusted. Clearly, I can believe that someone is
trustworthy without yet trusting her. If a person tells me that p, I might
already have stronger independent evidence for believing that p. In
such a case, I might believe the person to be trustworthy but not trust
her for the truth. Indeed, the very term “trustworthiness” appears to
refer to those characteristics of a person that make the person worthy
of trust. Believing that a person possess the characteristics that make
her worthy of trust is one thing, actually trusting her is quite another. I
might believe someone trustworthy without ever having the occasion to
trust her.
(15) Hieronymi (2008) argues convincingly that considerations of the

value or usefulness of trusting a person are not properly construed as
reasons for trusting the person. The only considerations that are
genuine reasons for trusting a person are considerations concerning
the person's trustworthiness. I agree. However, I am here adding an
additional layer to the story. The attitude of trusting a person to ϕ itself
involves believing that the person will ϕ, where this belief is justified by
an irreducibly second-personal reason for belief. This is what makes
this trusting belief different from other forms of belief—this is what
makes it the case that this belief doesn't involve the truster's coming to
her own conclusion about things.
(16) This feature of Baker's example is noted by Holton (1994: 73).
(17) Stroud (2006) argues for something similar, though I find her case
for this unconvincing. It seems to me that she overestimates the extent
to which the good friend ought to believe the best about her friends.
(18) Recall the discussion of Hume's argument against belief in the

occurrence of miracles on the basis of testimony in chapter 1. There I
suggested that one might object to Hume's argument by claiming that
Hume simply underestimates the extent to which particularly
Trusting A Person
trustworthy human testimony can outweigh the evidence of prior

experience.
Authority, Autonomy, and Second-Personal Reasons

Benjamin McMyler

Print ISBN-13: 9780199794331
DOI: 10.1093/acprof:oso/9780199794331.001.0001
Authority, Autonomy, and Second-

Personal Reasons
Benjamin McMyler
DOI:10.1093/acprof:oso/9780199794331.003.0006

One might object to the interpersonal account of testimony developed
thus far by arguing that there are no genuinely second-personal
reasons for belief, that the relations of authority and responsibility
between persons appealed to by this account of testimonial belief in
particular and trust-based belief more generally are simply the wrong
kind of thing to play an irreducible role in epistemically justifying
belief. Along these lines, one might accept that second-personal
considerations can play a genuine role in practical rationality but deny
that such relations play an analogous role in theoretical rationality. This
chapter argues that there is just as much reason to think that second-
personal considerations play an irreducible role in theoretical
rationality as there is to think that they play an irreducible role in
practical rationality. The rational significance of the second-person thus
spans whatever divide there may be between theoretical and practical
reason.
Keywords: the second person, second-personal reasons, theoretical rationality, practical

rationality, evidence, authority, autonomy
Philosophers commonly assume that interpersonal relations—relations

of authority and responsibility between persons—cannot play a
genuinely justificatory role in the processes of belief formation and
knowledge acquisition. What other people say and do can certainly give
us reason to believe certain things about them and about the world
around us, just as the behavior of ordinary objects can give us reason to
believe things about these objects and about their environment, but the
interpersonal relations that we bear towards others cannot themselves
generate reasons for belief. These relations might influence our beliefs
in all sorts of ways, but this influence cannot be genuinely justificatory;
it can only result in bias or prejudice. Such bias may be understandable
—it might even be necessary for carrying on our ordinary relationships
with others, necessary for being a good friend, a loving parent, or a
loyal sibling—but it is bias just the same. Insofar as our aim is truth, we
should not take interpersonal relations as a guide.
In chapter 1, I labeled this idea epistemic autonomy, and I argued that

it has a long and distinguished historical pedigree tracing back to the
early Enlightenment. According to epistemic autonomy, fully rational
cognitive agents are always solely epistemically responsible for the
justification of their own beliefs. Another person cannot bear any
genuinely epistemic responsibility for the justification of an individual's
beliefs, and hence relations of authority and responsibility between
persons cannot generate genuine reasons for belief. This isn't to deny
that such interpersonal relations can be conducive to the acquisition of
justified belief in the sense that they might facilitate a cognitive agent's
acquisition of other, noninterpersonal considerations that do in fact
(p.143) appropriately count in favor of belief. It also isn't to deny that
such interpersonal relations can play a significant, nonepistemic role in
the formation and revision of belief. Epistemic autonomy only denies
that interpersonal relations, qua interpersonal relations, generate
genuine reasons for belief.
Epistemic autonomy is an assumption about theoretical reasoning,

about reasoning that issues in belief. Philosophers have also been
attracted to an analogous assumption about practical reasoning, about
reasoning that issues in action. We might call this the idea of practical
autonomy. According to practical autonomy, fully rational practical
agents are always solely responsible for the justification of their own
actions. Another person cannot bear any genuinely rational
responsibility for the justification of an individual's action, and hence
relations of authority and responsibility between persons cannot
generate genuine reasons for action. As Robert Paul Wolff puts it, “The
autonomous man, insofar as he is autonomous, is not subject to the will
of another. He may do what another tells him, but not because he been
told to it. He is, therefore, in the political sense of the word,
free” (1990: 26–27). Autonomous rational agency requires that one
always come to one's own conclusion about what to do, and as a result,
relations of authority and responsibility between persons cannot
themselves generate genuine reasons for action.
Like epistemic autonomy, the idea of practical autonomy has a long and
distinguished historical pedigree tracing back to the early
Enlightenment.1 However, practical autonomy has come in for much
more sustained criticism in he history of post-Enlightenment thought
than has epistemic autonomy. Though the idea of practical autonomy
continues to exert a sustained influence on social and political thought
that extends well beyond the bounds of philosophical theorizing,
criticism of practical autonomy is not uncommon. What is extremely
uncommon, indeed almost nonexistent, is the parallel criticism of
epistemic autonomy.
A case in point is Stephen Darwall's recent book, The Second-Person

Standpoint (2006b). Darwall makes a compelling case for the existence
of what he calls “second-personal reasons for action,” reasons for
action that are grounded in ordinary relations of authority and
responsibility between persons. He thus rejects practical autonomy.2
(p.144) However, in the course of making his case for the existence of
second-personal reasons for action, Darwall argues that there are no

genuinely second-personal reasons for belief. More specifically, he
argues that the most natural candidate for a second-personal reason for
belief, the kind of reason for belief provided by a speaker's testimony, is
not genuinely or irreducibly so. This then makes for a sharp distinction
between theoretical and practical reasoning. While practical reasoning
admits of irreducibly second-personal reasons, theoretical reasoning
does not. In arguing against practical autonomy, Darwall thus endorses
epistemic autonomy. Even though he rejects the idea that fully rational
agents are always solely responsible for the justification of their own
actions, he accepts the idea that fully rational agents are always solely
responsible for the justification of their own beliefs.
I think that epistemic autonomy is just as mistaken as practical

autonomy, and for very much the same reasons. I have argued in this
book for an account of the epistemology of testimony according to
which knowledge and belief based on testimony is justified in virtue of
interpersonal relations of authority and responsibility existing between
a testimonial speaker and audience. In essence, then, I have argued
that testimony does in fact amount to a genuinely second-personal
reason for belief. In this final chapter, I want to use Darwall's own
account of second-personal reasons for action as a guide in making a

parallel case for the existence of genuinely second-personal reasons for
belief. I will argue that testimonial reasons are just as second-personal
as Darwall's leading cases of second-personal reasons for action, and
hence that epistemic autonomy is just as mistaken as practical
autonomy. If this is right, then the rational significance of the second
person actually spans whatever divide there may be between
theoretical and practical reason. Whereas I argued in chapter 4 that
belief justified in virtue of interpersonal relations extends beyond the
cases of testimonial belief to trusting belief generally, I will here show
that the kind of justification relevant to testimonial and trusting belief—
justification in virtue of interpersonal relations of authority and
responsibility—extends into the practical sphere as well. Theoretical
and practical rationality are in this respect continuous.
Section 5.1 gives a very general account of the nature of second-

personal reasons. Section 5.2 makes a prima facie case for the
existence of genuinely second-personal reasons for belief, and sections
5.3 and 5.4 argue that Darwall's reasons for denying the existence of
genuinely second-personal reasons for belief don't stand up to scrutiny.
(p.145) Finally, section 5.5 offers some general conclusions concerning
the role of the second person in human rationality.
5.1 Second and Third-Personal Reasons

What exactly is a second-personal reason? What does it mean to say of a
reason that it is second-personal? In the most general terms, a reason
is a consideration that counts in favor of something. If X is a reason for
Y, then X somehow counts in favor of Y. Depending on our interests,
there are any number of ways in which we can go about individuating
classes or kinds of reasons. One common way is to divide reasons
according to what they are reasons for, according to what it is that the
reasons count in favor of. In this sense, reasons can be divided into
practical and theoretical. A practical reason is a reason that counts in
favor of action, and a theoretical reason is a reason that counts in favor
of belief. Another common way of classifying reasons is according to
how the consideration counts in favor of whatever it is that it counts in
favor of. In this sense reasons can be divided into justificatory and
explanatory. A justificatory reason is a consideration that justifies
something, and an explanatory reason is a consideration that explains
something. Some reasons may be both justificatory and explanatory, but
not all justificatory reasons are explanatory and not all explanatory
reasons are justificatory.
Darwall's distinction between second-personal and third-personal

reasons is a distinction within the class of justificatory reasons. Both
second and third-personal reasons justify things, but they serve to
justify things in very different ways. Understanding the nature of
second-personal reasons thus requires understanding the particular
way in which such reasons justify. According to Darwall, a second-
personal reason is a consideration that justifies something in virtue of
an addresser and an addressee. The justificatory force of a second-
personal reason derives directly from the authority the addresser
assumes in addressing the reason to the addressee, and hence the
reason depends for its existence on the addressee seeing the addresser
as trying to give the reason to the addressee and as having the
authority to do so (Darwall 2006a: 274 and 2006b: 8). Third-personal
reasons can be understood in simple contrast to second-personal
reasons. A third-personal reason is a consideration that justifies
something but not in virtue of interpersonal relations of authority and
responsibility existing between an addresser and an addressee. The
justificatory force of a third-personal reason does not derive from the
authority the addresser assumes in addressing the reason to the
addressee, and hence the reason does not depend for its existence on
the addressee seeing the addresser as trying to give the reason to the
addressee and as having the authority to do so.
(p.146) The easiest way to cotton on to this distinction between second

and third-personal reasons is to consider a concrete example. Darwall
recognizes that both second and third-personal reasons can be
addressed from a speaker to an audience. What distinguishes second
from third-personal reasons is that second-personal reasons actually
derive their justificatory force from the interpersonal relations involved
in the speaker's addressing the reason to the audience.3 This can be
illustrated by considering Thomas Hobbes's distinction between
command and counsel.4 Counseling or advising a person to ϕ justifies
the person's ϕ-ing in virtue of providing the person with the materials
to come to her own conclusion about whether ϕ-ing is the thing to do.
The reason for action provided by counsel or advice thus gains its
justificatory force not from the authority relations involved in the act of
counseling or advising itself but from the particular considerations that
happen to be presented by the counsel or advice. These considerations
exist and tell in favor of ϕ-ing completely independently of the speaker's
presenting them to the audience. In this sense, the speaker is not
assuming any authority over the audience—she is not assuming any
rational responsibility for the audience's decision with regard to
whether to ϕ—and so the kind of reason for action provided by counsel

or advice is third-personal.
Commands are very different. Commanding a person to ϕ does not

