You are on page 1of 931


Second Edition

Edited by Ernest Sosa, Jaegwon Kim,

Jeremy Fantl, and Matthew McGrath

~ A Blackwell
'-II Publishing
Editorial material and organization 2000, 2008 by Blackwell Publishing Ltd

350 Main Street, Malden, MA 02148-5020, USA
9600 Garsington Road, Oxford OX4 2DQ, UK
550 Swanston Street, Carlton, Victoria 3053, Australia

The right of Ernest Sosa, Jaegwon Kim, Jeremy Fant!, and Matthew McGrath to be identified as the Authors of
the Editorial Material in this Work has been asserted in accordance with the UK Copyright, Designs, and
Patents Act 1988.

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, except as permitted
by the UK Copyright, Designs, and Patents Act 1988, without the prior permission of the publisher.

Designations used by companies to distinguish their products are often claimed as trademarks. All brand
names and product names used in this book are trade names, service marks, trademarks, or registered trade-
marks of their respective owners. The publisher is not associated with any product or vendor mentioned in
this book.

This publication is designed to provide accurate and authoritative information in regard to the subject matter
covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If
professional advice or other expert assistance is required, the services of a competent professional should be

First edition published 2000 by Blackwell Publishers Ltd

This edition published 2008 by Blackwell Publishing Ltd


Library of Congress Cataloging-in-Publication Data

Epistemology: an anthology / edited by Ernest Sosa, Jaegwon Kim, Jeremy Fant!, and Matthew McGrath - 2nd ed.
p. em. - (Blackwell philosophy anthologies)
Includes bibliographical references.
ISBN 978-1-4051-6967-7 (hardcover: alk. paper) - ISBN 978-1-4051-6966-0 (pbk. : alk. paper) l.
Knowledge, Theory of. I. So sa, Ernest.


A catalogue record for this title is available from the British Library.

Set in 9.511l.5pt Minion

by SPi Publisher Services, Pondicherry, India
Printed and bound in Singapore
by Markono Print Media Pte Ltd

The publisher's policy is to use permanent paper from mills that operate a sustainable forestry policy, and
which has been manufactured from pulp processed using acid-free and elementary chlorine-free practices.
Furthermore, the publisher ensures that the text paper and cover board used have met acceptable environmental
accreditation standards.

For further information on

Blackwell Publishing, visit our website:

Preface to the Second Edition ix

Acknowledgments xi

Part I Skepticism 1
Introduction 3
The Problem of the External World 7
Barry Stroud
2 Proof of an External World 26
G. E. Moore
3 Four Forms of Scepticism 29
G. E. Moore
4 Certainty 31
G. E. Moore
5 How a Pyrrhonian Skeptic Might Respond to Academic Skepticism 35
Peter Klein
6 Epistemological Realism 51
Michael Williams

Part II The Structure of Knowledge and Justification 73

Introduction 75

7 The Myth of the Given 80

Roderick M. Chisholm
8 Does Empirical Knowledge Have a Foundation? 94
Wilfrid Sellars
9 Epistemic Principles 99
Wilfrid Sellars
10 Can Empirical Knowledge Have a Foundation? 109
Laurence BonJour

11 A Coherence Theory of Truth and Knowledge 124

Donald Davidson

12 A Foundherentist Theory of Empirical Justification l34

Susan Haack

l3 The Raft and the Pyramid 145

Ernest Sosa

14 Human Knowledge and the Infinite Regress of Reasons 165

Peter D. Klein

Part III Defining Knowledge 187

Introduction 189
15 Is Justified True Belief Knowledge? 192
Edmund Gettier

16 Thought, Selections 194

Gilbert Harman

17 The Inescapability of Gettier Problems 207

Linda Zagzebski

18 A State of Mind 213

Timothy Williamson

Part IV Epistemic Closure 231

Introduction 233

19 Epistemic Operators 237

Fred Dretske

20 Skepticism, Relevant Alternatives, and Deductive Closure 247

Gail Stine

21 Knowledge and Skepticism 255

Robert Nozick

22 How to Defeat Opposition to Moore 280

Ernest Sosa

23 Are There Counterexamples to the Closure Principle? 290

Jonathan Vogel

Part V Theories of Epistemic Justification 303

Introduction 305
24 Evidentialism 310
Richard Feldman and Earl Conee

25 Skepticism and Rationality 322

Richard Foley

26 What Is Justified Belief? 333

Alvin I. Goldman

27 Reliabilism Leveled 348

Jonathan Vogel
28 Externalist Theories of Empirical Knowledge 363
Laurence BonJour
29 Internalism Exposed 379
Alvin I. Goldman
30 Externalism and Skepticism 394
Richard Fumerton
31 Internalism Defended 407
Richard Feldman and Earl Conee

Part VI Virtue Epistemology and the Value of Knowledge 423

Introduction 425
32 Warrant: A First Approximation 429
Alvin Plantinga
33 Virtues of the Mind, Selections 442
Linda Zagzebski
34 Virtues and Vices of Virtue Epistemology 454
John Greco
35 Cognitive Responsibility and the Epistemic Virtues 462
Duncan Pritchard
36 The Place of Truth in Epistemology 477
Ernest Sosa
37 Why Should Inquiring Minds Want to Know?: Meno Problems and Epistemological Axiology 492
Jonathan L. Kvanvig
38 True Enough 507
Catherine Z. Elgin

Part VII Naturalized Epistemology and the A Priori 521

Introduction 523
39 Epistemology Naturalized 528
W. V. Quine
40 What Is "Naturalized Epistemology"? 538
Jaegwon Kim
41 Quine as Feminist: The Radical Import of Naturalized Epistemology 552
Louise M. Antony
42 There is at Least One A Priori Truth 585
Hilary Putnam
43 Revisability, Reliabilism, and A Priori Knowledge 595
Albert Casullo
44 A Priori Knowledge and the Scope of Philosophy 612
George Bealer

45 Normativity and Epistemic Intuitions 625

Jonathan M. Weinberg, Shaun Nichols, and Stephen Stich

46 Investigating Knowledge Itself 647

Hilary Kornblith

Part VIII Knowledge and Context 661

Introduction 663
47 Solving the Skeptical Problem 669
Keith DeRose

48 Elusive Knowledge 691

David Lewis

49 Contextualist Solutions to Epistemological Problems: Scepticism, Gettier, and the Lottery 706
Stewart Cohen

50 Knowledge and Practical Interest, Selections 721

Jason Stanley

51 Evidence, Pragmatics, and Justification 742

Jeremy Fantl and Matthew McGrath

52 Sensitive Moderate Invariantism 760

John Hawthorne

53 The Assessment Sensitivity of Knowledge Attributions 779

John MacFarlane

PartIX Testimony, Memory, and Perception 801

Introduction 803
54 Trust and Rationality 807
Judith Baker

55 Against Gullibility 815

Elizabeth Fricker

56 Content Preservation 836

Tyler Burge

57 Testimonial Knowledge and Transmission 855

Jennifer Lackey

58 The Problem of Memory Knowledge 868

Michael Huemer

59 Criteria, Defeasibility, and Knowledge 876

John McDowell

60 Knowing How to Believe With Justification 892

Steven L. Reynolds

Index 903
Preface to the Second Edition

Epistemology is a philosophical inquiry into the The selections are collected in nine sections,
nature, conditions, and extent of human knowl- each of which opens with an introduction that
edge. It encompasses some of the most puzzling discusses the contained readings, and is fol-
and persistent issues in all of philosophy, ones lowed by a list of further readings on the subject
that extensively define its history. The problem of matter of that section. For further expert but
skepticism is one example, and the empiricism/ introductory discussion of the issues, the reader
rationalism controversy another, along with its is referred to the relevant Blackwell Companion
Kantian and Hegelian aftermath. Such issues, and Guide.
although alien to common sense at first sight, in The topics taken up in these nine sections by
fact derive naturally from straightforward reflection no means exhaust the field of epistemology. Space
on the most ordinary knowledge about the world limits have made it impossible to include all topics
around us, knowledge produced or sustained in the field. We have consciously selected central
through perception, memory, or induction. issues but we have also drawn from contempo-
Elementary reflection on such matters produces rary work some of the most novel and radical
puzzles and paradoxes that have engaged philoso- responses to those issues. The resulting collection
phers from ancient times to the present. brings together a variety of approaches and solu-
This anthology is meant to supplement tions, still under vigorous debate. The current
Blackwell's Companion to Epistemology and Guide edition departs from the first in expanding the
to Epistemology. We made a conscious effort to section on epistemological contextualism to take
include both selections that are representative of account of recent work on sensitive invariantism
the best current discussion on the most central and relativism. We have also added a section
issues in the field, and selections which, though devoted to perception, memory, and testimony,
relevant to current debates, are somewhat older significantly restructured and reorganized other
and appropriate for use in upper level undergrad- sections, and included some newer work. Space
uate epistemology courses. Though the former limitations have prevented us, once again, from
selections are inevitably demanding, all readings, including work on more specific issues - other
some of which are only excerpts, should all prove minds and induction, for example. On these
accessible in proper order to the attentive reader issues excellent work has been published and
who approaches these issues for the first time. continues to appear.

Special thanks go to Sara Bagg and Justin Nick Bellorini, our editor, has been exceedingly
McBrayer for their assistance, particularly in their patient and supportive.
contributions to the introductions to the various
sections, as well as to Michael DiRamio, Brie Jeremy Fantl
Gertler, Joseph Shieber, and Baron Reed, for their Jaegwon Kim
work on the first edition. Also helpful were the Matthew McGrath
comments and suggestions by anonymous referees. Ernest So sa

The editors and publisher gratefully acknowledge Doubts (Oxford, Blackwell Publishers, 1991).
the permission granted to reproduce the copy- 1991 by Michael Williams. Reprinted with
right material in this book: permission from the author.
7. Chisholm, Roderick, "The Myth of the
1. Stroud, Barry, "The Problem of the External Given;' pp. 261-86 in R. Chisholm, Philosophy
World;' Chapter 1 III The Significance of (Englewood Cliffs, NJ: Prentice-Hall, 1964).
Philosophical Skepticism (Oxford: Clarendon 1964 by Roderick M. Chisholm.
Press, 1984). 1984 by Barry Stroud. 8. Sellars, Wilfred, "Does Empirical Knowledge
Reprinted with permission from Oxford Have a Foundation?" pp. 293-300 in H. Feigel
University Press. and M. Scriven (eds.), The Foundations of
2. Moore, G. E., "Proof of an External World;' Science and the Concepts of Psychology and
extracted from pp.147-70 in Thomas Baldwin Psychoanalysis, Minnesota Studies in the
(ed.), G. E. Moore: Selected Writings (London Philosophy of Science, Vol. 1 (Minneapolis:
& New York: Routledge, 1993). 1993 by University of Minnesota Press, 1956). 1956.
Thomas Baldwin. Reprinted with permission Reprinted with permission from University
from Taylor & Francis Books (UK). of Minnesota Press.
3. Moore, G. E., "Four Forms of Scepticism;' pp. 9. Sellars, Wilfred, "Epistemic Principles;'
220-2 in Philosophical Papers (New York: pp. 332-49 in H. Casteneda (ed.), Action,
Collier Books, 1962). 1962 by G. E. Moore. Knowledge, and Reality (Indianapolis: Bobbs-
4. Moore, G. E., "Certainty;' extracted from pp. Merrill, 1975). 1975 by Wilfrid Sellars.
171-96 in Thomas Baldwin (ed.), G. E. Moore: 10. BonJour, Laurence,"Can Empirical Knowledge
Selected Writings, (London & New York: Have a Foundation?" pp. 1-13 in American
Routledge, 1993). 1993 by Thomas Baldwin. Philosophical Quarterly 15, 1 (1978). 1978
Reprinted with permission from Taylor & by American Philosophical Quarterly.
Francis Books (UK). Reprinted with permission from American
5. Klein, Peter, "How a Pyrrhonian Skeptic Philosophical Quarterly.
Might Respond to Academic Skepticism;' pp. 11. Davidson, Donald, "A Coherence Theory of
75-94 in Steven Luper (ed.), The Skeptics: Truth and Knowledge;' pp. 307-19 in Ernest
Contemporary Essays (Aldershot: Ashgate, Lepore (ed.), Truth and Interpretation:
2003). 2003 by Steven Luper. Reprinted Perspectives on the Philosophy of Donald
with permission from Ashgate Publishing. Davidson (New York: Blackwell, 1989).
6. Williams, Michael, "Epistemological Realism," 1989 by Ernest Lepore. Reprinted with
pp. 83-93,101-19,121-4,129-39 in Unnatural permission from Blackwell Publishing.

12. Haack, Susan, "A Foundherentist Theory of D. Reidel Publishing Co., 1976). 1976 by
Empirical Justification," pp. 283-93 in Louis Philosophical Studies. Reprinted with per-
Pojman (ed.), The Theory of Knowledge: mission from Springer Science and Business
Classical and Contemporary Readings (Belmont, Media.
CA: Wadsworth, 1999). 1999 by Susan 21. Nozick, Robert, "Knowledge and Skepticism;'
Haack. Reprinted by permission of Susan pp. 172-85, 197-217 in Philosophical
Haack, the copyright holder. Explanations (Cambridge, MA: Harvard
13. So sa, Ernest, "The Raft and the Pyramid;' University Press, 1981). 1981 by Robert
pp. 3-25 in Midwest Studies in Philosophy, Nozick. Reprinted with permission from
Vol. 5: Studies in Epistemology (Minneapolis: Harvard University Press.
University of Minnesota Press, 1980); an 22. Sosa, Ernest, "How to Defeat Opposition to
appendix to this paper is drawn from Ernest Moore;' pp. 141-53 in Philosophical
Sosa, pp. 113-22 "How Do You Know?" Perspectives 13, Epistemology (1999). 1999
American Philosophical Quarterly 11 (1974). by Philosophical Perspectives. Reprinted with
1980 by Midwest Studies in Philosophy. permission from Blackwell Publishing.
Reprinted with permission from the author 23. Vogel,Jonathan,"AreThereCounterexamples
and University of Minnesota Press. to the Closure Principle?;' pp. 13-27 in M. D.
14. Klein, Peter, "Human Knowledge and the Roth and G. Ross (eds.), Doubting, (Kluwer
Infinite Regress of Reasons;' pp. 297-325 Academic Publishers, 1990). 1990 Kluwer
in James Tomberlin (ed.), Philosophical Academic Publishers. Reprinted with per-
Perspedives, 13 Epistemology, 1999. 1999 mission from the author and Springer
by Philosophical Perspectives. Reprinted Science and Business Media.
with permission from Blackwell Publishing. 24. Feldman, Richard and Earl Conee,
15. Gettier, Edmund, "Is Justified True Belief "Evidentialism," pp. 15-34 in Philosophical
Knowledge?" pp. 121-3 in Analysis (1963). Studies 48 (Kluwer Academic Publishers, 1985).
1963 by Edmund Gettier. Reprinted with 1985 by Philosophical Studies. Reprinted
permission from the author. with permission from the authors and
16. Harman, Gilbert, Selections from Thought Springer Science and Business Media.
(Princeton: Princeton University Press, 1973). 25. Foley, Richard, "Skepticism and Rationality;'
1973 by Princeton University Press, 2001 pp. 69-81 in M. D. Roth and G. Ross (eds.),
renewed PUP. Reprinted with permission Doubting (Dordrecht, Netherlands: Kluwer
from Princeton University Press. Academic Publishers, 1990). 1990 by
17. Zagzebski, Linda, "The Inescapability of Gettier Kluwer Academic Publishers. Reprinted with
Problems;' pp. 65-73 in The Philosophical permission from the author and Springer
Quarterly 44, 174 (Oxford: Blackwell Science and Business Media.
Publishers, 1994). 1994 by The Editors of 26. Goldman, Alvin, "What is Justified Belief?"
The Philosophical Quarterly. Reprinted with pp. 1-23 in G. S. Pappas (ed.), Justification
permission from Blackwell Publishing. and Knowledge (Dordrecht, Netherlands:
18. Williamson, Timothy, "A State of Mind;' D. Reidel, 1976). 1976 by D. Reidel
pp. 21-48 in T. Williamson, Knowledge and Publishing Company. Reprinted with per-
Its Limits (Oxford: Oxford University Press, mission from the author and Springer
2000). 2000 by Timothy Williamson. Science and Business Media.
Reprinted with permission from Blackwell 27. Vogel, Jonathan, "Reliabilism Leveled;'
Publishing. pp. 602-23 in The Journal of Philosophy 97, 11
19. Dretske, Fred, "Epistemic Operators;' (Nov. 2000). 2000 The Journal of Philosophy,
pp. 1007-23 in The Journal of Philosophy 67, Inc. Reprinted with permission from the
24 (Dec. 24,1970). 1970 by The Journal of author and The Journal of Philosophy.
Philosophy, Inc. Reprinted with permission 28. BonJour, Laurence, "Externalist Theories
from the author and The Journal ofPhilosophy. of Empirical Knowledge," pp. 53-73 in
20. Stine, Gail, "Skepticism, RelevantAlternatives P. French, T. Uehling, Jr., and H. Wettstein
and Deductive Closure," pp. 249-61 in Philo- (eds.), Midwest Studies in Philosophy, Vol. 5
sophical Studies 29 (Dordrecht, Netherlands: (Minneapolis: University of Minnesota Press,

1980). 1980 by Midwest Studies in Philosophy. 37. Kvanvig, Jonathan 1., "Why Should
Reprinted with permission from the author Enquiring Minds Want to Know?: Meno
and University of Minnesota Press. Problems and Epistemological Axiology," pp.
29. Goldman, Alvin, "Internalism Exposed;' 426-52 in The Monist 81,3 (1998). 1998,
pp. 271-93 in The Journal of Philosophy 96, THE MONIST, La Salle, Illinois 61301, USA.
6 (Jun. 1999). 1999 The Journal of Reprinted with permission from the author
Philosophy, Inc. Reprinted with permission and publisher.
from the author and The Journal of 38. Elgin, Catherine Z., "True Enough;' pp.
Philosophy. 113-131 in Philosophical Issues 14, Episte-
30. Fumerton, Richard, "Externalism and mology (2004). 2004 by Philosophical Issues.
Skepticism;' pp. 159-81 in Metaepistemology Reprinted with permission from Blackwell
and Skepticism (Lanham, MD: Rowman and Publishing.
Littlefield, 1995). 1995 by Richard Fumerton. 39. Quine, W. V., "Epistemology Naturalized;'
Reprinted with permission from Rowman & pp. 68-90 in Ontological Relativity and Other
Littlefield Publishing Group. Essays (Columbia University Press, 1969). 1969
31. Feldman, Richard and Earl Conee, "Internalism by Columbia University Press. Reprinted with
Defended,"pp.1-18inAmericanPhilosophical permission of the publisher.
Quarterly 38, 1 (January 2001). 2001 by 40. Kim, Jaegwon, "What is 'Naturalized
American Philosophical Quarterly. Reprinted Epistemology'?" m J. Tomberlin (ed.),
with permission from American Philosophical Philosophical Perspectives 2. Epistemology
Quarterly. (Atascadero, CA: Ridgeview Publishing Co.,
32. Plantinga, Alvin, "Warrant: A First Approxi- 1988), pp. 381-405. 1998 by Philosophical
mation;' pp. 3-20 in Warrant and Proper Perspectives. Reprinted with permission from
Function (Oxford and New York: Oxford Blackwell Publishing.
University Press, 1993). 1993 by Alvin 41. Antony, Louise M., "Quine as Feminist: the
Plantinga. Reprinted with permission from Radical Import of Naturalized Epistemology;'
Oxford University Press. pp. 185-225 in 1. Antony and C. Witt (eds.),
33. Zagzebski, Linda, Selections from Virtues of A Mind ofOne's Own (Boulder, CO: Westview,
the Mind, pp. 134-7, 166-84 in 1. Zagzebski, 1993). 1993 by Westview Press, a member
Virtues of the Mind (Cambridge: Cambridge of the Perseus Books Group. Reprinted with
University Press, 1996). 1996 by Cambridge permission from Westview Press, a member
University Press. Reprinted with permission of the Perseus Books Group.
from the author and Cambridge University 42. Putnam, Hilary, "There is a Least One A
Press. Priori Truth;' pp. 153-170 in Erkenntnis 13
34. Greco, John, "Virtues and Vices of Virtue (Dordrecht,Netherlands: D. Reidel Publishing
Epistemology," pp. 413-32 in Canadian Co., 1978). 1978 by D. Reidel Publishing
Journal of Philosophy 23, 3 (1993). 1993 by Company. Reprinted with permission from
Canadian Journal of Philosophy. Reprinted by the author and Springer Science and Business
permission of University of Calgary Press. Media.
35. Pritchard, Duncan, "Cognitive Responsibility 43. Casullo, Albert, "Revisability, Reliabilism,
and the Epistemic Virtues;' pp. 181-201 in and A Priori Knowledge;' pp. 187-213 in
Epistemic Luck (Oxford: Oxford University Philosophy and Phenomenology Research 49,
Press, 2005). 2005 by Duncan Pritchard. 2 (Dec. 1988). 1988 by Philosophy and
Reprinted with permission from Oxford Phenomenology Research. Reprinted with
University Press, Inc. permission from Blackwell Publishing.
36. Sosa, Ernest, "The Place of Truth in 44. Bealer, George, "A Priori Knowledge and the
Epistemology;' pp. 155-79 in M. DePaul and Scope of Philosophy;' pp. 121-42 in
1. Zagzebski (eds.), Intellectual Virtue: Philosophical Studies 81 (Kluwer Academic
Perspectives From Ethics and Epistemology Publishers, 1996). 1996 Kluwer Academic
(Oxford and New York: Oxford University Publishers. Reprinted with permission from
Press, 2003). 2003 by Ernest Sosa. the author and Springer Science and Business
Reprinted with permission from the author. Media.