involve attempting to provide the person with the materials to come to
her own conclusion about whether to ϕ. A command doesn't present
materials or considerations that are themselves reasons to ϕ. Rather,
the command is itself the reason to ϕ. To obey the command to ϕ is to ϕ
on the basis of the speaker's having addressed one with the command
to ϕ and not on the basis of having come to one's own conclusion that ϕ-
ing is the thing to do.5 One must still come to one's own conclusion
about whether to obey the command, about whether to treat the
speaker as an authority, but this is different from coming to (p.147)
one's own conclusion that ϕ-ing is the thing to do. If one were to ϕ on
the basis of coming to one's own conclusion that ϕ-ing is the thing to
do, then one would not be obeying the command—one would be ϕ-ing
for the wrong kind of reason. In this respect, the kind of reason for
action provided by a speaker's command is genuinely second-personal.
It is a reason that wouldn't exist but for the authority that the speaker
assumes in addressing the reason to the audience.6
The distinction between command and counsel is a distinction within

the realm of practical reasons; it is a distinction between second and
third-personal reasons for action. Our earlier distinction between
second and third-personal reasons made no mention of what these
particular reasons are reasons for, whether action or belief. In this
sense, it amounted to a distinction between what we might call generic
second-personal and generic third-personal reasons. Importantly, much
of what Darwall says about the distinction between second and third-
personal reasons can be construed quite generically, without any
mention of what these reasons are reasons for. Still, this generic
distinction can be easily combined with the distinction between
practical and theoretical reasons to yield a distinction between second
and third-personal reasons for action and second and third-personal
reasons for belief. A second-personal reason for action is a
consideration that justifies an action in virtue of interpersonal relations
of authority and responsibility existing between an addresser and an
addressee, and a third-personal reason for action is a consideration that
justifies an action but not in virtue of such interpersonal relations.
Analogously, a second-personal reason for belief is a consideration that
justifies a belief in virtue of interpersonal relations of authority and
responsibility existing between an addresser and an addressee, and a
third-personal reason for belief is a consideration that justifies a belief

but not in virtue of such interpersonal relations.
Darwall doesn't have all of these distinctions squarely in view. His

particular concern is with reasons for action, and so he isn't much
concerned to distinguish generic second-personal from generic third-
personal reasons. This leads him to orient his account of second-
personal reasons around the practical case and ultimately to collapse
the second-personal/third-personal and practical/theoretical
distinctions. In the end, Darwall seems to think that there is something
essentially practical about second-personal reasons and something
essentially theoretical about third-personal reasons. This, I want to
argue, is a mistake.
(p.148) 5.2 Second-Personal Reasons for Belief

Before we turn to Darwall's reasons for denying the existence of
genuinely second-personal reasons for belief, it will prove helpful to
have a clearer view of what exactly it is that he is denying. As we've
seen, a second-personal reason for belief would have to be a reason for
belief that is consistent with our account of generic second-personal
reasons. It would have to be a consideration that justifies a belief in
virtue of interpersonal relations of authority and responsibility existing
between an addresser and an addressee. The justificatory force of the
reason would have to derive directly from the authority the speaker
assumes in addressing the reason to the audience, and hence it would
have to be the case that the reason for belief would not exist but for the
audience's seeing the speaker as addressing the reason to the audience
and as having the authority to do so.
Darwall notes that if there are genuinely second-personal reasons for

belief, the most natural candidate for such a reason would be the
reason for belief provided by a speaker's testimony. I think that this is
exactly right, and we can see how this is so by noting, as Darwall does
not, that there is an epistemic analogue to the practical distinction
between command and counsel. Indeed, this is just what we should
expect if there is in fact a genuine distinction in the epistemic realm
between second and third-personal reasons. The distinction between
the kind of reason for action provided by counsel and the kind of reason
for action provided by command is very much like the distinction
between the kind of reason for belief provided by arguing and the kind
of reason for belief provided by telling (by testifying). In fact, the
distinction between counsel and command can itself be construed (and
is often construed by Darwall himself) as a distinction between cases in
which a speaker argues for a given course of action and cases in which
a speaker comes out and tells an audience to perform a given course of
action. As we've seen, in counseling or advising a course of action, a
speaker is presenting an audience with considerations that tell in favor
of the course of action that the audience is itself charged with coming
to its own conclusion about. Similarly, in arguing that p a speaker is
presenting the audience with considerations that tell in favor of the
truth of p that the audience is charged with coming to its own
conclusion about. The reason for belief provided by a speaker's
argument thus gains its justificatory force not from the authority
relations involved in the act of arguing but from the particular
considerations that happen to be presented by the argument. These
considerations exist and tell in favor of belief in the conclusion of the
argument completely independently of the speaker's act of presenting
them to the audience. The whole point of providing argument is to help
the audience to see things for itself, to come to believe the conclusion
of the argument on the basis of the audience's own evaluation and
assessment of the strength of the argument, and so the justificatory
force of an argument does not derive (p.149) from any authority the
speaker assumes in presenting the argument to the audience. The kind
of reason for belief provided by argument is therefore third-personal.
Telling is very different. Like commanding a person to ϕ, telling an

audience that p does not involve attempting to provide the audience
with materials to come to her own conclusion about p. Unlike arguing,
telling doesn't present materials or considerations that are reasons to
believe that p. Rather, the telling is itself the reason to believe that p. To
believe that p on the basis of the speaker's telling is to believe that p on
the basis of the speaker's having addressed one with the testimony that
p and not on the basis of having come to one's own conclusion about p.7
One must still come to one's own conclusion about whether to believe
the speaker, about whether to treat the speaker as a theoretical
authority, but this is different from coming to one's own conclusion
about p. If one were to believe that p on the basis of coming to one's
own conclusion about p, perhaps by treating the speaker's testimony as
good inductive evidence for p, then one would not be believing the
speaker—one would be believing that p for the wrong kind of reason. In
this sense, the kind of reason for belief provided by a speaker's
testimony appears to be genuinely second-personal. It appears to be a
reason that would not exist but for the authority the speaker assumes in
addressing the reason to the audience.8
The analogy here between commanding and telling is further supported

by the way in which we are ordinarily entitled to defer challenges to
actions based on commands and to beliefs based on testimony. If I ϕ on
the basis of a speaker's command to ϕ, and if a third-party challenges
my action by arguing that, in the circumstances, ϕ-ing was not the thing
to do, then I am ordinarily entitled to defer the challenge back to the
authority who commanded the action. The authority thus appears to be
partially responsible for the justification of my action. I am still solely
responsible for coming to my own conclusion about whether to obey the
command, and therefore if a third-party challenges my action by
arguing that, in the circumstances, the speaker should not be treated
as an authority, then I am not entitled to defer the challenge. But even
though I am solely responsible for coming to my own conclusion about
whether to treat the speaker as an authority, when I do decide to treat
the speaker as an authority, I am not then solely responsible for the
justification of my action based on the speaker's command; and this is
shown by the fact that I am entitled to defer challenges to this action
back to the authority. This is not the case for action based on advice. If I
proceed to ϕ on the basis of accepting a speaker's advice, and if a third-
party challenges my action by arguing that, in the circumstances, ϕ-ing
was not the thing to do, then I am (p.150) not entitled to defer the
challenge back to the advisor. In ϕ-ing on the basis of a speaker's
advice, I am coming to my own conclusion that ϕ-ing is the thing to do,
and so I am solely responsible for the justification of my action.9
The same seems to hold for the epistemic realm. As we saw in chapter
2, if I believe that p on the basis of a speaker's testimony that p, and if a
third-party challenges my belief by producing evidence that tells
against p, then I am ordinarily entitled to defer the challenge back to
the original testifier. The original testifier thus appears to be partially
responsible for the justification of my belief. I am still solely responsible
for coming to my own conclusion about whether to believe the speaker,
and so if a third-party challenges my testimonial belief that p by
producing evidence that tells against the trustworthiness of the speaker
—say, that the speaker is known to be an inveterate liar—then I am not
entitled to defer the challenge back to the original testifier. But even
though I am solely responsible for coming to my own conclusion about
whether to believe the speaker, when I do decide to believe the speaker
I am not then solely responsible for the justification of my testimonial
belief; and this is shown by the fact that I am entitled to defer
challenges to the content of this belief back to the original testifier. This
is not the case for belief based on argument. If I come to believe that p
on the basis of accepting a speaker's argument, and if a third party
challenges my belief by producing evidence that tells against p, then I
am not entitled to defer the challenge back to the original testifier. In
believing that p on the basis of a speaker's argument, I am coming to
my own conclusion about p, and so I am solely epistemically responsible
for the justification of my belief.10
The kind of reason for belief provided by testimony thus appears to be

quite similar to the kind of reason for action provided by command. In
believing that p on the basis of a speaker's testimony and in ϕ-ing on
the basis of a speaker's command, the speaker is partially responsible
for the justification of the audience's belief or action, and hence the
justificatory force of the testimony or command derives directly from
the responsibility for meeting challenges that the speaker assumes in
addressing the reasons to the audience. Importantly, this is not to
construe testimony as a form of command. Belief cannot be
commanded, and hence there is something paradoxical about a
speaker's (p.151) commanding an audience to believe that p. The
reason for this is that commands aim to direct the will of an audience,
and therefore only things that can be willed can be fit objects of
command. Belief is not (directly) subject to the will, and therefore belief
cannot be (directly) commanded.11 But to say that belief cannot be
commanded is not yet to say that there cannot be reasons for belief that
serve to justify beliefs in a way analogous to that in which commands
serve to justify actions. Telling an audience that p does not involve
commanding an audience to believe that p. Testifying, unlike
commanding, aims to direct belief. Nevertheless, this is completely
consistent with the idea that the way in which testifying aims to direct
belief is similar to the way in which commanding aims to direct the will.
Both telling and commanding aim to direct their respective objects
second-personally.
5.3 Belief, Evidence, and Evidentialism