45. Weinberg, Jonathan M., Shaun Nichols, and Oxford University Press. Reprinted with per-
Stephen Stich, "Normativity and Epistemic mission from Oxford University Press, Inc.
Intuitions:' pp. 429-60 in Philosophical Topics 54. Baker, Judith, "Trust and Rationality,"
29, 1 & 2 (2001). 2001 by Philosophical pp. 1-13 in Pacific Philosophical Quarterly
Topics. Used with permission of the authors 68 (1987). 1987 Pacific Philosophical
and University of Arkansas Press, www. Quarterly. Reprinted with permission from Blackwell Publishing.
46. Kornblith, Hilary, "Investigating Knowledge 55. Fricker, Elizabeth, ''Against Gullibility:' pp.
Itself:' pp. 1-27 in H. Kornblith, Knowledge 125-61 in B. K. Matilal and A. Chakrabarti
and its Place in Nature (Oxford: Clarendon, (eds.) Knowing from Words (Dordrecht,
2002). 2002 by Hilary Kornblith. Reprinted Netherlands: Kluwer Academic Publishers,
with permission from Oxford University 1994). 1994 Kluwer Academic Publishers.
Press. Reprinted with permission from the author
47. DeRose, Keith, "Solving the Skeptical and Springer Science and Business Media.
Problem:'pp.1-7, 17-52 in The Philosophical 56. Burge, Tyler, "Content Preservation,"
Review 104, 1 (1995). 1995 by The pp. 457-88 in The Philosophical Review, 102,
Philosophical Review. Reprinted with per- 4 (Oct. 1993). 1993 by Cornell University
mission from The Philosophical Review. Press. All rights reserved. Used by permis-
48. Lewis, David, "Elusive Knowledge;' pp. 549-67 sion of the publisher.
in Australasian Journal of Philosophy 74, 4 57. Lackey, Jennifer, "Testimonial Knowledge and
(1996). 1996 by Australasian Journal of Transmission;' pp. 471-90 in The Philosophical
Philosophy. Reprinted with permission from Quarterly, 49, 197 (Oct. 1999). The Editors
Taylor & Francis (UK). of The Philosophical Quarterly, 1999.
49. Cohen, Stewart, "Contextualist Solutions Reprinted with permission from Blackwell
to Epistemological Problems: Scepticism, Publishing.
Gettier, and the Lottery;' pp. 289-306 in 58. Huemer, Michael, "The Problem of Memory
Australasian Journal of Philosophy 76, 2 Knowledge," pp. 346-57 in The Philosophical
(1998). 1998 by Australasian Journal of Quarterly, 80,197 (1999). 1999 University
Philosophy. Reprinted with permission from of Southern California and Blackwell
the author and Taylor & Francis (UK). Publishers Ltd. Reprinted with permission
50. Stanley, Jason, "Introduction" and "Knowledge from Blackwell Publishing.
Ascriptions and Context-Sensitivity;' pp. 1-15, 59. McDowell, John, "Criteria, Defeasibility, and
47-73 in Knowledge and Practical Interest Knowledge," pp. 455-79 in Proceedings of
(New York: Oxford University Press, 2005). the British Academy 68 (1982). Annual
2005 by Jason Stanley. Reprinted with per- Philosophical Lecture. 1983 by The British
mission from Oxford University Press, Inc. Academy. Reprinted with permission from
51. Fantl, Jeremy and Matthew McGrath, Proceedings of the British Academy LXVIII;
"Evidence, Pragmatics, and Justification," 1982.
pp. 67-94 in The Philosophical Review Ill, 1 60. Reynolds, Steven L., "Knowing How to
(Jan. 2002). 2002 by Cornell University Believe with Justification;' pp. 273-92 in
Press. All rights reserved. Used by permis- Philosophical Studies 64, 3 (Dec. 1991). 1991
sion of the publisher. by Philosophical Studies. Reprinted with
52. Hawthorne, John, "Sensitive Moderate permission from the author and Springer
Invariantism;' pp. 157-91 in Knowledge and Science and Business Media.
Lotteries (Oxford: Clarendon, 2004). 2004
by John Hawthorne. Reprinted with permis- Every effort has been made to trace copyright
sion from Oxford University Press. holders and to obtain their permission for the use
53. MacFarlane, John, "The Assessment Sensitivity of copyright material. The publisher apologizes
of Knowledge Attributions;' pp. 197-233 in for any errors or omissions in the above list and
Tamar Szab6 Gendler and John O'Leary would be grateful if notified of any corrections
Hawthorne (eds.), Oxford Studies in Epistemology, that should be incorporated in future reprints or
Vol. 1 (Oxford: Clarendon, 2005). 2005 by editions of this book.

Like Rene Descartes, we have all asked ourselves at one time or another "Couldn't
everything I seem to see, hear, etc. be illusory? Might I in fact be dreaming all this? If so,
what do I really know of the outside world?" The skeptic's answers are pessimistic: yes,
you could be dreaming, and so you know nothing of the outside world. The conclusion
is outlandish, and yet the reasoning behind it hardly seems strained at all. We feel the
pressure towards skepticism in the movement from the question about the trust-
worthiness of our senses to the question of our ability to know. Given that the bulk of
our knowledge of the outside world derives from the senses, how can we know any-
thing about the world unless we first show that our senses can be trusted? The core of
the skeptical strategy is more general: how can one gain knowledge using a source of
belief unless one first shows that the source is trustworthy?
In his selection, Barry Stroud presents the skeptic's argument in its most favorable
light. The skeptic does not hold us up to an uncommonly high standard of knowledge
only to make the obvious point that we fail to meet it. The skeptic invokes only
the standards presupposed in everyday knowledge attributions. To use an example of
Stroud's, if no goldfinch could possibly be a canary, then if one is to know that the bird
one sees is a goldfinch, one must be able to rule out its being a canary. More generally,
to know that p, one must be able to rule out every possibility one knows to be incom-
patible with one's knowing that p. The skeptic then has her wedge: to know that you're
sitting beside a warm fire, you must be able to rule out any possibility which excludes
this knowledge, including innumerable "skeptical possibilities:' such as that you're
dreaming, that you're being deceived by a malicious demon, and that you're a brain in
a vat stimulated to have the experiences and apparent memories you now have. But it's
hard to see how you can rule these out.
In each of the selections from the work of G. E. Moore, the tables are turned on the
skeptic. Moore provides a counter-argument in "Proof of an External World." A good
proof, he explains, proceeds from known premises to a distinct conclusion to which
they can be seen to lead. He then produces an example: raising his hands, one after the
other, he exclaims "Here is a hand. Here is another hand," and he concludes "There are

hands:' If asked to prove his premises, he would reject the demand, for not everything
that is known can be proved.
Moore nevertheless takes the skeptic seriously. In "Certainty;' he grants that if he
doesn't know he is not dreaming, he doesn't know he is standing up giving a lecture.
Still he asks why there is any more plausibility in using this premise as part of a modus
ponens argument to conclude that he doesn't know he is standing up than in using it as
a part of the corresponding modus tollens argument to conclude that he does know
after all that he is not dreaming.
In "Four Forms of Scepticism;' Moore fully admits that skeptical scenarios are
logically possible, but he finds it more certain that something has gone afoul in the
skeptical argument than that he lacks knowledge that he has hands (or is holding a
pencil). Moreover, he concludes that since the only way he could know this is through
some inductive or analogical argument from the character of his experience, such an
argument must exist.
The selections from Stroud and Moore concern our knowledge of the external
world. One might hope that, even if it is hard to answer the skeptical challenge for
knowledge, at least it could be satisfactorily answered for justification. Peter Klein calls
the view that we cannot be justified in our beliefs about how things are (as opposed to
how they seem) "Academic skepticism" and contrasts it with an older form of skepti-
cism: Pyrrhonism. Pyrrhonism, in Klein's view, is a more moderate skepticism than its
Academic cousin, for Pyrrhonism allows that our beliefs can be conditionally or pro-
visionally justified. But it is still a form of skepticism, because it denies that our beliefs
can be completely justified. Only if reasoning could settle the matter of whether a
belief is true could that belief be completely justified. But how can reasoning settle
anything? If it were legitimate to end reasoning with a proposition for which we could
not provide a further reason, then it seems reasoning could settle some matter. But this
is not legitimate. Nor is it legitimate to reason in a circle. Therefore, the only way for
reasoning to settle matters would be to complete an infinite regress of non-repeating
reasons (a view Klein refers to as "infinitism," discussed in more detail in his contribu-
tion to Part 11). While this would be a legitimate way to settle some matter, it cannot,
in fact, be done.
The lesson for Academic skepticism is that the arguments invoked in favor of
Academic skepticism are themselves fallacious in that they either rely on arbitrary
premises or beg the question in favor of their conclusion. Thus, consider the Academic
skeptic's claim that we cannot know whether we are dreaming or deceived by a mali-
cious demon. This claim is central to the argument for Academic skepticism. If it is
unsupported, it is arbitrary. To support the claim, the Academic skeptic must first dem-
onstrate that we cannot know, say, that there is a table in front of us. But "I cannot know
there is a table in front of me" is the ostensible conclusion of the skeptical argument.
Therefore, Academic skepticism, like the inadequate models of reasoning, must either
rely on arbitrary premises or beg the question.
Michael Williams argues that if there is such a thing as knowledge of the external
world, the kind of knowledge the Cartesian skeptic questions, it seems impossible for us
to see ourselves as having it. That is, the skeptic would carry the day. But he asks: is there
such a thing as knowledge of the world? His answer is no. The concept of knowledge of
the external world is a theoretical concept, and so, unlike practical concepts such as the
concept of a chair, it lacks application entirely unless there is an appropriate unified

domain of reality whose contours are there for it to match. But there is no such epi-
stemic domain. There could be only if (empirical) beliefs divided into two classes: those
that could only be known on the basis of beliefs about immediate experience, i.e.,
beliefs about the external world, and those that could be known directly from immedi-
ate experience. Yet an examination of our practices in attributing knowledge and justi-
fication suggests that beliefs do not divide into these epistemic categories nor into any
objective epistemic categories.
Williams describes his view as a form of contextualism. But it is a version of context-
ualism quite different from those appearing in Part VIII of this volume. The contextual-
ist theories of DeRose and Cohen, and to a lesser extent Lewis, presuppose the existence
of a unified range of objective characteristics which, given a speech context, comprise the
truth-conditions for knowledge attributions in that context. For DeRose, there are the
objective (context-invariant) notions of sensitivity and strength of epistemic position,
and for Cohen objective notions of strength of evidence or justification. For Lewis,
there are the objective factors of one's evidence and which possibilities it rules out. For
all three of these epistemologists, the function of context is to set the bar on which (or
what degree) of a relatively unified range of objective factors count. Thus, for them,
there is an independent place for epistemological inquiry into the nature of these objec-
tive factors as well as into how they feed into the semantics of knowledge attribution.
According to Williams, by contrast, there is no range of objective factors, with
the result that there is nothing at all to serve as an object of theoretical investigation for
the epistemologist.
Part and parcel of repudiating skepticism, then, is repudiating traditional epistemol-
ogy. Both rely for their livelihood on the assumption that Williams calls "epistemological
realism:' viz. that there are objective relations of epistemic priority waiting to be

Further Reading

Annas, Julia and Jonathan Barnes, The Modes of Landesman, Charles and Roblin Meeks (eds),
Scepticism (Cambridge: Cambridge University Philosophical Skepticism: From Plato to Rorty
Press, 1985). (Oxford: Blackwell, 2002).
Burnyeat, M. F. (ed.), The Skeptical Tradition Klein, P., Certainty: A Refutation of Scepticism
(Berkeley: University of California Press, 1983). (Minneapolis: University of Minnesota Press,
Clarke, Thompson, "The Legacy of Skepticism;' 1991).
Journal of Philosophy 69 (1972), pp. 754-69. Nozick, Robert, Philosophical Explanations
DeRose, K. and T. Warfield (eds), Skepticism: A (Oxford: Oxford University Press, 1981).
Contemporary Reader (Oxford: Oxford Popkin, Richard, Scepticism from Erasmus to
University Press, 1999). Spinoza (Berkeley: University of California
Fumerton, R., Metaepistemology and Skepticism Press, 1979).
(Langham, MD: Rowman and Littlefield, 1995). Roth, Michael D. and Glenn Ross (eds),Doubting:
Hadot, Pierre, Philosophy as a Way ofLife (Oxford: Contemporary Perspectives on Skepticism
Blackwell, 1995). (Dordrecht: Kluwer Academic Publishers,
Hankinson, R. J., The Sceptics (London: Routledge, 1990).
1995). Sosa, Ernest, "The Skeptic's Appeal;' in Marjorie
Huemer, Michael, Skepticism and the Veil of Clay and Keith Lehrer (eds), Knowledge and
Perception (Langham, MD: Rowman and Skepticism (Boulder, CO: Westview Press,
Littlefield, 2001). 1989), pp. 51-68.

Strawson, P. F., Skepticism and Naturalism: Some - - , The Significance of Philosophical Scepticism
Varieties (London: Methuen, 1985). (Oxford: Oxford University Press, 1984).
Stroud, Barry, "Understanding Human Knowledge Unger, Peter, Ignorance: A Case for Scepticism
in General;' in Marjorie Clay and Keith Lehrer (Oxford: Clarendon Press, 1975).
(eds), Knowledge and Skepticism (Boulder, CO: Williams, Michael, Unnatural Doubts (Princeton:
Westview Press, 1989), pp. 31-50. Princeton University Press, 1996).
The Problem of the
External World

Barry Stroud

Since at least the time of Descartes in the seven- the philosophical problem of our knowledge of
teenth century there has been a philosophical the external world.
problem about our knowledge of the world The problem arose for Descartes in the course
around us. 1 Put most simply, the problem is to of reflecting on everything he knows. He reached a
show how we can have any knowledge of the point in his life at which he tried to sit back and
world at all. The conclusion that we cannot, that reflect on everything he had ever been taught or
no one knows anything about the world around told, everything he had learned or discovered
us, is what I call "scepticism about the external or believed since he was old enough to know or
world", so we could also say that the problem is to believe anything. 2 We might say that he was reflect-
show how or why scepticism about the external ing on his knowledge, but putting it that way could
world is not correct. My aim is not to solve the suggest that what he was directing his attention to
problem but to understand it. I believe the prob- was indeed knowledge, and whether it was knowl-
lem has no solution; or rather that the only answer edge or not is precisely what he wanted to deter-
to the question as it is meant to be understood is mine. "Among all the things I believe or take to be
that we can know nothing about the world around true, what amounts to knowledge and what does
us. But how is the question meant to be under- not?"; that is the question Descartes asks himself.
stood? It can be expressed in a few English words It is obviously a very general question, since it asks
familiar to all of us, but I hope to show that an about everything he believes or takes to be true,
understanding of the special philosophical char- but in other respects it sounds just like the sort of
acter of the question, and of the inevitability of question we are perfectly familiar with in everyday
an unsatisfactory answer to it, cannot be guaran- life and often know how to answer.
teed by our understanding of those words alone. For example, I have come to accept over the
To see how the problem is meant to be under- years a great many things about the common
stood we must therefore examine what is per- cold. I have always been told that one can catch
haps best described as its source - how the cold by getting wet feet, or from sitting in a
problem arises and how it acquires that special draught, or from not drying one's hair before
character that makes an unsatisfactory negative going outdoors in cold weather. I have also learned
answer inevitable. We must try to understand that the common cold is the effect of a virus
transmitted by an already infected person. And
Originally published in B. Stroud, The Significance of I also believe that one is more vulnerable to colds
Philosophical Skepticism (Oxford: Clarendon Press, when over-tired, under stress, or otherwise in less
1984), ch. 1. than the best of health. Some of these beliefs seem

to me on reflection to be inconsistent with some In pointing out that we are perfectly familiar
others; I see that it is very unlikely that all of them with the idea of investigating or reviewing our
could be true. Perhaps they could be, but I knowledge on some particular matter or in some
acknowledge that there is much I do not under- general area I do not mean to suggest that it is
stand. If I sit back and try to think about all my always easy to settle the question. Depending on
"knowledge" of the common cold, then, I might the nature of the case, it might be very difficult,
easily come to wonder how much of it really perhaps even impossible at the time, to reach a
amounts to knowledge and how much does not. firm conclusion. For example, it would probably
What do I really know about the common cold? If be very difficult if not impossible for me to trace
I were sufficiently interested in pursuing the and assess the origins of many of those things I
matter it would be natural to look into the source believe about the common cold. But it is equally
of my beliefs. Has there ever been any good reason true that sometimes it is not impossible or even
for thinking that colds are even correlated with especially difficult to answer the question. We do
wet hair in cold weather, for example, or with sit- sometimes discover that we do not really know
ting in a draught? Are the people from whom I what we previously thought we knew. I might
learned such things likely to have believed them find that what I had previously believed is not
for good reasons? Are those beliefs just old wives' even true - that sitting in draughts is not even
tales, or are they really true, and perhaps even correlated with catching a cold, for example. Or I
known to be true by some people? These are ques- might find that there is not or perhaps never was
tions I might ask myself, and I have at least a any good reason to believe what I believed - that
general idea of how to go about answering them. the man's alibi was concocted and then falsely tes-
Apart from my impression of the implausibil- tified to by his friends. I could reasonably con-
ity of all my beliefs about the common cold being clude in each case that I, and everyone else for
true together, I have not mentioned any other that matter, never did know what I had previously
reason for being interested in investigating the thought I knew. We are all familiar with the ordi-
state of my knowledge on that subject. But for the nary activity of reviewing our knowledge, and
moment that does not seem to affect the intelligi- with the experience of reaching a positive verdict
bility or the feasibility of the reflective project. in some cases and a negative verdict in others.
There is nothing mysterious about it. It is the sort Descartes's own interest in what he knows and
of task we can be led to undertake for a number how he knows it is part of his search for what he
of reasons, and often very good reasons, in so far calls a general method for "rightly conducting
as we have very good reasons for preferring knowl- reason and seeking truth in the sciences".3 He
edge and firm belief to guesswork or wishful wants a method of inquiry that he can be assured
thinking or simply taking things for granted. in advance will lead only to the truth if properly
Reflection on or investigation of our putative followed. I think we do not need to endorse the
knowledge need not always extend to a wide area of wisdom of that search or the feasibility of that
interest. It might be important to ask whether some programme in order to try to go along with
quite specific and particular thing I believe or have Descartes in his general assessment of the posi-
been taking for granted is really something I know. tion he is in with respect to the things he believes.
As a member of a jury I might find that I have been He comes to find his putative knowledge wanting
ruling out one suspect in my mind because he was a in certain general respects, and it is in the course
thousand miles away, in Cleveland, at the time of the of that original negative assessment that the prob-
crime. But I might then begin to ask myself whether lem I am interested in arises. I call the assessment
that is really something that I know. I would reflect "negative" because by the end of his First
on the source of my belief, but reflection in this case Meditation Descartes finds that he has no good
need not involve a general scrutiny of everything I reason to believe anything about the world
take myself to know about the case. Re-examining around him and therefore that he can know noth-
the man's alibi and the credentials of its supporting ing of the external world.
witnesses might be enough to satisfy me. Indeed How is that assessment conducted, and how
I might find that its reliability on those counts is closely does it parallel the familiar kind of review
precisely what I had been going on all along. of our knowledge that we all know how to conduct

in everyday life? The question in one form or the addition of the number one to a given number;
another will be with us for the rest of this book. It I would still have to do the same for the addition
is the question of what exactly the problem of our of two, and then the addition of three, and so on.
knowledge of the external world amounts to, and And even that would exhaust only my beliefs
how it arises with its special philosophical charac- about addition; all my other mathematical beliefs,
ter. The source of the problem is to be found not to mention all the rest of my knowledge,
somewhere within or behind the kind of thinking would remain so far unexamined. Obviously the
Descartes engages in. job cannot be done piecemeal, one by one. Some
One way Descartes's question about his knowl- method must be found for assessing large classes
edge differs from the everyday examples I consid- of beliefs all at once.
ered is in being concerned with everything he One way to do this would be to look for
believes or takes to be true. How does one go common sources or channels or bases of our
about assessing all of one's knowledge all at once? beliefs, and then to examine the reliability of
I was able to list a few of the things I believe about those sources or bases, just as I examined the
the common cold and then to ask about each of source or basis of my belief that the suspect was
them whether I really know it, and if so how. But in Cleveland. Descartes describes such a search as
although I can certainly list a number of the a search for "principles" of human knowledge,
things I believe, and I would assent to many more "principles" whose general credentials he can
of them as soon as they were put to me, there then investigate (HR, 145). If some "principles"
obviously is no hope of assessing everything I are found to be involved in all or even most of our
believe in this piecemeal way. For one thing, it knowledge, an assessment of the reliability of
probably makes no sense, strictly speaking, to talk those "principles" could be an assessment of all or
of the number of things one believes. If I am most of our knowledge. If I found good reason to
asked whether it is one of my beliefs that I went to doubt the reliability of the suspect's alibi, for
see a film last night I can truly answer "Yes". If I example, and that was all I had to go on in my
were asked whether it is one of my beliefs that belief that he was in Cleveland, then what I earlier
I went to the movies last night I would give the took to be my knowledge that he was in Cleveland
same answer. Have I thereby identified two, or would have been found wanting or called into
only one, of my beliefs? How is that question ever question. Its source or basis would have been
to be settled? If we say that I identified only one of undermined. Similarly, if one of the "principles"
my beliefs, it would seem that I must also be said or bases on which all my knowledge of the world
to hold the further belief that going to see a film depends were found to be unreliable, my knowl-
and going to the movies are one and the same edge of the world would to that extent have been
thing. So we would have more than one belief found wanting or called into question as well.
after all. The prospects of arriving even at a prin- Are there any important "principles" of human
ciple for counting beliefs, let alone at an actual knowledge in Descartes's sense? It takes very little
number of them, seem dim. reflection on the human organism to convince us
Even if it did make sense to count the things of the importance of the senses - sight, hearing,
we believe it is pretty clear that the number would touch, taste, and smell. Descartes puts the point
be indefinitely large and so an assessment of our most strongly when he says that "all that up to the
beliefs one by one could never be completed present time I have accepted as most true and
anyway. This is easily seen by considering only certain I have learned either from the senses or
some of the simplest things one knows, for through the senses" (HR, 145). Exactly what he
example in arithmetic. One thing I know is that would include under "the senses" here is perhaps
one plus one equals two. Another thing I know is somewhat indeterminate, but even if it is left
that one plus two is three, and another, that one vague many philosophers would deny what
plus three is four. Obviously there could be no Descartes appears to be saying. They would hold
end to the task of assessing my knowledge if I had that, for example, the mathematical knowledge I
to investigate separately the source of each one of mentioned earlier is not and could not be acquired
my beliefs in that series. And even if I succeeded I from the senses or through the senses, so not every-
would only have assessed the things I know about thing I know is known in that way. Whether