Darwall admits that testimonial reasons appear to be second-personal,
that a speaker's testimony at least appears to derive its justificatory
force from the relations of authority and responsibility existing between
addresser and addressee. Nevertheless, he goes on to claim that
testimonial reasons are not genuinely or irreducibly second-personal,
that they are not second-personal all the way down (2006b: 57). It is
helpful to put this point in perspective. If one is attracted to what I have
called practical autonomy, then one will naturally be inclined to think
that whatever appears to be a second-personal reason for action must
ultimately bottom out in third-personal reasons, in reasons for action
that a subject is in the position of coming to her own conclusion about.
Darwall rejects this line of thought in the practical domain. He rejects
the idea that what appear to be second-personal reasons for action
ultimately bottom out in third-personal reasons. But when Darwall
turns to the epistemic domain, he accepts a perfectly parallel line of
thought. He accepts the idea that what appear to be second-personal
reasons for belief ultimately bottom out in third-personal reasons. What
we need, then, is some reason for thinking that this line of thought is
actually correct when it comes to the case of reasons for belief—some
reason for thinking that even though practical autonomy is mistaken,
epistemic autonomy is correct. In the remainder of this chapter, I want
to pull out and elaborate on what appear to be Darwall's reasons for
thinking that testimonial reasons for belief must ultimately bottom out
in third-personal reasons.
(p.152) One might naturally associate Darwall's claim about the

reducibility of testimonial reasons for belief with reductionism about
testimony. Importantly, however, Darwall doesn't commit himself to
either reductionism or anti-reductionism about testimony. As these
positions are generally understood, both reductionism and anti-
reductionism about testimony are consistent with the denial of the
possibility of irreducibly second-personal reasons for belief.12 Darwall's
claim about the reducibility of testimonial reasons for belief thus
doesn't have anything to do with the details of the epistemology of
testimony. Instead, his claim appears to be that the reasons for belief
provided by testimony cannot be irreducibly second-personal simply
because they are reasons for belief. Reasons for belief simply can't take
an irreducibly second-personal form.
This suggests that there might be a very general reason for denying the
possibility of genuinely second-personal reasons for belief, something
about the very nature of belief such that reasons for belief cannot take
an irreducibly second-personal form. In line with this, Darwall's
discussions of the reducibility of testimonial reasons typically begin by
noting a very general feature of the concept of belief. Belief, it is often

said, constitutively aims at truth.13 In this sense, belief is different from
other cognitive attitudes like supposing or assuming. Both supposing
and assuming involve representing something as true, but they are not
aimed at truth in the same way as belief. Belief involves not only
representing something as true but doing so with the aim of getting it
right. Belief is an attitude aimed at representing as true only that which
is true (or, alternatively, an attitude governed by a norm stating that the
attitude is correct just in case what it represents as true is true).14
(p.153) It is plausible to think that this feature of the concept of belief

yields a constraint on reasons for belief. If a reason is a reason for
belief, then it must be consistent with belief's constitutive aim; it must
be a reason for representing something as true that is appropriately
aimed at getting it right. Reasons for belief must be, as Darwall puts it,
“grounded in” what is the case believer-neutrally (57). We might call
this the evidentialist constraint on reasons for belief. Some
epistemologists take the way in which belief aims at truth to support
evidentialism, the view that only evidence can amount to a genuine
reason for belief.15 The term “evidence” is here being construed
broadly and precisely in terms of belief's constitutive aim. On this
construal, a consideration is evidence for the proposition p just in case
it counts in favor of the truth of, confirms, or probabilifies p. Employing
this very broad conception of evidence, we can say that, according to
the evidentialist constraint on reasons for belief, a consideration is a
reason for believing that p just in case it is evidence for p.
As here construed, the evidentialist constraint on reasons for belief

rules out the possibility of purely prudential or pragmatic reasons for
belief.16 It rules out the possibility that considerations of the
desirability or efficaciousness of a belief can amount to genuine reasons
for belief. Considerations of the desirability of the belief that p may be
reasons to want to believe that p and hence may be reasons to attempt
to bring about or induce in oneself the belief that p, but they are not
themselves reasons to believe that p. The only considerations that can
be reasons to believe that p are considerations that count in favor of the
truth of p. In this sense, the fact that the cuckolded husband would be
better off believing his wife to be faithful, perhaps because this would
be less painful for him, may be a reason for him to want to believe that
his wife is faithful and hence may be a reason for him to put himself in
positions that might bring about this belief, but it is not a reason for
him to believe that his wife is faithful. The only thing that can be a
reason for believing that his wife is faithful is evidence that his wife is
faithful, a consideration that actually counts in favor of the truth of his

wife's being faithful.
Importantly, however, even if we accept the evidentialist constraint on

reasons for belief, this doesn't go any distance towards ruling out the
possibility of irreducibly second-personal reasons for belief. As here
construed, the evidentialist constraint only rules out the possibility of
purely prudential reasons for belief, and so it would rule out (p.154) the
possibility of genuinely second-personal reasons for belief only if
second-personal reasons for belief were shown to be merely
prudential.17 As far as I know, however, no philosopher has gone so far
as to claim that testimonial reasons are merely prudential. Several
philosophers have sought to reject what, in chapter 3, I labeled the
evidential model of epistemic dependence, but rejecting this model
does not require claiming that testimonial reasons are merely
prudential. Moran, for example, rejects the idea that testimony can be
construed as a kind of ordinary evidence, but in so doing he is working
with a much narrower conception of evidence than that which appears
in the evidentialist constraint. Moran clearly holds that testimony
amounts to a genuinely epistemic, nonprudential reason for belief.
Still, if one is in the grips of epistemic autonomy, one might persist in

thinking that the sense in which testimonial reasons are evidential
simply cannot have anything to do with the second-personal features
involved in the activity of giving and receiving testimony. On this kind
of view, testimony amounts to perfectly good evidence for belief, but not
in virtue of there being anything second-personal about it. A speaker's
testimony serves to justify an audience's belief in the same way as any
other ordinary evidence. A speaker's testimony that p is a consideration
that counts in favor of the truth of p in roughly the same way as a
perceptual experience or another belief.18 It may very well be that the
act of testifying also partakes of various second-personal features
insofar as it involves addressing an audience with a claim, but the
justificatory force of the testimony, that in virtue of which it amounts to
genuine evidence for belief, has nothing to do with these second-
personal features. In this sense, it is not a genuinely second-personal
reason for belief.
Darwall's desire to structure the category of second-personal reasons

around a distinction between the practical and the theoretical leads
him to be attracted to something like this view, though as far as I can
tell, he never explicitly comes out and endorses it. It seems to me,
however, that trying to decouple the sense in which testimonial reasons
(p.155) are evidential from the sense in which they are second-personal
flies in the face of our earlier distinction between arguing and telling.
If, as we've seen, the distinction between arguing and telling is

structurally analogous to Darwall's distinction between advising and
commanding, then the justificatory force of a speaker's testimony
cannot simply be of a piece with any other ordinary kind of evidence. If
a speaker's testimony that p counted in favor of the truth of p in roughly
the same way as a perceptual experience or another belief, then the
way in which a speaker's testimony justifies an audience's belief
wouldn't look relevantly different from the way in which a speaker's
argument justifies an audience's belief. Both perception and inference
appear to involve a subject's coming to her own conclusion about
things. But as we've seen, belief based on testimony is distinguished
from belief based on argument precisely because it doesn't involve a
subject's coming to her own conclusion about things. An audience can
certainly treat a speaker's testimony that p as she would a speaker's
argument or advice, as a consideration with respect to which she must
come to her own conclusion as to whether it counts in favor of the truth
of p. To do so, however, is to believe that p for the wrong kind of reason.
It is not to believe that p on the basis of a testimonial reason, and this is
evident from the fact that, in so believing that p, an audience is not
entitled to defer challenges back to the original testifier.
So not only are testimonial reasons evidential, not only do they

genuinely count in favor of the truth of the proposition believed, but
they do so precisely in virtue of the interpersonal relations of authority
and responsibility existing between addresser and addressee. The
sense in which testimonial reasons are evidential thus cannot be
decoupled from the sense in which they are second-personal. Testimony
amounts to evidence precisely in virtue of its second-personal features,
or so it appears.
5.4 Theoretical Versus Practical Reasons

To his credit, Darwall doesn't go so far as to offer a straightforward
argument from the way in which belief aims at truth to the impossibility
of irreducibly second-personal reasons for belief. In the end, he seems
to be moved by the thought that, despite the way in which testimonial
reasons appear to be second-personal, there are significant
disanalogies between testimonial reasons and second-personal practical
reasons. The existence of these disanalogies might then still give us
good reason to deny that testimonial reasons are genuinely or
irreducibly second-personal.19
Darwall's most sustained discussion of the disanalogy between

theoretical and practical reasons occurs in chapter 11 of The Second-
Person Standpoint. There he offers the (p.156) following example
intended to illustrate the way in which testimonial reasons must

ultimately bottom out in third-personal reasons:
Suppose, first, that you are considering what to believe about the
economy's future direction, say, whether it will recover in the next
quarter. You examine various evidence and determine that it will
not. You then talk to me, who is convinced that the economy will
bounce back . . . In the course of listening to my reasons, you
become persuaded that what I am taking as evidence is
misleading in various respects and that, even if it weren't, it
would nonetheless be overridden by better reasons for thinking
the economy won't recover for at least another two quarters. But I
persist. “Trust me,” I say, “things have to get better soon.” Is it
possible for you to believe that the economy will soon improve
just because I have told you it will? That is, can you believe that it
will get better simply for the reason that I have told you (or, at
least tried to tell you) that it will? You could certainly believe this
if you suspected that, despite the weakness of my stated reasons,
my beliefs might respond to other evidence that I cannot perhaps
articulate . . . But without your supposing some such connection
between my beliefs and facts about the world as they are anyway,
it would simply be impossible for you to believe that the economy
will recover presently just for the reason that I say it will (though
I could of course cause you to believe it). To give me authority in
reasoning about what to believe, you must take me to have some
(epistemic) authority on the questions of fact my beliefs concern.
My claim on your beliefs and your freedom of belief are both
constrained by belief's substantive aim: accurate representation
of the world as it is anyways. (2006b: 288-289)
The example here seems to be one in which the speaker offers the
audience an argument to the conclusion that the economy will turn
around, the audience finds this argument unconvincing, and so as a last
resort the speaker simply tells the audience that the economy will turn
around. Darwall claims that what the speaker says here cannot amount
to a genuine reason for belief due to the fact that the audience does not
judge the speaker to be epistemically competent and authoritative
about the facts. However, it seems that what the speaker says here
cannot amount to a genuine reason for belief simply because the
speaker's speech act amounts to a dogmatic re-assertion of the
conclusion of a failed argument. In this respect, it's noteworthy that
Darwall's example concerns a subject matter that is far more fit for
argument than it is for simple telling. Arguably, predictions about the
future direction of the economy are even more speculative than
predictions about the weather, and so one might question whether such
predictions ever attain the status of knowledge. If testimony is

fundamentally a matter of knowledge-transmission, then there will be
something inherently degenerate about (p.157) outright testimony
(rather than, say, educated guesswork) concerning a subject matter
that is not clearly fit for knowledge. The instance of telling in Darwall's
example is not put forward as an educated guess, and this is part of
what makes it suspect. Someone who thinks that this is the kind of
thing that can be straightforwardly testified to shouldn't be trusted. As
Darwall's example even suggests, the proper home of educated
guesswork is in the context of argument, not testimony.
Despite this problem with the example, however, I think that what
Darwall goes on to conclude here is perfectly correct. In order for a
speaker's testimony that p to amount to a reason for the audience to
believe that p, the audience must judge the speaker to be epistemically
competent or authoritative with respect to p. Taking the speaker's
testimony to be a reason for belief involves (among other things) seeing
the speaker as a theoretical authority, as one standing in a privileged
and authoritative relation to the facts. What I don't see, however, is how
this tells against the possibility of testimony providing irreducibly
second-personal reasons for belief.
Darwall goes on to claim that the fact that testimony can provide a
reason for belief only if the speaker is judged to be competent and
authoritative about the facts makes the case of testimony analogous to
the practical case of advice.
Now, as we have noted before, there is a kind of practical case,

that of advice, that is structurally identical to the theoretical case.
In fact, we might regard it as a special instance of theoretical
reasoning, one in which the beliefs in question concern the
practical question of what one should do. If you ask me for advice
on where to invest your retirement funds during the next quarter,
and you trust me, then any authority you accord me will similarly
depend on what authority you take me to have on an independent
question of fact, namely, what would be a sensible investment in
the current economic circumstances. If you take me to have no
epistemic authority on this question, then it will simply be
impossible for you to treat my advice as reason-giving in the
normal way, that is, to give you any reason to think you should do
something when I say you should. An advisor's claim on your
reasoning, as well as your freedom to treat it as advice, are both
constrained by belief's substantive aim: accurate representation,
in this case, of (non-second-personal) reasons for acting, whose

status as reasons has nothing to do with one person's authority to
give them to another. (2006b: 289)
I have claimed that the epistemic distinction between telling and