Descartes is really denying the views of those who really are a certain way, I might still be wrong.
believe in the non-sensory character of mathe- We have all found at one time or another that we
matical knowledge, and whether, if he were, he have been misled by appearances; we know that
would be right, are issues we can set aside for the the senses are not always reliable. Should we not
moment. It is clear that the senses are at least very conclude, then, that as a general source of know1-
important for human knowledge. Even restrict- edge the senses are not to be trusted? As Descartes
ing ourselves to the traditional five senses we can puts it, is it not wiser never "to trust entirely to
begin to appreciate their importance by reflect- any thing by which we have once been deceived"
ing on how little someone would ever come to (HR, 145)? Don't we have here a quite general
know without them. A person blind and deaf way of condemning as not fully reliable all of our
from birth who also lacked taste buds and a sense beliefs acquired by means of the senses?
of smell would know very little about anything, I think the answer to that question is "No, we
no matter how long he lived. To imagine him also do not", and I think Descartes would agree with
anaesthetized or without a sense of touch is per- that answer. It is true that he does talk of the
haps to stretch altogether too far one's conception senses "deceiving" us on particular occasions, and
of a human organism, or at least a human organ- he does ask whether that is not enough to con-
ism from whom we can hope to learn something demn the senses in general as a source of knowl-
about human knowledge. The importance of the edge, but he immediately reminds us of the
senses as a source or channel of knowledge seems obvious fact that the circumstances in which the
undeniable. It seems possible, then, to acknowl- senses "deceive" us might be special in certain
edge their importance and to assess the reliability ascertainable ways, and so their occasional fail-
of that source, quite independently of the difficult ures would not support a blanket condemnation
question of whether all our knowledge comes to of their reliability.
us in that way. We would then be assessing the Sometimes, to give an ancient example, a
credentials of what is often called our "sensory" or tower looks round from a distance when it is
"experiential" or "expirical" knowledge, and that, actually square. If we relied only on the appear-
as we shall see, is quite enough to be going on with. ances of the moment we might say that the dis-
Having found an extremely important "prin- tant tower is round, and we would be wrong. We
ciple" or source of our knowledge, how can we also know that there are many small organisms
investigate or assess all the knowledge we get from invisible to the naked eye. If the table before me is
that source? As before, we are faced with the prob- covered with such organisms at the moment but
lem of the inexhaustibility of the things we believe I look at it and say there is nothing on the table at
on that basis, so no piecemeal, one-by-one proce- all, once again I will be wrong. But all that follows
dure will do. But perhaps we can make a sweeping from these familiar facts, as Descartes points out,
negative assessment. It might seem that as soon as is that there are things about which we can be
we have found that the senses are one of the wrong, or there are situations in which we can get
sources of our beliefs we are immediately in a false beliefs, if we rely entirely on our senses at
position to condemn all putative knowledge that moment. So sometimes we should be careful
derived from them. Some philosophers appear to about what we believe on the basis of the senses,
have reasoned in this way, and many have even or sometimes perhaps we should withhold our
supposed that Descartes is among them. The idea assent from any statement about how things are -
is that if I am assessing the reliability of my beliefs when things are too far away to be seen properly,
and asking whether I really know what I take for example, or too small to be seen at all. But that
myself to know, and I come across a large class of obviously is not enough to support the policy of
beliefs which have come to me through the senses, never trusting one's senses, or never believing
I can immediately dismiss all those beliefs as anything based on them. Nor does it show that I
unreliable or as not amounting to knowledge can never know anything by means of the senses.
because of the obvious fact that I can sometimes If my car starts promptly every morning for two
be wrong in my beliefs based on the senses. Things years in temperate weather at sea level but then
are not always as they appear, so if on the basis of fails to start one morning in freezing weather at
the way they appear to me I believe that they the top of a high mountain, that does not support

the policy of never trusting my car to start again position any of us can ever be in for knowing
once I return to the temperate lower altitude from things about the world around us on the basis of
which I so foolishly took it. Nor does it show that the senses. What is true of a representative case, if
I can never know whether my car will ever start it is truly representative and does not depend on
again. It shows only that there are certain circum- special peculiarities of its own, can legitimately
stances in which my otherwise fully reliable car support a general conclusion. A demonstration
might not start. So the fact that we are sometimes that a particular isosceles triangle has a certain
wrong or "deceived" in our judgements based on property, for example, can be taken as a demon-
the senses is not enough in itself to show that the stration that all isosceles triangles have that prop-
senses are never to be trusted and are therefore erty, as long as the original instance was typical or
never reliable as a source of knowledge. representative of the whole class. Whether
Descartes's negative assessment of all of his Descartes's investigation of the general reliability
sensory knowledge does not depend on any such of the senses really does follow that familiar pat-
reasoning. He starts his investigation, rather, in tern is a difficult question. Whether, or in pre-
what would seem to be the most favourable con- cisely what sense, the example he considers can be
ditions for the reliable operation of the senses as a treated as representative of our relation to the
source of knowledge. While engaging in the very world around us is, I believe, the key to under-
philosophical reflections he is writing about in standing the problem of our knowledge of the
his First Meditation Descartes is sitting in a warm external world. But if it turns out that there is
room, by the fire, in a dressing gown, with a piece nothing illegitimate about the way his negative
of paper in his hand. He finds that although he conclusion is reached, the problem will be prop-
might be able to doubt that a distant tower that erlyposed.
looks round really is round, it seems impossible For the moment I think at least this much can
to doubt that he really is sitting there by the fire in be said about Descartes's reasoning. He chooses
his dressing gown with a piece of paper in his the situation in which he finds himself as repre-
hand. The fire and the piece of paper are not too sentative of the best position we can be in for
small or too far away to be seen properly, they are knowing things about the world in the sense that,
right there before his eyes; it seems to be the best if it is impossible for him in that position to know
kind of position someone could be in for getting that he is sitting by the fire with a piece of paper in
reliable beliefs or knowledge by means of the his hand then it is also impossible for him in other
senses about what is going on around him. That situations to know anything about the world
is just how Descartes regards it. Its being a best- around him on the basis of his senses. A negative
possible case of that kind is precisely what he verdict in the chosen case would support a nega-
thinks enables him to investigate or assess at one tiveverdict everywhere else. The example Descartes
fell swoop all our sensory knowledge of the world considers is in that sense meant to be the best kind
around us. The verdict he arrives at about his of case there could be of sensory knowledge about
putative knowledge that he is sitting by the fire the world around us. I think we must admit that it
with a piece of paper in his hand in that particu- is very difficult to see how Descartes or anyone
lar situation serves as the basis for a completely else could be any better off with respect to know-
general assessment of the senses as a source of ing something about the world around him on the
knowledge about the world around us. basis of the senses than he is in the case he consid-
How can that be so? How can he so easily reach ers. But if no one could be in any better position
a general verdict about all his sensory knowledge for knowing, it seems natural to conclude that any
on the basis of a single example? Obviously not negative verdict arrived at about this example, any
simply by generalizing from one particular exam- discovery that Descartes's beliefs in this case are
ple to all cases of sensory knowledge, as one might not reliable or do not amount to knowledge, could
wildly leap to a conclusion about all red-haired safely be generalized into a negative conclusion
men on the basis of one or two individuals. about all of our sensory "knowledge" of the world.
Rather, he takes the particular example of his If candidates with the best possible credentials
conviction that he is sitting by the fire with a piece are found wanting, all those with less impressive
of paper in his hand as representative of the best credentials must fall short as well.

It will seem at first sight that in conceding that With this thought, if he is right, Descartes has
the whole question turns on whether Descartes lost the whole world. He knows what he is experi-
knows in this particular case we are conceding encing, he knows how things appear to him, but
very little; it seems obvious that Descartes on that he does not know whether he is in fact sitting by
occasion does know what he thinks he knows the fire with a piece of paper in his hand. It is, for
about the world around him. But in fact Descartes him, exactly as if he were sitting by the fire with a
finds that he cannot know in this case that he is piece of paper in his hand, but he does not know
sitting by the fire with a piece of paper in his whether there really is a fire or a piece of paper
hand. If the case is truly representative of our sen- there or not; he does not know what is really hap-
sory knowledge in general, that will show that no pening in the world around him. He realizes that
one can know anything about the world around if everything he can ever learn about what is hap-
us. But how could he ever arrive at that negative pening in the world around him comes to him
verdict in the particular case he considers? How through the senses, but he cannot tell by means of
could anyone possibly doubt in such a case that the senses whether or not he is dreaming, then all the
the fire and the piece of paper are there? The sensory experiences he is having are compatible
paper is in Descartes's hand, the fire is right there with his merely dreaming of a world around him
before his open eyes, and he feels its warmth. while in fact that world is very different from the
Wouldn't anyone have to be mad to deny that he way he takes it to be. That is why he thinks he
can know something about what is going on must find some way to tell that he is not dream-
around him in those circumstances? Descartes ing. Far from its being mad to deny that he knows
first answers "Yes". He says that if he were to in this case, he thinks his recognition of the pos-
doubt or deny on that occasion that he is sitting sibility that he might be dreaming gives him "very
by the fire with a piece of paper in his hand he powerful and maturely considered" (HR, 148)
would be no less mad than those paupers who reasons for withholding his judgement about how
say they are kings or those madmen who think things are in the world around him. He thinks it is
they are pumpkins or are made of glass. But his eminently reasonable to insist that ifhe is to know
reflections continue: that he is sitting by the fire he must know that he
is not dreaming that he is sitting by the fire. That
At the same time I must remember that I am a is seen as a necessary condition of knowing some-
man, and that consequently I am in the habit of thing about the world around him. And he finds
sleeping, and in my dreams representing to myself that that condition cannot be fulfilled. On careful
the same things or sometimes even less probable reflection he discovers that "there are no certain
things, than do those who are insane in their indications by which we may clearly distinguish
waking moments. How often has it happened to wakefulness from sleep': He concludes that he
me that in the night I dreamt that I found myself knows nothing about the world around him
in this particular place, that I was dressed and because he cannot tell that he is not dreaming; he
seated near the fire, whilst in reality I was lying cannot fulfil one of the conditions necessary for
undressed in bed! At this moment it does indeed
knowing something about the world.
seem to me that it is with eyes awake that I am
The Cartesian problem of our knowledge of the
looking at this paper; that this head which I move
external world therefore becomes: how can we know
is not asleep, that it is deliberately and of set pur-
anything about the world around us on the basis of
pose that I extend my hand and perceive it; what
the senses if the senses give us only what Descartes
happens in sleep does not appear so clear nor so
distinct as does all this. But in thinking over this I says they give us? What we gain through the senses
remind myself that on many occasions I have in is on Descartes's view only information that is com-
sleep been deceived by similar illusions, and in patible with our dreaming things about the world
dwelling carefully on this reflection I see so mani- around us and not knowing anything about the
festly that there are no certain indications by world. How then can we know anything about the
which we may clearly distinguish wakefulness world by means of the senses? The Cartesian argu-
from sleep that I am lost in astonishment. And my ment presents a challenge to our knowledge, and
astonishment is such that it is almost capable of the problem of our knowledge of the external world
persuading me that I now dream. (HR, 145-6) is to show how that challenge can be met.

When I speak here of the Cartesian argument they accept it. There seems to me no doubt about
or of Descartes's sceptical conclusion or of his the force and the fascination - I would say the
negative verdict about his knowledge I refer of almost overwhelming persuasiveness - of his
course only to the position he finds himself in by reflections. That alone is something that needs
the end of his First Meditation. Having at that accounting for. I cannot possibly do justice to all
point discovered and stated the problem of the reasonable reactions to them here. In the rest of
external world, Descartes goes on in the rest of his this chapter I want to concentrate on deepening
Meditations to try to solve it, and by the end of the and strengthening the problem and trying to
Sixth Meditation he thinks he has explained how locate more precisely the source of its power.
he knows almost all those familiar things he began There are at least three distinct questions that
by putting in question. So when I ascribe to could be pressed. Is the possibility that Descartes
Descartes the view that we can know nothing might be dreaming really a threat to his knowl-
about the world around us I do not mean to sug- edge of the world around him? Is he right in
gest that that is his final and considered view; it is thinking that he must know that he is not dream-
nothing more than a conclusion he feels almost ing if he is to know something about the world
inevitably driven to at the early stages of his around him? And is he right in his "discovery"
reflections. But those are the only stages of his that he can never know that he is not dreaming? If
thinking I am interested in here. That is where the Descartes were wrong on any of these points it
philosophical problem of our knowledge of the might be possible to avoid the problem and per-
external world gets posed, and before we can con- haps even to explain without difficulty how we
sider possible solutions we must be sure we know things about the world around us.
understand exactly what the problem is. On the first question, it certainly seems right
I have described it as that of showing or to say that if Descartes were dreaming that he is
explaining how knowledge of the world around sitting by the fire with a piece of paper in his hand
us is possible by means of the senses. It is impor- he would not then know that he is sitting by the
tant to keep in mind that that demand for an fire with a piece of paper in his hand. When you
explanation arises in the face of a challenge or dream that something is going on in the world
apparent obstacle to our knowledge of the world. around you you do not thereby know that it is.
The possibility that he is dreaming is seen as an Most often, of course, what we dream is not even
obstacle to Descartes's knowing that he is sitting true; no one is actually chasing us when we are
by the fire, and it must be explained how that lying asleep in bed dreaming, nor are we actually
obstacle can either be avoided or overcome. It climbing stairs. But although usually what we
must be shown or explained how it is possible for dream is not really so, that is not the real reason
us to know things about the world, given that the for our lack of knowledge. Even if Descartes were
sense-experiences we get are compatible with our in fact sitting by the fire and actually had a piece
merely dreaming. Explaining how something is of paper in his hand at the very time he was
nevertheless possible, despite what looks like an dreaming that he is sitting by the fire with a piece
obstacle to it, requires more than showing merely of paper in his hand, he would not thereby know
that there is no impossibility involved in the he was sitting there with that paper. He would be
thing - that it is consistent with the principles of like a certain Duke of Devonshire who, according
logic and the laws of nature and so in that sense to G. E. Moore, once dreamt he was speaking in
could exist. The mere possibility of the state of the House of Lords and woke up to find that he
affairs is not enough to settle the question of how was speaking in the House of Lords. 4 What he
our knowledge of the world is possible; we must was dreaming was in fact so. But even if what you
understand how the apparent obstacle is to be are dreaming is in fact so you do not thereby
got round. know that it is. Even if we allow that when you are
Descartes's reasoning can be examined and dreaming that something is so you can be said, at
criticized at many different points, and has been least for the time being, to think or to believe that
closely scrutinized by many philosophers for cen- it is so, there is still no real connection between
turies. It has also been accepted by many, perhaps your thinking or believing what you do and its
by more than would admit or even realize that being so. At best you have a thought or a belief