arguing is analogous to the practical distinction between command and
counsel. Darwall is here claiming, to the contrary, that testimony is
analogous to advice. However, he does this by claiming that the kind of
reason for action provided by advice amounts to “a special instance of
(p.158) theoretical reasoning.”20 This, I think, is a mistaken
characterization of advice. The distinction between commanding and
advising (or counseling) was originally introduced as a distinction
between second and third-personal reasons for action, but here Darwall
construes advice in such a way as to make it questionable whether
advice amounts to a practical reason at all. Advice is here construed as
the giving of a theoretical reason with a practical content, as the giving
of a reason that counts in favor of beliefs about action rather than
action itself. Darwall's desire to structure the category of second-
personal reasons around the practical case is thus narrowing his
picture of the practical to the point where it looks like advice cannot
amount to the giving of a genuinely practical reason.21 Darwall isn't
alone in this regard. Joseph Raz, for example, has also claimed that
“advice, whatever the hopes of the adviser may be, is given with the
intention that its utterance will be taken as a reason for belief, not for
action” (1979: 21). It seems to me, however, that advising involves
more than informing an audience that certain reasons for action exist.
It involves also intending that the audience take these reasons into
account in her practical reasoning and thereby intending to influence
the will of the audience. In this sense, advising does amount to the
giving of a practical reason, though, unlike commanding, it is the giving
of a practical reason that the audience is charged with coming to its
own conclusion about. The distinction between arguing and telling is
helpful here. Advising is arguing for an audience to ϕ, while
commanding is telling an audience to ϕ.
Darwall's desire to equate testimony with advice thus leads him to

misconstrue advice. Not only does he misconstrue the giving of advice
as the giving of a kind of theoretical reason, but he also misconstrues
the competence required of an advisor as consisting in a kind of
theoretical expertise. The example Darwall offers is one in which, in
order for the audience to see the speaker's advice as genuinely reason-
giving, the audience must judge the speaker to have a relevant degree
of expertise with respect to financial investments. Such theoretical
expertise is sometimes required of an advisor. (p.159) In such cases, the
speaker's advice will take the form of the speaker's testifying to various
facts concerning, say, the relevant investment opportunities and their
associated risks and benefits. On the basis of these facts and facts
concerning the audience and her financial position, the speaker will
then advise a particular course of action: “I think you should ϕ.” Having
listened to the speaker's testimony, the audience must then decide
whether to accept the speaker's advice, which will involve the
audience's assessing both the speaker's testimony and the practical
advice that the speaker has offered based on the facts presented in the
testimony. If the audience accepts the speaker's testimony and accepts
the relevant practical conclusions the speaker has drawn from the facts
presented in the testimony, then the audience will accept the speaker's
advice to ϕ. Clearly, this whole process requires that the audience judge
the speaker to have a relevant degree of theoretical expertise with
respect to financial matters, but this is simply because this instance of
the giving of advice also involves a whole lot of testimony. Not all
advice is like this. Advising a friend that she shouldn't give in to her
parents' demands and should instead pursue the career that she herself
wants to pursue likely will not require testifying to any facts previously
unknown to the audience, and hence it does not seem to require any
specific theoretical expertise. It may still require a kind of practical
wisdom, and so the speaker's advice may here be genuinely reason-
giving only if the audience judges the speaker to have this kind of
practical competence or expertise, but it would be a mistake to
construe the speaker's practical wisdom in this case as involving her
occupying a privileged position with respect to an objective order of
facts.
Again, however, despite these problems, I think that much of what

Darwall goes on to conclude here is correct. It is true, as Darwall
claims, that testimony is like advice in that both testimony and advice
typically generate reasons for an audience only if the audience judges
the speaker to be relevantly competent with respect to her testimony or
advice—though as we've seen, the relevant competence may be quite
different in the theoretical and practical cases. So there is a
circumscribed sense in which testimony is analogous to advice.
However, it is not clear that this analogy between testimony and advice
is at all relevant to their status as producing second-personal reasons.
As we've seen, the reason that advice provides only third-personal
reasons is that, paradigmatically, the giving of advice involves
presenting the audience with considerations that the audience is
charged with coming to its own conclusion about. It is in this sense that
advice is different from command. But testimony is not analogous to
advice in this sense. As we've seen, telling differs from arguing
precisely in that it does not involve presenting the audience with

considerations that the audience is charged with coming to its own
conclusion about. In this sense, even if the kind of reason for belief
generated by a speaker's testimony depends on the audience's judging
the speaker to be competent and authoritative about the facts, the
speaker's testimony still does not call (p.160) for the audience to come
to its own conclusion about the content of the testimony. The speaker's
testimony still calls for the audience to believe the content of the
testimony on the basis of the authority that the speaker assumes in
addressing her testimony to the audience, and so the reason for belief
provided by testimony still looks to be genuinely second-personal. This
means that even if it is true that advice amounts to the giving of a
special kind of theoretical reason, the kind of theoretical reason given
by advice still looks more like the kind of theoretical reason given by
argument than like that given by testimony. Even if advising is
distinguished from commanding in virtue of being theoretical, it is also
distinguished from commanding in virtue of being third-personal, and
testifying does not appear to be distinguished from commanding in this
latter respect.
I have claimed that Darwall points to a genuine though circumscribed

parallel between testimony and advice, but that this parallel is
irrelevant with respect to the question of whether testimony provides
irreducibly second-personal reasons. Perhaps this is mistaken. Perhaps
Darwall's point is that, despite the way in which I originally presented
the distinction between second and third-personal reasons, the
particular parallel between testimony and advice that Darwall has
identified here—the parallel stating that both testimony and advice
generate reasons only if the audience judges the speaker to be
relevantly competent—is enough to show that testimony cannot provide
irreducibly second-personal reasons. If this is the point, then my
previous objection is off-target.
Darwall does seem to be attracted to something like this idea. He goes

on:
Suppose, to vary our familiar example, that it is you who has your
foot on top of mine. If you recognize my claim to your removing
your foot and my authority to make it, you give me second-person
standing in your practical reasoning. You recognize that you
should move your foot because I have the authority to demand
this as, indeed, does anyone on my behalf. The authority you thus
accord me in regulating your conduct by my claim is
fundamentally different from that of an advisor . . . Unlike advice,

it neither depends upon nor can be defeated by the addresser's
epistemic relation to any facts that are themselves independent of
the addresser's second-personal authority. (2006b: 289)
Leaving aside the fact that Darwall again construes the relevant
trustworthiness required of an advisor as overtly epistemic, the idea
here seems to be that the reason-giving force of irreducibly second-
personal reasons does not essentially depend on the audience's judging
the speaker to be relevantly competent. Irreducibly second-personal
reasons derive their justificatory force not from any particular kind of
competence the speaker may have with respect to a theoretical or
practical question, but rather from (p.161) the bare second-personhood
of the speaker, from something like her mere status as one free and
rational agent among others.
As Darwall describes his project: “A major claim of this book is that

second-personal address has certain presuppositions built into it in
general. To enter intelligibly into the second-person stance and make
claims on and demands of one another at all, I argue, you and I must
presuppose that we share a common second-person authority,
competence, and responsibility simply as free and rational
agents” (2006b: 5). This is an interesting and important claim. If I
understand him, Darwall is here pointing to a logical or conceptual
feature of second-personal relations. Entering into second-personal
relations with another requires recognizing the other as a person. It
requires recognizing the other as a being in a position to address
claims to one and to reciprocally recognize one as a being in a position
to address claims to her. These relations of mutual recognition are
essential to the second-person standpoint, and they are also essential to
the generation of second-personal reasons. Second-personal reasons
actually derive their justificatory force from these relations of
recognition, and these relations of recognition do not essentially
depend on the speaker's having any particular kind of competence or
authority with respect to a theoretical or practical question. They do
depend on seeing the other person as having something like bare
second-person authority, but this bare second-person authority simply
goes along with recognizing the other as a person. It doesn't essentially
depend upon and cannot be defeated by considerations of the person's
epistemic or third-personal competence with respect to some particular
subject matter.
I think that Darwall is on to something important here about how the

existence of second-personal reasons shows that the category of other
persons plays a distinctive and irreducible role in human rationality.
What I don't see, however, is how this serves to distinguish testimony

from commands. Even if it is true that second-personal reasons derive
their justificatory force from the bare second-person authority of the
speaker, from the relations of recognition involved in entering into the
second-person standpoint itself, I don't see how this requires denying
the possibility of genuinely second-personal reasons for belief. Darwall
freely admits that having the second-personal practical authority to
address particular commands, for example, often requires having a
certain practical wisdom or expertise. Only the sergeant is in the
position to command her soldiers to perform certain actions, and so her
authority to address such particular irreducibly second-personal
reasons may depend on her having a kind of practical expertise with
respect to which sorts of actions ought to be performed in which sorts
of situations. Nevertheless, Darwall's point is that “the [second-
personal] standing itself neither is, nor simply follows from, any form of
third-personal or epistemic authority” (2006b: 13, my emphasis).
Practical expertise with regard to a particular field of action may be a
necessary background condition for having the authority to address
commands, but the kind of reason for action provided by (p.162) the
command isn't simply a function of, and doesn't simply follow from, this
background condition. Rather, the justificatory force of the command
derives from the second-personal authority the speaker assumes in
actively addressing the reason to the audience, and this second-
personal authority derives from the bare personhood of the speaker.
This is precisely what distinguishes the kind of reason for action
provided by command from the kind of reason for action provided by
counsel.
This all seems equally true of testimony, however. While it is true that
the kind of reason for belief provided by a speaker's telling typically
requires that the audience see the speaker as standing in a position of
authority with respect to the facts, it is nevertheless the case that the
speaker's second-personal standing to address the reason for belief
provided by the telling is not simply a function of and doesn't simply
follow from this. The competence and reliability of the speaker is simply
a background condition that must be in place if the speaker is to be in a
position to herself generate the reason for belief provided by her
coming out and telling the audience that p. The kind of third-personal
competence typically required of a teller thus doesn't appear to play a
relevantly different role from the kind of third-personal competence
often required of a commander. Both function as backgrounds against
which the speaker herself provides the reason-giving force to what she
says.
Significantly, in both the practical and theoretical cases, an audience

can judge that the relevant background is in place without yet
according the speaker any second-personal standing. If a speaker
commands an audience to ϕ, the audience might judge the speaker to
have the relevant practical expertise concerning what to do in these
particular circumstances and so might take the speaker's practical
judgment expressed in her command to be a good reason to ϕ.
However, this can be done without obeying the speaker's command,
without giving over to the speaker any rational responsibility for the
audience's action. If the audience simply treats the speaker's
commands as reliable indicators of what to do, then the audience is not
treating the speaker as a genuine authority, and this is shown by the
fact that, if a third-party challenges the audience by arguing that in the
circumstances ϕ-ing is not the thing to do, the audience is not here
entitled to defer the challenge back to the commander. The audience is
solely responsible for the relevant practical conclusions it draws from
what it takes to be reliable indicators, and so it is not entitled to defer
challenges to these conclusions back to the indicators. Merely reliable
indicators do not stand in second-personal relations, and so the
audience's reason for action here is not genuinely second-personal.
As we saw in chapters 2 and 3, the same seems to go for testimony. If a