which just happens to be true, but that is no might be a physicist who knows a great deal about
more than coincidence and not knowledge. So the way things are which the child does not know.
Descartes's first step relies on what seems to be an If the man also dreams that things are that way he
undeniable fact about dreams: if you are dream- can once again be said to be dreaming that some-
ing that something is so you do not thereby know thing is so and also to know that it is so. There is
that it is so. therefore no incompatibility between dreaming
This bald claim needs to be qualified and more and knowing. That is true, but I do not think it
carefully explained, but I do not think that will affects Descartes's argument. He is led to consider
diminish the force of the point for Descartes's how he knows he is not dreaming at the moment
purposes. Sometimes what is going on in the by reflecting on how he knows at that moment
world around us has an effect on what we dream; that he is sitting by the fire with a piece of paper
for example, a banging shutter might actually in his hand. If he knows that at all, he thinks, he
cause me to dream, among other things, that a knows it on the basis of the senses. But he real-
shutter is banging. If my environment affects me izes that his having the sensory experiences he is
in that way, and if in dreams I can be said to think now having is compatible with his merely dream-
or believe that something is so, would I not in that ing that he is sitting by the fire with a piece of
case know that a shutter is banging? It seems to paper in his hand. So he does not know on the
me that I would not, but I confess it is difficult to basis of the sensory experiences he is having at
say exactly why I think so. That is probably the moment that he is sitting by the fire. Nor, of
because it is difficult to say exactly what is required course, did the man in my examples know the
for knowledge. We use the term "know" confi- things he was said to know on the basis of the sen-
dently, we quite easily distinguish cases of knowl- sory experiences he was having at that moment.
edge from cases of its absence, but we are not He knew certain things to be so, and he was
always in a position to state what we are going on dreaming those things to be so, but in dreaming
in applying or withholding the term in the ways them he did not thereby know them to be so.
we do. I think that in the case of the banging shut- But as long as we allow that the sleeping man
ter it would not be knowledge because I would be does know certain things about the world around
dreaming, I would not even be awake. At least it him, even if he does not know them on the basis
can be said, I think, that even if Descartes's sitting of the very dreams he is having at the moment,
by the fire with a piece of paper in his hand (like isn't that enough to show that Descartes must
the banging shutter) is what in fact causes him to nevertheless be wrong in his conclusion that no
dream that he is sitting by the fire with a piece of one can know anything about the world around
paper in his hand, that is still no help to him in him? No. It shows at most that we were hasty or
coming to know what is going on in the world were ignoring Descartes's conclusion in conced-
around him. He realizes that he could be dream- ing that someone could know something about
ing that he is sitting by the fire even if he is in fact the world around him. If Descartes's reasoning is
sitting there, and that is the possibility he finds he correct the dreaming physicist, even when he is
has to rule out. awake, does not really know any of the things we
I have said that if you are dreaming that some- were uncritically crediting him with knowing
thing is so you do not thereby know that it is so, about the way things are - or at least he does not
and it might seem as if that is not always true. know them on the basis of the senses. In order to
Suppose a man and a child are both sleeping. I say know them on the basis of the senses there would
of the child that it is so young it does not know have to have been at least some time at which he
what seven times nine is, whereas the grown man knew something about what was going on around
does know that. If the man happens at that very him at that time. But if Descartes is right he could
moment to be dreaming that seven times nine is not have known any such thing unless he had
sixty-three (perhaps he is dreaming that he is established that he was not dreaming at that time;
computing his income tax), then he is a man who and according to Descartes he could never estab-
is dreaming that something is so and also knows lish that. So the fact about dreams that Descartes
that it is so. The same kind of thing is possible relies on - that one who dreams that something is
for knowledge about the world around him. He so does not thereby know that it is so - is enough

to yield his conclusion if the other steps of his it is very plausible to say that there is nothing we
reasoning are correct. could not dream about, nothing that could be the
When he first introduces the possibility that case that we could not dream to be the case. I say
he might be dreaming Descartes seems to be rely- it is very plausible; of course I cannot prove it to
ing on some knowledge about how things are or be true. But even if it is not true with complete
were in the world around him. He says "I remind generality, we must surely grant that it is possible
myself that on many occasions I have in sleep been to dream that one is sitting by a fire with a piece
deceived by similar illusions", so he seems to be of paper in one's hand, and possible to dream of
relying on some knowledge to the effect that he countless other equally obvious and equally mun-
has actually dreamt in the past and that he remem- dane states of affairs as well, and those possibili-
bers having been "deceived" by those dreams. That ties are what Descartes sees as threatening to his
is more than he actually needs for his reflections knowledge of the world around him.
about knowledge to have the force he thinks they There seems little hope, then, of objecting that
have. He does not need to support his judgement it is simply not possible for Descartes to dream
that he has actually dreamt in the past. The only that he is sitting by the fire with a piece of paper
thought he needs is that it is now possible for him in his hand. Nor is it any more promising to say
to be dreaming that he is sitting by the fire, and that even if he were dreaming it would not follow
that if that possibility were realized he would not that he did not know that he was sitting there.
know that he is sitting by the fire. Of course it was I think both those steps or assumptions of
no doubt true that Descartes had dreamt in the Descartes's reasoning are perfectly correct, and
past and that his knowledge that he had done so further defence of them at this stage is unneces-
was partly what he was going on in acknowledg- sary. If his argument and the problem to which it
ing the possibility of his dreaming on this partic- gives rise are to be avoided, it might seem that the
ular occasion. But neither the fact of past dreams best hope is therefore to accept his challenge and
nor knowledge of their actual occurrence would show that it can be met. That would be in effect to
seem to be strictly required in order to grant what argue that Descartes's alleged "discovery" is no
Descartes relies on - the possibility of dreaming, discovery at all: we can sometimes know that we
and the absence of knowledge if that possibility are not dreaming.
were realized. The thought that he might be This can easily seem to be the most straight-
dreaming that he is sitting by the fire with a piece forward and most promising strategy. It allows
of paper in his hand, and the fact that if he were that Descartes is right in thinking that knowing
he wouldn't know he was sitting there, is what that one is not dreaming is a condition of know-
gives Descartes pause. That would worry him in ing something about the world around us, but
the way it does even if he had never actually had wrong in thinking that that condition can never
any dreams exactly like it in the past - if he had be met. And that certainly seems plausible. Surely
never dreamt about fires and pieces of paper at it is not impossible for me to know that I am not
all. In fact, I think he need never have actually dreaming? Isn't that something I often know, and
dreamt of anything before, and certainly needn't isn't it something I can sometimes find out if the
know that he ever has, in order to be worried in question arises? If it is, then the fact that I must
the way he is by the thought that he might be know that I am not dreaming if I am to know
dreaming now. anything about the world around me will be no
The fact that the possibility of dreaming is all threat to my knowledge of the world.
Descartes needs to appeal to brings out another However obvious and undeniable it might be
truth about dreams that his argument depends that we often do know that we are not dreaming,
on - that anything that can be going on or that I think this straightforward response to Descartes's
one can experience in one's waking life can also challenge is a total failure. In calling it straightfor-
be dreamt about. This again is only a statement of ward I mean that it accepts Descartes's conditions
possibility - no sensible person would suggest for knowledge of the world and tries to show that
that we do at some time dream of everything that they can be fulfilled. That is what I think cannot
actually happens to us, or that everything we be done. To put the same point in another way:
dream about does in fact happen sometime. But I think Descartes would be perfectly correct in

saying "there are no certain indications by which of knowing something about the world around us
we may clearly distinguish wakefulness from on the basis of the senses. Since he thinks the pos-
sleep", and so we could never tell we are not sibility of his dreaming must be ruled out in the
dreaming, ifhe were also right that knowing that case he considers, and the case he considers is
one is not dreaming is a condition of knowing regarded as typical and without special character-
something about the world around us. That is istics of its own, he thinks that the possibility that
why I think one cannot accept that condition and he is dreaming must be ruled out in every case of
then go on to establish that one is not dreaming. knowing something about the world by means of
I do not mean to be saying simply that Descartes the senses.
is right - that we can never know that we are not If that really is a condition of knowing some-
dreaming. But I do want to argue that either we thing about the world, I think it can be shown
can never know that we are not dreaming or else that Descartes is right in holding that it can
what Descartes says is a condition of knowing never be fulfilled. That is what the straightfor-
things about the world is not really a condition in ward response denies, and that is why I think
general of knowing things about the world. The that response must be wrong. We cannot accept
straightforward strategy denies both alternatives. the terms of Descartes's challenge and then hope
I will try to explain why I think we must accept to meet it.
one alternative or the other. Suppose Descartes tries to determine that he is
When Descartes asks himself how he knows not dreaming in order to fulfil what he sees as a
that he is sitting by the fire with a piece of paper necessary condition of knowing that he is sitting
in his hand why does he immediately go on to ask by the fire with a piece of paper in his hand. How
himself how he knows he is not dreaming that he is he to proceed? He realizes that his seeing his
is sitting by the fire with a piece of paper in his hand and seeing and feeling a piece of paper
hand? I have suggested that it is because he recog- before him and feeling the warmth of the fire - in
nizes that ifhe were dreaming he would not know fact his getting all the sensory experiences or all
on the basis of his senses at the moment that he is the sensory information he is then getting - is
sitting there, and so he thinks he must know that something that could be happening even if he
that possibility does not obtain if he is to know were dreaming. To establish that he is not dream-
that he is in fact sitting there. But this particular ing he would therefore need something more
example was chosen, not for any peculiarities it than just those experiences or that information
might be thought to possess, but because it could alone. He would also need to know whether those
be taken as typical of the best position we can experiences and that information are reliable, not
ever be in for coming to know things about the merely dreamt. If he could find some operation
world around us on the basis of the senses. What or test, or if he could find some circumstance or
is true of this case that is relevant to Descartes's state of affairs, that indicated to him that he was
investigation of knowledge is supposed to be true not dreaming, perhaps he could then fulfil the
of all cases of knowledge of the world by means condition - he could know that he was not dream-
of the senses; that is why the verdict arrived at ing. But how could a test or a circumstance or a
here can be taken to be true of our sensory state of affairs indicate to him that he is not dream-
knowledge generally. But what Descartes thinks ing if a condition of knowing anything about the
is true of this particular case of sensory knowl- world is that he know he is not dreaming? It could
edge of the world is that he must know he is not not. He could never fulfil the condition.
dreaming if he is to know that he is sitting by the Let us suppose that there is in fact some test
fire with a piece of paper in his hand. That is which a person can perform successfully only if
required, not because of any peculiarities of this he is not dreaming, or some circumstance or state
particular case, but presumably because, accord- of affairs which obtains only if that person is not
ing to Descartes, it is a necessary condition of any dreaming. Of course for that test or state of affairs
case - even a best possible case - of knowledge of to be of any use to him Descartes would have to
the world by means of the senses. That is why I know of it. He would have to know that there is
ascribed to Descartes the quite general thesis that such a test or that there is a state of affairs that
knowing that one is not dreaming is a condition shows that he is not dreaming; without such