speaker tells an audience that p, the audience might judge the audience
to have the relevant theoretical competence or expertise with respect
to this particular subject matter and so might take the speaker's
judgment purportedly expressed in her testimony to be a good reason
to believe that p. However, this can be done without believing the
speaker that p, without (p.163) trusting the speaker for the truth, and
so without giving over to the speaker any epistemic responsibility for
the audience's belief. If the audience simply treats the speaker's
testimony as a reliable indicator of the facts, then the audience is not
treating the speaker as a genuine authority, and this is shown by the
fact that if a third party challenges the audience's belief by producing
evidence that tells against p, the audience is not entitled to defer the
challenge back to the original testifier. The audience is solely
responsible for the relevant theoretical conclusions it draws from what
it takes to be reliable indicators, and so it is not entitled to defer
challenges to these conclusions back to the indicators. Again, merely
reliable indicators do not stand in second-personal relations, and so the
audience's reason for belief here is not genuinely second-personal.
In this sense, the second-personal standing required to generate the

distinctive kind of reason provided by both commands and testimony
neither is, nor simply follows from, any form of third-personal authority.
One can possess the third-personal expertise typically required of a

commander or a testifier without yet being accorded any second-
personal standing and so without generating for an audience genuinely
second-personal reasons for belief. In order to generate the distinctive
kind of reason provided by both commands and tellings, the speaker
must be recognized by the audience as a second person, as one
standing in interpersonal relations of authority and responsibility with
the addressee. Absent such recognition, neither the command nor the
testimony is genuinely accepted and so second-personal reasons,
reasons that distinctively involve a right of deferral, are not generated.
Moreover, it is important to recognize that an audience needn't always

take a speaker to occupy a sophisticated position of theoretical
expertise with respect to the subject matter of her testimony in order
for her testimony to amount to a reason for belief. A speaker might
simply tell her audience that she wants eggs for breakfast or that she is
afraid of spiders. Arguably, in the case of such testimonial avowals, no
sophisticated third-personal authority on the part of the speaker is
required, and hence they come very close to Darwall's purest practical
case of the address of second-personal reasons, the case of my simply
demanding that a person remove her foot from on top of mine. In this
sense, generating the kind of second-personal reason for belief
provided by telling an audience my mind might not require any
particular third-personal competence or expertise whatsoever.
Still, one might think that a significant difference remains. Even in such
“minimal” cases of testimony, we can imagine situations in which we
would not take the speaker's testimony to be good reason to believe
what she says—situations in which the person might actually be
mistaken about what she says due to short-term memory loss or a
failure of emotional self-knowledge, for example. There is a fact of the
matter concerning the speaker's desires or fears that is independent of
the speaker's second-person standing itself, and we can easily imagine
scenarios in which we have good reason to believe that (p.164) the
speaker's testimony does not track these independent facts. If an
audience has reason to believe that the speaker's testimony is mistaken
in one of these ways, then the reason for belief provided by the
speaker's testimony is defeated. In the case of commands, however, it
seems that we can imagine cases in which an audience judges the
speaker's command to be mistaken but where the reason for action
provided by the command is not thereby defeated. In this sense,
Darwall might be motivated to think that only second-personal practical
reasons can derive from the bare second-person authority of the

speaker by imagining the disanalogy between cases like the following:
Case 1: A speaker in a position of practical authority

commands an audience to ϕ. The audience then thinks
to itself, “The speaker is mistaken; ϕ-ing is not the
(appropriate, valuable, desirable) thing to do.
Nevertheless, I still ought to ϕ: the speaker told me to.”
Case 2: A speaker in a position of theoretical authority
tells an audience that p. The audience then thinks to
itself, “The speaker is mistaken; p isn't true.
Nevertheless, I still ought to believe that p: the speaker
told me so.”22
The audience's thought in Case 1 makes sense. The audience's thought

in Case 2, I submit, does not. There thus appears to be a disanalogy
between commanding and telling. Commanding can provide an
undefeated reason for action independently of the perceived
appropriateness, value, or desirability of the action. Telling, on the
other hand, cannot provide an undefeated reason for belief
independently of the perceived truth of the belief.
One might try to deny that there is a genuine disanalogy here by

claiming either that in Case 1 the audience ought not to ϕ or that in
Case 2 the audience ought to believe that p. The latter seems far less
plausible than the former, but even the former seems to me implausible.
It seems to me possible that in Case 1 the speaker does have a reason
to ϕ that is not defeated by the audience's judgment that ϕ-ing is not the
thing to do. For example, a private might judge her sergeant's order to
be mistaken or misguided and still have an undefeated reason to do as
she is ordered. In this sense, I am willing to accept that there is a
genuine disanalogy here. The question is how to explain it.23
(p.165) One might claim that the reason the audience ought to ϕ in
Case 1 but ought not to believe that p in Case 2 is that the speaker
herself can play a more substantive role in generating reasons for
action through her commands than she can reasons for belief through
her testimony. In Case 1, the speaker has the ability to generate a
reason for action all on her own, through something like shear force of
will, while in Case 2 the speaker is unable to simply will that her
testimony amount to a reason for belief. This might then lend support to
the view that the kind of reason for action provided by a speaker's
command is more second-personal than the kind of reason for belief
provided by a speaker's testimony. Since the speaker in Case 1 is able
to generate a second-personal reason for action through sheer force of
will, the kind of reason for action provided by her command derives
from her mere status as a free and rational agent in a way that the kind
of reason for belief provided by testimony does not.
I think this is incorrect. As we've seen, the speaker does play an

essential role in constituting her utterance as a reason for belief
through making it an instance of testimony rather than something less
than this. If the speaker doesn't actively assume the relevant epistemic
responsibilities involved in testifying, then no matter her third-personal
standing with respect to the facts, what she says cannot amount to a
reason for belief that entitles the audience to defer challenges back to
her. In this sense, the will of the speaker is essentially drawn upon in
the generation of testimonial reasons for belief.
What the above disanalogy shows is simply that, as we have already

seen, testimony aims to direct belief while commands do not. Testifying
involves the giving of a particular kind of reason for belief. To tell an
audience that p is to give the audience a particular kind of reason to
believe that p. Accepting a speaker's testimony thus requires believing
what the speaker says, believing that p. If the audience is not in a
position to believe that p, then the audience is not in a position to
accept the speaker's testimony. Commanding, on the other hand,
involves the giving of a particular kind of reason for action. To tell an
audience to ϕ is to give the audience a particular kind of reason to ϕ.
Obeying a speaker's command thus requires doing as one is told, but
importantly it does not require believing that what one is told is the
thing to do. One can thus be in a position to obey a command to ϕ by
proceeding to ϕ on the basis of the command even when one is not in a
position to believe that ϕ-ing is the (appropriate, valuable, desirable)
thing to do.
In this sense, the disanalogy between the above two cases does not
show that commanding is more second-personal than telling. Rather, it
simply shows that commanding and telling are distinguished by their
respective practical and theoretical aims. Since commanding aims to
direct the will, the reason for action provided by command can operate
independently of the audience's beliefs concerning whether the action
is the thing to do. The audience can have good reason to believe that, in
the circumstances, the action is not the thing to do and nevertheless
still have an undefeated reason to act as ordered. Since testifying aims
to direct belief, however, the (p.166) reason for belief provided by
testifying cannot operate independently of the audience's beliefs
concerning the truth of the testimony. The audience cannot have good
reason to believe that what the speaker says is false and still have an
undefeated reason to believe that p. Considerations bearing on the
truth of the belief that p cannot help but bear on the reason for belief
provided by a speaker's testimony that p. But since commands aim to
direct the will rather than belief, considerations bearing on the truth of
the belief that ϕ-ing is the thing to do needn't necessarily bear on the
reason for action provided by the command to ϕ. Of course, such
considerations often will bear on the reason for action provided by a
command. A soldier's judgment that her superior officer's order is
unlawful or unethical may definitely serve to defeat the reason for
action provided by the order. If this is the case, then the soldier ought
not to do as ordered. All that I have claimed is that it is at least possible
that there may be cases in which an audience's belief that the
commanded action is not the thing to do does not serve to defeat the
reason for action provided by the command.
Note that I have not here claimed that actions based on second-
personal practical reasons involve trusting a person. On the account of
interpersonal trust developed in chapter 4, trust is a species of belief—
trusting a person to ϕ involves believing that the person will ϕ on the
basis of an at least implicit second-personal reason that serves to justify
the belief in virtue of the interpersonal relationship existing between
truster and trusted. In this chapter I have argued that actions can be
justified in virtue of the very same kind of interpersonal relationships as
trusting beliefs, but these actions themselves do not involve trusting a
person. On the account I have offered, trust is a form of belief, not
action. Actions may be said to be “based on trust” in various ways
insofar as the actions themselves are supported by trusting beliefs.
Still, a person can have a second-personal reason for action in cases in
which the action is not supported by any trusting beliefs. This is often
the case with commands. If, in the appropriate circumstances, a
sergeant orders a private to fall in, the private has a second-personal
reason for action even if she doesn't trust the sergeant with respect to
any particular matters of fact. The reason is second personal in that it
derives from the relations of authority and responsibility existing
between addresser and addressee, but acting on the basis of the reason
needn't involve trust in that it needn't involve believing anything on the
basis of such interpersonal relations. Moreover, a person can have a
third-personal reason for action that is in fact supported by trusting
beliefs. If, in advising a course of action, a speaker testifies to various
facts, an audience may trust the speaker for the truth with respect to
those facts. Nevertheless, the reason for action provided by the advice
is still third-personal in that it is not itself generated by the
interpersonal relations assumed in the advisor's testimony.24
(p.167) Darwall thinks that there is something essentially practical

about second-personal reasons. I have argued that this is not the case.
The kind of reason for belief provided by a speaker's telling appears to
be genuinely second-personal, and Darwall's reasons for thinking the
contrary are unconvincing. The phenomenon of the second person and
of second-personal reasons thus appears to span whatever divide there
is between practical and theoretical reasoning. This might sound like a
serious indictment of Darwall's project. After all, he places a great deal
of emphasis on drawing a distinction between epistemic and genuinely
second-personal reasons.25 However, viewed from a broader
perspective, what I have argued is that the phenomenon of the second
person has an even greater significance than Darwall himself realizes.
Interpersonal relations of authority and responsibility play a genuinely
justificatory role in both practical and theoretical reasoning, and hence
they are an important and irreducible part of our lives as rational
beings.
Clearly, this shouldn't be overstated. We shouldn't believe just anything

another person tells us, just as we shouldn't do just anything another
person commands us. We are always rationally responsible for
assessing the competence of the speaker and for determining the
relevant extent of her practical or theoretical authority. Still, when we
do judge that, in a particular circumstance, a person is indeed an
authority with respect to what to do or believe, we are then in a
position to acquire a distinctive kind of reason for action or for belief—a
second-personal reason that justifies the action or belief in virtue of
addresser and addressee.
5.5 Conclusion: Authority, Sociality, and Cognition

The idea that the category of other persons plays a distinctive and
irreducible role in practical rationality is common ground for many
moral philosophers, though there is of course much debate about how
exactly this should be understood. The parallel idea that the category of
other persons plays a distinctive and irreducible role in theoretical
(p.168) rationality is far more radical. There is a powerful philosophical
inclination to think that relations of authority and responsibility
between persons simply cannot play a genuinely justificatory role in
cognition. They can certainly play a social role, and they might have
some kind of moral significance, but they can't have an irreducibly
epistemic significance. They can't play a role in actually justifying
knowledge and belief. This is to assume epistemic autonomy; it is to
assume that fully rational cognitive agents are always solely

epistemically responsible for the justification of their own beliefs.
In this book I have argued that epistemic autonomy is mistaken. Fully

rational cognitive agents are not always solely epistemically responsible
for the justification of their own beliefs. In the particular case of
knowledge and belief based on testimony, the speaker is partially
responsible for the justification of the audience's testimonial belief, and
this is shown by the fact that the audience is entitled to defer certain
challenges to its testimonial belief back to the original speaker.
Epistemic responsibility for an audience's testimonial belief is thus
shared out interpersonally between speaker and audience. I have also
argued that this sharing out of epistemic responsibility extends beyond
the cases of outright testimony to the class of trusting belief generally,
and that it closely parallels the way in which responsibility for action
can be shared out interpersonally. If this is right, then relations of
authority and responsibility between persons play an important and
distinctive role in human rationality.
Importantly, rejecting epistemic autonomy doesn't require denying the

value of personal autonomy in the epistemic realm. I have insisted
throughout this book that a mature epistemic agent is always
responsible for coming to her own conclusion about the trustworthiness
of purported theoretical authorities. This general responsibility to
rationally discriminate between authorities is not something that can be
simply ignored or offloaded onto others, and hence we are quick to
criticize others who are overly trusting and easily gulled. In this
respect, a genuinely autonomous epistemic agent is one who, among
other things, is responsibly deferential to authority. This is consistent
with the denial of what I have called epistemic autonomy. One can be
responsibly deferential to theoretical authority and still not be solely
responsible for the justification of one's testimonial beliefs.
In addition, speakers can exert forms of power and influence over the
beliefs of others that fail to amount to exercises of legitimate
theoretical authority, and thus impede the others' autonomy. Forms of
mind control, brainwashing, and indoctrination all seek to cause others
to form beliefs, and in the most insidious cases, they may even pass
themselves off as exercises of legitimate theoretical authority.
Nevertheless, these forms of influence all seek to bypass the agency of
others in a way that the legitimate exercise of theoretical authority
does not. The legitimate exercise of theoretical authority directly
engages the agency of others by providing others (p.169) with a
second-personal reason for belief, a reason for belief that presupposes
the others' second-personal competence to recognize and respond to
the speaker's address. A genuinely autonomous epistemic agent is one

whose beliefs are actually the result of the exercise of her epistemic
agency; but again, this is consistent with the denial of epistemic
autonomy. If what I have argued in this book is correct, then the
exercise of some human agential capacities—Reid's social operations of
mind—is something that necessarily takes two, and hence responsibility
for the outcome of the exercise of these capacities is something that is
shared out interpersonally. Rational responsibility for beliefs based on
testimony and for actions based on commands is shared out between
addresser and addressee in a way that is a direct result of the social
nature of the agential capacities from which they result. Genuinely
autonomous, mature epistemic agency requires that one be in a
position to responsibly exercise one's agency in a way that is free from
forms of external influence that may block or bypass the exercise of
one's rational capacities. Nevertheless, for beings like us, the very
exercise of our rational capacities is often an irreducibly social affair.
Genuinely giving up on epistemic autonomy requires admitting that our

cognitive relation to much of the world around us is at the mercy of the
will of our fellow human beings. It requires admitting that our cognitive
lives are irreducibly dependent on the goodwill of others, and as with
any such dependence, this can seem regrettable. In the end, epistemic
autonomy may ultimately be motivated by an all too human desire to
secure our cognitive relation to the world all on our own, without
dependence on the goodwill of others.26 Nevertheless, I hope that in
this book I have at least begun to (p.170) demonstrate how such
cognitive dependence is an irreducible feature of our lives as social
beings, not something to be lamented or regretted. We might be able to
imagine a world in which rational agents are able to secure their
cognitive relation to the world all on their own, a world in which
rational agents only employ cognitive capacities for which they are
solely epistemically responsible, but such a world simply doesn't appear
to be our own. We are rationally social creatures not only with respect
to practical affairs but with respect to theoretical matters as well, and
the relevant sense of sociality at issue doesn't appear to be something
that can be reduced to our using others as a means to come to our own
conclusion about what to do or believe. Our nature as rational beings is
irreducibly social, and if it were not, then we wouldn't be the kind of
rational beings that we are.
Notes:
(1) For a particularly illuminating Enlightenment defense of practical
autonomy, see Godwin (1971). The idea is also present in the social
psychologist Stanley Milgram's discussion of the implications of his

infamous experiments on obedience to authority in Milgram (1974).
(2) As I have formulated epistemic and practical autonomy, they are

ideas about the nature of rational agency. They claim that an individual
subject is herself solely responsible for some domain of the exercise of
her rational agency. To reject either epistemic or practical autonomy is
not, I think, to endorse some kind of heteronomy, if heteronomy
involves some form of nonrational interference with a subject's rational
agency. See Darwall (2006b: 35). The denial of either epistemic or
practical autonomy is consistent with what most philosophers are
typically concerned with when they are concerned with the nature of
autonomy—namely, an agent's ability to determine her thought and
action according to reasons that she is in a position to endorse, identify
with, or defend. To say that an individual subject is in the position to
rationally determine her own thought and action free from external and
internal nonrational influences is not yet to say that the subject is solely
responsible for the justification of her own thought and action. A
subject may freely determine what to think or do on the basis of
reasons she has even if some of the reasons she has are not reasons
that render her solely responsible for the resulting belief or action.
(3) Importantly, Darwall doesn't hold that second-personal reasons

must be explicitly addressed from an addresser to an addressee.
Second-personal reasons for action can be generated by relations of
authority between persons even where nothing is said from one person
to another. Hence, in one of Darwall's favorite recurring examples, I
can have a second-personal reason to get off of your foot simply in
virtue of my recognizing you as a person. You needn't tell me to get off
of your foot. The fact that I recognize you as a person and thereby as
one standing in interpersonal relations with me is enough to itself
generate a reason for me to get off of your foot, a reason that would not
exist but for our standing in these interpersonal relations.
(4) See Hobbes (1996: 176). This is roughly the distinction in terms of
which Darwall introduces his conception of second-personal reasons in
The Second-Person Standpoint (2006b: 5). In this sense, it seems to
function for Darwall as a paradigmatic case of the distinction between
second and third-personal reasons.
(5) As Hart (1990) puts it, the reason for action provided by a speaker's
command is both content-independent and peremptory. It is content
independent in that the reason derives from the will of the speaker
independently of the specific character of the action to be performed,
and it is peremptory in that it is intended to preclude or cut off the

agent's acting on her own private assessment of the merits of the
action.
(6) The distinction between second and third-personal reasons for

action may often be quite fine. On many occasions there may be little if
any difference between telling a person to ϕ and telling a person she
should ϕ (telling a person that, all things considered, the reasons count
in favor of her ϕ-ing). In this sense, much that goes under the label of
advice, especially between friends and family, is often quite second-
personal. Still, there is a genuine distinction between clearly third-
personal cases of advice and clearly second-personal cases of
command, and this is all that Darwall needs in order to get the
distinction between second and third-personal reasons on the table.
(7) For this reason, Hart claims that testimony provides a reason for
belief that is both content-independent and peremptory (1990: 107).
(8) Friedman argues for a similar parallel between authority over action
and authority over belief, claiming that both actions and beliefs based
on authority involve “the surrender of private judgment” (1990: 63–68).
(9) This is consistent with the fact that, in practice, the giving of advice
often involves testimony—testifying to various premises and then
drawing particular practical conclusions from these premises. In such
cases, an audience is not in the position of coming to its own conclusion
about the speaker's testimony, but it is still in the position of coming to
its own practical conclusion from the testimony offered in the advice.
See section 5.4.
(10) Analogous to the case of advice (see the previous footnote), the
giving of argument often involves testimony—testifying to various
premises and then drawing particular theoretical conclusions from
these premises. In such cases, an audience is not in the position of
coming to its own conclusion about the content of the speaker's
testimony, but it is still in the position of drawing its own theoretical
conclusion from the premises provided by the testimony.
(11) Note that insofar as there is a sense in which belief can be

indirectly willed, there is a corresponding sense in which belief can be
indirectly commanded. Belief can be indirectly willed in the sense that
one can will to perform actions intended to bring about a desired belief,
actions like taking a pill or collecting more evidence. These actions are
fit objects of command, and so there is nothing paradoxical in
commanding an audience to perform them. Belief can thus be indirectly

commanded in the sense that one can command an audience to perform
actions intended to bring about a desired belief.
(12) I argued in chapter 2 that reductionism about testimony requires

denying that testimonial reasons are irreducibly second-personal. If
testimonial reasons for belief are genuinely second-personal, then
reductionism must be mistaken. Most particular anti-reductionist
positions are also committed to denying that testimonial reasons are
genuinely second-personal, though anti-reductionism stated generally
(as the mere denial of reductionism) does not require this.
(13) See Darwall (2006b: 56 and 287). Darwall cites Velleman (2000),
Shah (2003), and Shah and Velleman (2005), but see also Williams
(1973), Moran (2001), and Adler (2002). Much recent discussion of this
issue has concerned the related issue of the transparency of belief—the
phenomenon according to which, from a first-person deliberative
perspective, the question whether to believe that p is transparent to the
question whether p in the sense that one gets oneself in a position to
answer the former, self-directed question by getting oneself in a
position to answer the latter, world-directed question.
(14) Different accounts have been offered of the precise way in which
belief is oriented to truth. On a teleological account of belief, belief is
the attitude of taking something to be true with the aim of getting it
right, with the aim of taking to be true only that which is true. On a
normative account of belief, belief is an attitude of taking something to
be true governed by a norm stating that the attitude is correct just in
case what it takes to be true is in fact true. Velleman (2000) offers a
teleological account of belief. Wedgwood (2002), Boghossian (2003),
and Shah (2003) offer normative accounts. Velleman joins Shah in
offering a normative account in Shah and Velleman (2005). Steglich-
Petersen (2006) argues for a teleological account with a critical
discussion of Shah and Velleman. Nothing I say in the text turns on
which of these accounts is correct.
(15) See, for example, Adler (2002) and Shah (2006). Shah (2006) holds
that evidentialism doesn't follow directly from the transparency of
belief but that, given a plausible constraint on deliberation, the best
explanation of transparency entails it.
(16) Note that the evidentialist constraint is consistent with there being
pragmatic influences on justification. For example, how much evidence
counts as enough evidence in particular circumstances might depend
on how much is practically at stake for the subject. The evidentialist

constraint only denies that such pragmatic considerations amount to
genuine reasons for belief.
(17) One might try to further narrow the conception of evidence as it

appears in the evidentialist constraint in such a way as to actually
exclude more than purely prudential reasons, but this would just raise
the question of whether such a narrower conception of evidence could
actually figure in a constraint on the general category of reasons for
belief. All that plausibly follows from the concept of belief concerning
the necessary nature of reasons for belief is that reasons for belief must
be consistent with belief's constitutive aim—they must be
considerations that genuinely count in favor of the truth of the
proposition believed. If we further narrow our conception of evidence,
this may allow us to exclude second-personal reasons from the category
of evidence, but it will go no distance towards excluding second-
personal reasons from the category of genuine reasons for belief.
(18) As we've seen, much of the dispute between reductionists and anti-
reductionists about testimony can be construed as a dispute concerning
whether the evidential status of testimony should be modeled on that of
perceptual representation (anti-reductionism) or that of inference from
another belief (reductionism). If testimony amounts to an irreducibly
second-personal reason for belief, then both of these models must be
mistaken.
(19) In conversation Darwall has insisted that he is open to the general