information he would be no better off for telling successfully or that he established that the state of
that he is not dreaming than he would be if there affairs obtains. How could that in turn be known?
were no such test or state of affairs at all. To have Obviously the particular test or state of affairs
acquired that information he would at some already in question cannot serve as a guarantee of
time have to have known more than just some- its own authenticity, since it might have been
thing about the course of his sensory experience, merely dreamt, so some further test or state of
since the connection between the performance of affairs would be needed to indicate that the origi-
a certain test, or between a certain state of affairs, nal test was actually performed and not merely
and someone's not dreaming is not itself just a dreamt, or that the state of affairs in question was
fact about the course of that person's sensory actually ascertained to obtain and not just dreamt
experience; it is a fact about the world beyond his to obtain. But this further test or state of affairs is
sensory experiences. Now strictly speaking if it is subject to the same general condition in turn.
a condition of knowing anything about the world Every piece of knowledge that goes beyond one's
beyond one's sensory experiences that one know sensory experiences requires that one know one is
that one is not dreaming, there is an obvious not dreaming. This second test or state of affairs
obstacle to Descartes's ever having got the infor- will therefore be of use only if Descartes knows
mation he needs about that test or state of affairs. that he is not merely dreaming that he is perform-
He would have to have known at some time that ing or ascertaining it, since merely to dream that
he was not dreaming in order to get the informa- he had established the authenticity of the first test
tion he needs to tell at any time that he is not is not to have established it. And so on. At no
dreaming - and that cannot be done. point can he find a test for not dreaming which he
But suppose we forget about this difficulty can know has been successfully performed or a
and concede that Descartes does indeed know state of affairs correlated with not dreaming
(somehow) that there is a test or circumstance or which he can know obtains. He can therefore
state of affairs that unfailingly indicates that he is never fulfil what Descartes says is a necessary
not dreaming. Still, there is an obstacle to his ever condition of knowing something about the world
using that test or state of affairs to tell that he is around him. He can never know that he is not
not dreaming and thereby fulfilling the condition dreaming.
for knowledge of the world. The test would have I must emphasize that this conclusion is
to be something he could know he had performed reached only on the assumption that it is a condi-
successfully, the state of affairs would have to be tion of knowing anything about the world around
something he could know obtains. If he com- us on the basis of the senses that we know we are
pletely unwittingly happened to perform the test, not dreaming that the thing is so. I think it is his
or if the state of affairs happened to obtain but he acceptance of that condition that leads Descartes
didn't know that it did, he would be in no better to "see so manifestly that there are no certain
position for telling whether he was dreaming than indications by which we may clearly distinguish
he would be if he had done nothing or did not wakefulness from sleep". And I think Descartes is
even know that there was such a test. But how is absolutely right to draw that conclusion, given
he to know that the test has been performed suc- what he thinks is a condition of knowledge of the
cessfully or that the state of affairs in question world. But all I have argued on Descartes's behalf
does in fact obtain? Anything one can experience (he never spells out his reasoning) is that we
in one's walking life can also be dreamt about; it is cannot both accept that condition of knowledge
possible to dream that one has performed a cer- and hope to fulfil it, as the straightforward
tain test or dream that one has established that a response hopes to do. And of course if one of the
certain state of affairs obtains. And, as we have necessary conditions of knowledge of the world
seen, to dream that something about the world can never be fulfilled, knowledge of the world
around you is so is not thereby to know that it is so. around us will be impossible.
In order to know that his test has been performed I think we have now located Descartes's reason
or that the state of affairs in question obtains for his negative verdict about sensory knowledge
Descartes would therefore have to establish that he in general. If we agree that he must know that he
is not merely dreaming that he performed the test is not dreaming if he is to know in his particular

case that he is sitting by the fire with a piece of circumstances when nothing very important turns
paper in his hand, we must also agree that we can on the outcome, we cannot know a particular thing
know nothing about the world around us. unless we have ruled out certain possibilities that
Once we recognize that the condition Descartes we recognize are incompatible with our knowing
takes as necessary can never be fulfilled if he is that thing.
right in thinking it is indeed necessary, we are Suppose that on looking out the window I
naturally led to the question whether Descartes is announce casually that there is a goldfinch in the
right. Is it really a condition of knowing some- garden. If I am asked how I know it is a goldfinch
thing about the world that one know one is not and I reply that it is yellow, we all recognize that
dreaming? That is the second of the three ques- in the normal case that is not enough for knowl-
tions I distinguished. It is the one that has received edge. "For all you've said so far," it might be
the least attention. In asking it now I do not mean replied, "the thing could be a canary, so how do
to be going back on something I said earlier was you know it's a goldfinch?" A certain possibility
undeniably true, viz., that if one is dreaming that compatible with everything I have said so far has
something about the world is so one does not been raised, and if what I have said so far is all I
thereby know that it is so. That still seems to me have got to go on and I don't know that the thing
undeniable, but it is not the same as Descartes's in the garden is not a canary, then I do not know
assumption that one must know that one is not that there is a goldfinch in the garden. I must be
dreaming if one is to know something about the able to rule out the possibility that it is a canary if
world. The undeniable truth says only that you I am to know that it is a goldfinch. Anyone who
lack knowledge if you are dreaming; Descartes speaks about knowledge and understands what
says that you lack knowledge if you don't know others say about it will recognize this fact or con-
that you are not dreaming. Only with the stronger dition in particular cases.
assumption can his sceptical conclusion be In this example what is said to be possible is
reached. something incompatible with the truth of what I
Is that assumption true? In so far as we find claim to know - if that bird were a canary it would
Descartes's reasoning convincing, or even plausi- not be a goldfinch in the garden, but a canary.
ble, I think it is because we too on reflection find What I believe in believing it is a goldfinch would
that it is true. I said that not much attention had be false. But that is not the only way a possibility
been paid to that particular part of Descartes's can work against my knowledge. If I come to sus-
reasoning, and I think that too is because, as he pect that all the witnesses have conspired and
presents it, the step seems perfectly convincing made up a story about the man's being in
and so only other parts of the argument appear Cleveland that night, for example, and their testi-
vulnerable. Why is that so? Is it because Descartes's mony is all I have got to go on in believing that he
assumption is indeed true? Is there anything we was in Cleveland, I might find that I no longer
can do that would help us determine whether it is know whether he was there or not until I have
true or not? The question is important because I some reason to rule out my suspicion. If their tes-
have argued so far that if it is true we can never timony were all invented I would not know that
know anything about the world around us on the the man was in Cleveland. But strictly speaking
basis of the senses, and philosophical scepticism his being in Cleveland is not incompatible with
about the external world is correct. We would their making up a story saying he was. They might
have to find that conclusion as convincing or as have invented a story to protect him, whereas in
plausible as we find the assumption from which it fact, unknown to them, he was there all the time.
is derived. Such a complicated plot is not necessary to bring
Given our original favourable response to out the point; Moore's Duke of Devonshire is
Descartes's reasoning, then, it can scarcely be enough. From the fact that he was dreaming that he
denied that what I have called his assumption or was speaking in the House ofLords it did not follow
condition seems perfectly natural to insist on. that he was not speaking in the House of Lords. In
Perhaps it seems like nothing more than an instance fact he was. The possibility of dreaming - which
of a familiar commonplace about knowledge. We was actual in that case - did not imply the falsity
are all aware that, even in the most ordinary of what was believed. A possible deficiency in the

basis of my belief can interfere with my knowl- conditions? I have already said that it seems unde-
edge without itself rendering false the very thing I niable that it fulfils the first. If he were dreaming
believe. A hallucinogenic drug might cause me to Descartes would not know what he claims to
see my bed covered with a huge pile of leaves, for know. Someone who is dreaming does not thereby
example. s Having taken that drug, I will know the know anything about the world around him even
actual state of my bed only if I know that what I if the world around him happens to be just the
see is not just the effect of the drug; I must be able way he dreams or believes it to be. So his dream-
to rule out the possibility that I am hallucinating ing is incompatible with his knowing. But does it
the bed and the leaves. But however improbable it fulfil the second condition? Is it a possibility
might be that my bed is actually covered with which must be known not to obtain if Descartes
leaves, its not being covered with leaves does not is to know that he is sitting by the fire with a piece
follow from the fact that I am hallucinating that it of paper in his hand? I think it is difficult simply
is. What I am hallucinating could nevertheless be to deny that it is. The evident force of Descartes's
(unknown to me) true. But a goldfinch simply reasoning when we first encounter it is enough to
could not be a canary. So although there are two show that it certainly strikes us as a relevant pos-
different ways in which a certain possibility can sibility, as something that he should know not to
threaten my knowledge, it remains true that there obtain if he is to know where he is and what is
are always certain possibilities which must be happening around him.
known not to obtain if I am to know what I claim When that possibility strikes us as obviously
to know. relevant in Descartes's investigation we might
I think these are just familiar facts about come to think that it is because of a simple and
human knowledge, something we all recognize obvious fact about knowledge. In the case of the
and abide by in our thought and talk about know- goldfinch we immediately recognize that I must
ing things. We know what would be a valid chal- know that it is not a canary if I am to know it is a
lenge to a claim to know something, and we can goldfinch. And it is very natural to think that that
recognize the relevance and force of objections is simply because its being a canary is incompat-
made to our claims to know. The question before ible with its being a goldfinch. If it were a canary
us is to what extent Descartes's investigation of it would not be a goldfinch, and I would there-
his knowledge that he is sitting by the fire with a fore be wrong in saying that it is; so if I am to
piece of paper in his hand follows these recog- know it is a goldfinch I must rule out the possi-
nized everyday procedures for assessing claims to bility that it is a canary. The idea is that the two
know. If it does follow them faithfully, and yet conditions I distinguished in the previous para-
leads to the conclusion that he cannot know graph are not really separate after all. As soon as
where he is or what is happening around him, we we see that a certain possibility is incompatible
seem forced to accept his negative conclusion with our knowing such-and-such, it is suggested,
about knowledge in general just as we are forced we immediately recognize that it is a possibility
to accept the conclusion that I do not know it is a that must be known not to obtain if we are to
goldfinch or do not know the witness was in know the such-and-such in question. We see that
Cleveland because I cannot rule out the possibili- the dream-possibility satisfies that first condition
ties which must be ruled out if! am to know such in Descartes's case (if he were dreaming, he
things. Is Descartes's introduction of the possibil- wouldn't know), and that is why, according to
ity that he might be dreaming just like the intro- this suggestion, we immediately see that it is rel-
duction of the possibility that it might be a canary evant and must be ruled out. Something we all
in the garden or that the alibi might be contrived recognize about knowledge is what is said to
or that it might be a hallucination of my bed cov- make that obvious to us.
ered with leaves? But is the "simple and obvious fact about
Those possibilities were all such that if they knowledge" appealed to in this explanation really
obtained I did not know what I claimed to know, something that is true of human knowledge even
and they had to be known not to obtain in order in the most ordinary circumstances? What exactly
for the original knowledge-claim to be true. Does is the "fact" in question supposed to be? I have
Descartes's dream-possibility fulfil both of those described it so far, as applied to the case of the