possibility of irreducibly second-personal reasons for belief, but that he
is nevertheless impressed by the disanalogies that exist between the
practical and the theoretical cases.
(20) See also Darwall (2006a: 274).
(21) This is a problem for Darwall right from the start, and it is a result
(I think) of his not having squarely in view a fully generic distinction
between second and third-personal reasons. When Darwall first
introduces the notion of second-personal reasons on pp. 5–8 of The
Second-Person Standpoint, he immediately characterizes the giving of
third-personal reasons as epistemic and sets out to contrast genuinely
second-personal reasons with epistemic reasons. I suspect that Darwall
immediately appeals to the practical/epistemic distinction in order to
cash out the second-personal/third-personal distinction because he
doesn't see any other way of drawing the distinction. Note, however,
that the way I have drawn the distinction between commanding and
advising and between telling and arguing does not require appeal to
the practical/epistemic distinction. Instead, it appeals to a generic

distinction between cases in which a subject does and does not come to
her own conclusion about things. A subject can reason to her own
practical conclusions (to actions) or reason to her own theoretical
conclusions (to beliefs), and this then gives us a way of drawing the
second-personal/third-personal distinction without appealing to the
practical/theoretical distinction.
(22) This way of formulating the issue was originally suggested to me

by David Finkelstein. Friedman seems to have this disanalogy in mind
when he claims: “Belief on authority calls for internal assent, whereas
the notion of acting in conformity to the commands of authority allows
for the dissociation of thought and action” (1990: 72).
(23) Note that if one denies that there is a genuine disanalogy here—if
one denies, for example, that the audience in Case 1 has a genuine
reason to ϕ—then this is all the worse for Darwall. Darwall wants to
find a disanalogy between telling and commanding in virtue of which he
can deny that telling amounts to a genuinely second-personal reason
for belief. I am willing to admit that there is a genuine disanalogy here,
but even so, I don't think it supports the idea that commanding is
somehow more second-personal than telling.
(24) Thanks to an anonymous referee for encouraging me to clarify

these points.
(25) A central part of Darwall's project in The Second-Person

Standpoint is to establish that moral obligation and the dignity of
persons are something like transcendental conditions for the possibility
of generating and appreciating second-personal reasons. This is an
extremely ambitious project that I am not in a position to evaluate here.
It is worth noting, however, that adopting a fully generic distinction
between second and third-personal reasons and admitting the existence
of genuinely second-personal reasons for belief might have the effect of
weakening the specifically moral implications of the second-person
standpoint. There might thus be a sense in which Darwall's larger
ambitions with respect to moral theory encourage him to tie the
phenomenon of the second-person to the specifically practical case. I
think that the account of second-personal reasons that I have offered in
this chapter makes sense and is also consistent with the bulk of what
Darwall says about second-personal reasons. My account differs from
Darwall's only in that it doesn't also assert that second-personal reasons
are essentially practical. I have argued that, as of yet, I don't see good
reason for this further claim.
(26) In this respect, the thesis of epistemic autonomy may be related to

something that McDowell has called the interiorization of the space of
reasons: the idea “that reason can ensure that we have only acceptable
standings in the space of reasons, without being indebted to the world
for favors received; if we exercise reason properly, we cannot arrive at
defective standings in the space of reasons, in a way that could only be
explained in terms of the world's unkindness” (1998b: 405). McDowell's
target here is a very general idea about the nature of theoretical
rationality—the idea that the proper exercise of theoretical rationality
doesn't constitutively depend on the way that the world is, that
theoretical rationality is in this sense autonomous. If what I have
argued in this book is correct, then this general interiorization of the
space of reasons can be construed as taking a distinctive form when it
comes to the epistemology of testimony. To give it a label, we might call
this particular form of interiorization the individualization of the space
of reasons: the idea that reason can ensure that we have only
acceptable standings in the space of reasons without being indebted to
others for favors received, that if we exercise reason properly, we
cannot arrive at defective standings in the space of reasons in a way
that can only be explained by the unkindness of others. The space of
reasons is individualized when it is construed as a space in which a
subject is solely epistemically responsible for the justification of her
own beliefs. Importantly, this is very different from what Brandom has
referred to as the individualization of the space of reasons. Brandom
(1995) claims that the problem with the interiorization of the space of
reasons is that interiorizing the space of reasons involves
individualizing it. In this sense, Brandom's claim, unlike mine, is a claim
about the very nature of standings in the space of reasons, that to be a
standing in the space of reasons is to be a certain kind of social
standing. I agree with McDowell (2002) that the particular way in
which Brandom makes out this claim actually succumbs to the
interiorization of the space of reasons. In contrast to Brandom, all that I
have claimed in this book is that the justification of some (though by no
means all) standings in the space of reasons cannot be made out in
terms of what a subject is herself solely epistemically responsible for.
This is a claim about a particular class of standings in the space of
reasons, and the way in which I will make out this claim does not
succumb to the interiorizing move that McDowell attacks.
(p.171) References

Benjamin McMyler

Print ISBN-13: 9780199794331
DOI: 10.1093/acprof:oso/9780199794331.001.0001
(p.171) References
Bibliography references:
Adler, J. 1997. “Lying, Deceiving, or Falsely Implicating.” The Journal of

Philosophy 94: 435–52.
—— . 2002. Belief's Own Ethics (Cambridge, Mass.: The MIT Press).
Anscombe, E. 1979. “What Is It to Believe Someone?” in C. F. Delaney

(ed.), Rationality and Religious Belief (Notre Dame: Notre Dame
University Press).
Audi, R. 1997. “The Place of Testimony in the Fabric of Knowledge and

Justification.” American Philosophical Quarterly 34: 405–22.
Augustine. 1961. Confessions, Pine-Coffin (trans.) (London: Penguin).
Austin, J. 1962. Sense and Sensibilia (Oxford: Oxford University Press).
—— . 1975. How to Do Things With Words (Cambridge, Mass.: Harvard

University Press).
—— . 1979. “Other Minds,” in Philosophical Papers (Oxford: Oxford

University Press).
Baier, A. 1994. “Trust and Antitrust,” in Moral Prejudices (Cambridge,

Mass.: Harvard University Press).
(p.171) References
Bar-On, D. 2004. Speaking My Mind (Oxford: Clarendon Press).
Baker, J. 1987. “Trust and Rationality.” Pacific Philosophical Quarterly

68: 1–13.
Becker, L. 1996. “Trust as Noncognitive Security about Motives.” Ethics

107: 43–61.
Boghossian, P. 2003. “The Normativity of Content.” Philosophical Issues

13: 31–45.
Brandom, R. 1983. “Asserting.” Nous 17: 637–50.
—— . 1994. Making It Explicit (Cambridge, Mass.: Harvard University

Press).
—— . 1995. “Knowledge and the Social Articulation of the Space of

Reasons.” Philosophy and Phenomenological Research 55: 895–908.
Brewer, S. 2006. “Scientific Expert Testimony and Intellectual Due

Process,” in Selinger and Crease (eds.), The Philosophy of Expertise
(New York: Columbia University Press).
Burge, T. 1979. “Individualism and the Mental.” Midwest Studies in

—— . 1993. “Content Preservation.” The Philosophical Review 102: 457–

88.
—— . 1997. “Interlocution, Perception, and Knowledge.” Philosophical

Studies 86: 21–47.
Burnyeat, M. 1980. “Socrates and the Jury: Paradoxes in Plato's

Distinction Between Knowledge and True Belief.” Aristotelian Society
Supplement 54: 173–91.
—— . 1990 “Introduction,” in The Theaetetus of Plato (Indianapolis:

Hacket Publishing).
Cavell, S. 1976. Must we mean what we say? (Cambridge: Cambridge

University Press).
—— . 1979. The Claim of Reason (Oxford: Oxford University Press).
Coady, C. 1992. Testimony: A Philosophical Study (Oxford: Oxford

University Press).
(p.171) References
—— . 2004. “Reid and the Social Operations of Mind,” in Cuneo and Van
Woudenberg (eds.), The Cambridge Companion to Thomas Reid
(Cambridge: Cambridge University Press).
Craig, E. 1990. Knowledge and the State of Nature (Oxford: Oxford

University Press).
Daston, L. 1988. Classical Probability in the Enlightenment (Princeton,

N.J.: Princeton University Press).
Darwall, S. 2006a. “The Value of Autonomy and Autonomy of the Will.”

Ethics 116: 263–84.
—— . 2006b. The Second Person Standpoint (Cambridge, Mass.:

Harvard University Press).
(p.172) Descartes, R. 1985. The Philosophical Writings of Descartes,

Vol. I, Cottingham, Stoothoff, and Murdoch (trans.) (Cambridge:
Cambridge University Press).
Faulkner, P. 2000. “The Social Character of Testimonial Knowledge.”

The Journal of Philosophy 97: 581–601.
—— . 2007a. “On Telling and Trusting.” Mind 116: 875–902.
—— . 2007b. “What Is Wrong with Lying?” Philosophy and

Phenomenological Research 75: 535–57.
Finkelstein, D. 2003. Expression and the Inner (Cambridge, Mass.:

Harvard University Press).
Fogelin, R. 2003. A Defense of Hume on Miracles (Princeton, N.J.:

Princeton University Press).
Fricker, E. 1987. “The Epistemology of Testimony.” Proceedings of the

Aristotelian Society Supplement 61: 57–83.
—— . 1994. “Against Gullibility,” in Matilal and Chakrabarti (eds.),

Knowing From Words (Dordrecht, NL: Klewer).
—— . 1995. “Telling and Trusting: Reductionism and Anti-Reductionism

in the Epistemology of Testimony.” Mind 104: 393–411.
—— . 2006a. “Testimony and Epistemic Autonomy,” in Lackey and Sosa

(eds.), The Epistemology of Testimony (Oxford: Oxford University
Press).
(p.171) References
—— . 2006b. “Second-Hand Knowledge.” Philosophy and

Friedman, R. 1990. “On the Concept of Authority in Political

Philosophy” in Raz (ed.), Authority (New York: New York University
Press).
Geach, P. 1977. The Virtues (Cambridge: Cambridge University Press).
Gettier, E. 1963. “Is Justified True Belief Knowledge.” Analysis 23: 121–
23.
Godwin, W. 1971. Enquiry Concerning Political Justice (Oxford: Oxford

University Press).
Goldberg, S. 2006. “Reductionism and the Distinctiveness of

Testimonial Knowledge,” in Lackey and Sosa (eds.), The Epistemology
of Testimony (Oxford: Oxford University Press).
—— . 2007. Anti-Individualism: Mind and Language, Knowledge and

Justification (Cambridge: Cambridge University Press).
—— . 2009. “The Social Virtues: Two Accounts.” Acta Analytica 24: 237–
48.
—— . 2010. Relying on Others: An Essay in Epistemology (New York:

Oxford University Press).
Graham, P. 2006. “Testimonial Justification: Inferential or Non-

Inferential?” The Philosophical Quarterly 56: 84–95.
Greco, J. 2007. “The Nature of Ability and the Purpose of Knowledge.”

Philosophical Issues 17: 57–69.
Grice, P. 1989. Studies in the Way of Words (Cambridge, Mass.: Harvard

University Press).
Hacking, I. 1975. The Emergence of Probability (Cambridge: Cambridge

University Press).
Hardin, R. 2002. Trust and Trustworthiness (New York: Russell Sage

Foundation).
—— . 2006. Trust (Cambridge: Polity Press).
Hardwig, J. 1985. “Epistemic Dependence.” Journal of Philosophy 82:

335–49.
(p.171) References
—— . 1991. “The Role of Trust in Knowledge.” Journal of Philosophy 88:

693–704.
Hart, H. 1990. “Commands and Authoritative Legal Reasons,” in Raz

(ed.), Authority (New York: New York University Press).
Hertzberg, L. 1994. “On the Attitude of Trust.” Inquiry 31: 307–22.
Hieronymi, P. 2008. “The Reasons of Trust.” Australasian Journal of

Hinchman, E. 2005. “Telling as Inviting to Trust.” Philosophy and

Hobbes, T. 1996. Leviathan, Tuck (ed.) (Cambridge: Cambridge

University Press).
(p.173) Holton, R. 1994. “Deciding to Trust, Coming to Believe.”

Australasian Journal of Philosophy 72: 63–76.
Hume, D. 1999. An Enquiry Concerning Human Understanding, Selb-

Bigge (ed.) (Oxford: Oxford University Press).
Jones, K. 1996. “Trust as an Affective Attitude.” Ethics 107: 4–25.
Kant, I. 1983. “An Answer to the Question: What is Enlightenment?” in

Humphrey (trans.), Perpetual Peace and Other Essays (Indianapolis:
Hackett Publishing).
Keren, A. 2007. “Epistemic Authority, Testimony and the Transmission

of Knowledge.” Episteme 4: 368–81.
Kusch, M. 2002. Knowledge by Agreement (Oxford: Oxford University

Press).
Lackey, J. 1999. “Testimonial Knowledge and Transmission.” The

Philosophical Quarterly 49: 471–90.
—— . 2006a. “The Nature of Testimony.” Pacific Philosophical Quarterly

87: 177–79.
—— . 2006b. “It Takes Two to Tango: Reductionism and Non-

Reductionism in the Epistemology of Testimony,” in Lackey and Sosa
(eds.), The Epistemology of Testimony (Oxford: Oxford University
Press).
—— . 2006c. “Knowing from Words.” Philosophy and Phenomenological

Research 73: 77–101.
(p.171) References
—— . 2008. Learning from Words: Testimony as a Source of Knowledge

(Oxford: Oxford University Press).
Locke, J. 1975. An Essay Concerning Human Understanding, Nidditch

(ed.) (Oxford: Clarendon Press).
—— . 1996. Some Thoughts Concerning Education and Of the Conduct

of the Understanding, Grant and Tarcov (ed.) (Cambridge: Hackett
Publishing).
Lyons, J. 1997. “Testimony, Induction, and Folk Psychology.”

Australasian Journal of Philosophy 75: 163–77.
McDowell, J. 1998a. “Knowledge by Hearsay,” in Meaning, Knowledge,

and Reality (Cambridge, Mass.: Harvard University Press).
—— . 1998b. “Knowledge and the Internal,” in Meaning, Knowledge,

and Reality (Cambridge, Mass.: Harvard University Press).
—— . 2002. “Knowledge and the Internal Revisited.” Philosophy and

McMyler, B. 2007. “Knowing at Second Hand.” Inquiry 50: 511–40.
—— . 2011. “Believing What the Man Says About His Own Feelings,” in
Gustafsson and Sørli (eds.), The Philosophy of J. L. Austin: New Essays
—— . forthcoming. “The Epistemic Significance of Address.” Synthese.
Milgram, S. 1974. Obedience to Authority (New York: Perennial

Classics).
Moran, R. 2001. Authority and Estrangement (Princeton, N.J.: Princeton

University Press).
—— . 2004. “Problems of Sincerity.” Proceedings of the Aristotelian

Society 105: 341–61.
—— . 2005. “Getting Told and Being Believed.” Philosopher's Imprint 5:

1–29.
Orwell, G. 1949. 1984 (New York: Harcourt Brace Jovanovich, Inc).
Owens, D. 2000. Reason Without Freedom (New York: Routledge).
—— . 2006. “Testimony and Assertion.” Philosophical Studies 130: 105–

29.
(p.171) References
Plato 1981. Five Dialogues: Euthyphro, Apology, Crito, Meno, Phaedo,

Grube (trans.) (Indianapolis: Hackett Publishing).
—— . 1990. Theaetetus, Levett (tr.) (Indianapolis: Hackett Publishing).
—— . 2004. Republic, Reeve (tr.) (Indianapolis: Hackett Publishing).
Pritchard, D. 2006. What is this Thing Called Knowledge? (New York:

Routledge).
(p.174) Proust, M. 1999. In Search of Lost Time: Vol. V, The Captive &
The Fugitive (New York: Modern Library).
Putnam, H. 1975. “The Meaning of ‘Meaning.’ ” Language, Mind, and

Knowledge, Minnesota Studies in the Philosophy of Science 7: 131–93.
Raz, J. 1979. “Legitimate Authority,” in The Authority of Law: Essays on

Law and Morality (Oxford: Clarendon Press).
Reid, T. 1997. An Inquiry into the Human Mind (University Park: Penn
State University Press).
—— . 2002. Essays on the Intellectual Powers of Man (University Park:

Penn State University Press).
Ross, A. 1986. “Why Do We Believe What We Are Told?” Ratio 27: 69–
88.
Schmitt, F. 2006. “Testimonial Justification and Transindividual

Reasons” in Lackey and Sosa (eds), The Epistemology of Testimony
Searle, J. 1969. Speech Acts (Cambridge: Cambridge University Press).
—— . 1979. “A Taxonomy of Illocutionary Acts,” in Expression and

Meaning (Cambridge: Cambridge University Press).
Shah, N. 2003. “How Truth Governs Belief.” Philosophical Review 112:

447–82.
—— and Velleman, D. 2005. “Doxastic Deliberation.” Philosophical

Review 114: 497–534.
—— . 2006. “A New Argument for Evidentialism.” The Philosophical

Quarterly 56: 481–98.
Shapin, S. 1994. A Social History of Truth: Civility and Science in

Seventeenth-Century England (Chicago: University of Chicago Press).
(p.171) References
Shogenji, T. 2006. “A Defense of Reductionism about Testimonial

Justification of Beliefs.” Nous 40: 331–46.
Snyder, J. 1993. “What Happens by Itself in Photography?” in Cohen et

al. (eds.), Pursuits of Reason: Essays in Honor of Stanley Cavell
(Lubbock: Texas Tech University Press).
Sosa, E. 2007. A Virtue Epistemology: Apt Belief and Reflective

Knowledge, Volume I (Oxford: Oxford University Press).
Sperber, D. and Wilson, D. 1991. “Inference and Implicature,” in Davis

(ed.), Pragmatics: A Reader (Oxford: Oxford University Press).
—— . 1995. Relevance (Oxford: Blackwell Publishing).
Steglich-Petersen, A. 2006. “No Norm Needed: On the Aim of Belief.”

The Philosophical Quarterly 56: 499–516.
Stevenson, L. 1993. “Why Believe What People Say?” Synthese 94: 429–
51.
Strawson, P. 1974. “Freedom and Resentment,” in Freedom and

Resentment and Other Essays (London: Methuen & Co. Ltd.).
Stroud, S. 2006. “Epistemic Partiality in Friendship.” Ethics 116: 498–

524.
Thompson, M. 2004. “What is it to Wrong Someone? A Puzzle about

Justice” in Wallace et al. (eds.), Reason and Value (Oxford: Oxford
University Press).
Van Cleve, J. 2006. “Reid on the Credit of Human Testimony,” in Sosa

and Lackey (eds.), The Epistemology of Testimony (Oxford: Oxford
University Press).
Velleman, D. 2000. The Possibility of Practical Reason (Oxford: Oxford

University Press).
Walton, K. 1984. “Transparent Pictures: On the Nature of Photographic

Realism.” Critical Inquiry 11: 246–77.
Wedgwood, R. 2002. “The Aim of Belief.” Philosophical Perspectives 16:

267–97.
Weiner, M. 2003. “Accepting Testimony.” The Philosophical Quarterly

53: 256–64.
(p.171) References
Welbourne, M. 1993. The Community of Knowledge (Hampshire: Gregg

Revivals).
Williams, B. 1973. “Deciding to Believe,” in Problems of the Self

(Cambridge: Cambridge University Press).
(p.175) Wittgenstein, L. 1972. On Certainty (New York: Harper and

Row).
—— . 1997. Philosophical Investigations (Oxford: Blackwell).
Wolff, R. 1990. “The Conflict Between Authority and Autonomy,” in Raz

(ed.), Authority (New York: New York University Press).
Wolterstorff, N. 1996. John Locke and the Ethics of Belief (Cambridge:

Cambridge University Press).
Zagzebski, L. 2007. “Ethical and Epistemic Egoism and the Ideal of

Autonomy.” Episteme 4: 252–63.
—— . manuscript. Epistemic Authority: An Investigation of Trust,

Authority, and Autonomy in Belief. (p.176)

Testimony, Trust, & Authority

Uploaded by

Copyright:

Available Formats

Testimony, Trust, & Authority

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

Testimony, Trust, & Authority

Uploaded by

Copyright:

Available Formats

What is the main topic of the book?

What is the main topic of the book?

What philosophical tradition is the author interested in exploring?

What philosophical tradition is the author interested in exploring?

Title Pages

University Press Scholarship Online

Oxford Scholarship Online

Testimony, Trust, and Authority

Print publication date: 2011

(p.iii) Testimony, Trust, and Authority

(p.1) Testimony, Trust, and Authority (p.2)

Oxford University Press, Inc., publishes works that further

Copyright © 2011 Oxford University Press

Published by Oxford University Press, Inc.

Oxford is a registered trademark of Oxford University

All rights reserved. No part of this publication may be

Library of Congress Cataloging-in-Publication Data

(Cavell 1979: 391)

In the narrowest terms, this book is about the epistemology of

The project of this book, however, is to demonstrate that the topic of

This is an extremely unorthodox philosophical thesis, but it is somewhat

majority of epistemologists, however, such terms are (and indeed must

I have tried in this book to focus as much as possible on the actual

In a sense, then, I aim here to rehabilitate the intuitive idea that

rationality, that fully rational cognitive agents are epistemically

The standard contemporary theories concerning the epistemology of

In chapter 2, I provide an alternative, essentially interpersonal account

belief by producing evidence that tells against p, then the audience is

Importantly, to reject epistemic autonomy is not to endorse gullibility or

Unfortunately, giving up on epistemic autonomy might not seem like a

In chapters 4 and 5, I consider and respond to two very general reasons

Finally, in chapter 5, I examine the notion of a second-personal reason

reason to think that the rational significance of the second-person

The upshot of this book is that human rationality is constitutively

(2) Notable exceptions include Faulkner (2007a) and (2007b), Keren

Testimony as a Philosophical Problem

Abstract and Keywords

Keywords: knowledge, understanding, authority, testimony, faith, autonomy, evidence,

Napoleon lost the Battle of Waterloo. I know that, as I write, the

All of these things I know on the basis of what epistemologists call

As epistemologists use the term, ‘testimony’ refers to something much

Testimony is without a doubt one very common source of knowledge,

More strictly, perception, memory, and inference are three cognitive

There is one important respect, however, in which the capacity for

Enlightenment philosopher Thomas Reid refers to as “a social operation

Of course, this assumes that there is an intelligible sense in which such

In many respects, the epistemological problem of testimony is a

Enlightenment is man's emergence from his self-imposed

In his groundbreaking book on testimony, C. A. J. Coady argues that a

In the post-Renaissance Western world the dominance of an

obscured the extent to which much real-world rational inquiry requires

Section 1.1 argues that, prior to the Enlightenment, beliefs based on

reductionism, a dispute that sets the agenda for contemporary debates

1.1 Testimony, Knowledge, and Understanding

How's that? What art do you mean?

The art of the greatest representatives of wisdom—the men called

And by ‘persuading them’, you mean ‘causing them to judge’, don't

As Myles Burnyeat has noted, this passage poses several interpretive

Second, it is unclear why exactly the lawyer is so limited. On the one

Socrates goes on to speak of a juror forming a judgment with respect to