goldfinch, as the fact that if I know something p all those things that I know to be incompatible
(it's a goldfinch) I must know the falsity of all with p. Since I claim to know that the bird is a
those things incompatible with p (e.g., it's a goldfinch, and I know that its being a goldfinch
canary). If there were one of those things that I implies that it is not a canary, I must for that
did not know to be false, and it were in fact true, I reason know that it is not a canary if my original
would not know that p, since in that case some- claim is true. In claiming to know it is a goldfinch
thing incompatible with p would be true and so p I was, so to speak, committing myself to knowing
would not be true. But to say that I must know that it is not a canary, and I must honour my
that all those things incompatible with p are false commitments.
is the same as saying that I must know that truth This requirement as it stands, even if it does
of all those things that must be true if p is true. explain why I must know that the bird is not a
And it is extremely implausible to say that that is canary, does not account for the relevance of the
a "simple and obvious fact" we all recognize about other sorts of possibilities I have mentioned. The
human knowledge. reason in the goldfinch case was said to be that I
The difficulty is that there are no determinate know that its being a canary is incompatible with
limits to the number of things that follow from its being a goldfinch. But that will not explain
the things I already know. But it cannot be said why I must rule out the possibility that the wit-
that I now know all those indeterminately many nesses have invented a story about the man's
things, although they all must be true if the things being in Cleveland, or the possibility that I am
that I already know are true. Even granting that I hallucinating my bed covered with a pile ofleaves.
now know a great deal about a lot of different Nor will it explain why Descartes must rule out
things, my knowledge obviously does not extend the possibility that he is dreaming. What I claimed
to everything that follows from what I now know. to know in the first case is that the man was in
If it did, mathematics, to take only one example, Cleveland that night. But, as we saw earlier, it is
would be a great deal easier than it is - or else not a consequence of his being in Cleveland that
impossibly difficult. In knowing the truth of the no one will invent a story to the effect that he was
simple axioms of number theory, for example, I in Cleveland; they might mistakenly believe he
would thereby know the truth of everything that was not there and then tell what they think is a lie.
follows from them; every theorem of number Nor is it a consequence of my bed's being covered
theory would already be known. Or, taking the with leaves that I am not hallucinating that it is.
pessimistic side, since obviously no one does But we recognize that in order to know in those
know all the theorems of number theory, it would cases I nevertheless had to rule out those possi-
follow that no one even knows that those simple bilities. Similarly, as the Duke of Devonshire
axioms are true. reminds us, it is not a consequence of Descartes's
It is absurd to say that we enjoy or require such sitting by the fire with a piece of paper in his hand
virtual omniscience, so it is more plausible to that he is not dreaming that he is. So if it is obvi-
hold that the "simple and obvious fact" we all rec- ous to us that Descartes must know that he is not
ognize about knowledge is the weaker require- dreaming if he is to know that he is sitting by the
ment that we must know the falsity of all those fire, it cannot be simply because the possibility in
things that we know to be incompatible with the question is known to be incompatible with what
things we know. I know that a bird's being a he claims to know. It is not.
canary is incompatible with its being a goldfinch; If there is some "simple and obvious fact about
that is not some farflung, unknown consequence knowledge" that we recognize and rely on in
of its being a goldfinch, but something that responding to Descartes's reasoning it must there-
anyone would know who knew anything about fore be more complicated than what has been
goldfinches at all. And the idea is that that is why suggested so far. Reflecting even on the uncontro-
I must know that it is not a canary if I am to know versial everyday examples alone can easily lead us
that it is a goldfinch. Perhaps, in order to know to suppose that it is something like this: if some-
something, p, I do not need to know the falsity of body knows something, p, he must know the fal-
all those things that are incompatible with p, but sity of all those things incompatible with his
it can seem that at least I must know the falsity of knowing that p (or perhaps all those things he

knows to be incompatible with his knowing that day and scientific life. We have no notion of
pl. I will not speculate further on the qualifica- knowledge other than what is embodied in those
tions or emendations needed to make the princi- procedures and practices. So if that requirement
ple less implausible. The question now is whether is a "fact" of our ordinary conception of knowl-
it is our adherence to any such principle or edge we will have to accept the conclusion that no
requirement that is responsible for our recogni- one knows anything about the world around us.
tion that the possibility that the bird is a canary or
the possibility that the witnesses made up a story I now want to say a few more words about the
must be known not to obtain if I am to know the position we would all be in if Descartes's conclu-
things I said I knew in those cases. What exactly sion as he understands it were correct. I described
are the procedures or standards we follow in the him earlier as having lost the whole world, as
most ordinary, humdrum cases of putative knowing at most what he is experiencing or how
knowledge? Reflection on the source of Descartes's things appear to him, but knowing nothing about
sceptical reasoning has led to difficulties in how things really are in the world around him. To
describing and therefore in understanding even show how anyone in that position could come to
the most familiar procedures we follow in every- know anything about the world around him is
day life. That is one of the rewards of a study of what I am calling the problem of our knowledge
philosophical scepticism. of the external world, and it is worth dwelling for
The main difficulty in understanding our a moment on just how difficult a problem that
ordinary procedures is that no principle like those turns out to be if it has been properly raised.
I have mentioned could possibly describe the way If we are in the predicament Descartes finds
we proceed in everyday life. Or, to put it less dog- himself in at the end of his First Meditation we
matically, if our adherence to some such require- cannot tell by means of the senses whether we are
ment were responsible for our reactions in those dreaming or not; all the sensory experiences we
ordinary cases, Descartes would be perfectly cor- are having are compatible with our merely dream-
rect, and philosophical scepticism about the ing of a world around us while that world is in
external world would be true. Nobody would fact very different from the way we take it to be.
know anything about the world around us. If, in Our knowledge is in that way confined to our
order to know something, we must rule out a sensory experiences. There seems to be no way of
possibility which is known to be incompatible going beyond them to know that the world
with our knowing it, Descartes is perfectly right around us really is this way rather than that. Of
to insist that he must know that he is not dream- course we might have very strongly held beliefs
ing if he is to know that he is sitting by the fire about the way things are. We might even be unable
with a piece of paper in his hand. He knows his to get rid of the conviction that we are sitting by
dreaming is incompatible with his knowing. the fire holding a piece of paper, for example. But
I have already argued that if he is right in insisting if we acknowledge that our sensory experiences
that that condition must be fulfilled for knowl- are all we ever have to go on in gaining knowledge
edge of the world around us he is also right in about the world, and we acknowledge, as we must,
concluding that it can never be fulfilled; fulfilling that given our experiences as they are we could
it would require knowledge which itself would be nevertheless be simply dreaming of sitting by the
possible only if the condition were fulfilled. So fire, we must concede that we do not know that
both steps of Descartes's reasoning would be valid we are sitting by the fire. Of course, we are in no
and his conclusion would be true. position to claim the opposite either. We cannot
That conclusion can be avoided, it seems to conclude that we are not sitting by the fire; we
me, only if we can find some way to avoid the simply cannot tell which is the case. Our sensory
requirement that we must know we are not experience gives us no basis for believing one
dreaming if we are to know anything about the thing about the world around us rather than its
world around us. But that requirement cannot be opposite, but our sensory experience is all we
avoided if it is nothing more than an instance of a have got to go on. So whatever unshakeable con-
general procedure we recognize and insist on in viction we might nevertheless retain, that convic-
making and assessing knowledge-claims in every- tion cannot be knowledge. Even if we are in fact

holding a piece of paper by the fire, so that what on in the world outside. For all he can know, what-
we are convinced of is in fact true, that true con- ever is producing the patterns he can see on the
viction is still not knowledge. The world around screens in front of him might be something other
us, whatever it might be like, is in that way beyond than well-function cameras directed on to the
our grasp. We can know nothing of how it is, no passing show outside the room. The victim might
matter what convictions, beliefs, or opinions we switch on more of the sets in the room to try to get
continue, perhaps inevitably, to hold about it. more information, and he might find that some of
What can we know in such a predicament? We the sets show events exactly similar or coherently
can perhaps know what sensory experiences we related to those already visible on the screens he
are having, or how things seem to us to be. At can see. But all those pictures will be no help to
least that much of our knowledge will not be him without some independent information,
threatened by the kind of attack Descartes makes some knowledge which does not come to him
on our knowledge of the world beyond our expe- from the pictures themselves, about how the pic-
riences. What we can know turns out to be a great tures he does see before him are connected with
deal less than we thought we knew before engag- what is going on outside the room. The problem
ing in that assessment of our knowledge. Our of the external world is the problem of finding
position is much more restricted, much poorer, out, or knowing how we could find out, about the
than we had originally supposed. We are confined world around us if we were in that sort of predica-
at best to what Descartes calls "ideas" of things ment. It is perhaps enough simply to put the
around us, representations of things or states of problem this way to convince us that it can never
affairs which, for all we can know, might or might be given a satisfactory solution.
not have something corresponding to them in But putting the problem this way, or only this
reality. We are in a sense imprisoned within those way, has its drawbacks. For one thing, it encour-
representations, at least with respect to our ages a facile dismissive response; not a solution to
knowledge. Any attempt to go beyond them to try the problem as posed, but a rejection of it. I do
and tell whether the world really is as they repre- not mean that we should not find a way to reject
sent it to be can yield only more representations, the problem - I think that is our only hope - but
more deliverances of sense experience which this particular response, I believe, is wrong, or
themselves are compatible with reality's being at the very least premature. It is derived almost
very different from the way we take it to be on the entirely from the perhaps overly dramatic descrip-
basis of our sensory experiences. There is a gap, tion of the predicament I have just given.
then, between the most that we can ever find out I have described Descartes's sceptical conclu-
on the basis of our sensory experience and the sion as implying that we are permanently sealed
way things really are. In knowing the one we do off from a world we can never reach. We are
not thereby know the other. restricted to the passing show on the veil of per-
This can seem to leave us in the position of ception, with no possibility of extending our
finding a barrier between ourselves and the world knowledge to the world beyond. We are confined
around us. There would then be a veil of sensory to appearances we can never know to match or to
experiences or sensory objects which we could not deviate from the imperceptible reality that is for-
penetrate but which would be no reliable guide to ever denied us. This way of putting it naturally
the world beyond the veil. If we were in such a encourages us to minimize the seriousness of the
position, I think it is quite clear that we could not predicament, to try to settle for what is undenia-
know what is going on beyond the veil. There bly available to us, or perhaps even to argue that
would be no possibility of our getting reliable sen- nothing that concerns us or makes human life
sory information about the world beyond the veil; worthwhile has been left out.
all such reports would simply be more representa- If an imperceptible "reality", as it is called on
tions, further ingredients of the evermore-com- this picture, is forever inaccessible to us, what
plicated veil. We would know nothing but the veil concern can it be of ours? How can something we
itself. We would be in the position of someone can have no contact with, something from which
waking up to find himselflocked in a room full of we are permanently sealed off, even make sense to
television sets and trying to find out what is going us at all? Why should we be distressed by an

alleged limitation of our knowledge if it is not reason. On the contrary; so far we have every
even possible for the "limitation" to be overcome? reason to think that Descartes has revealed the
If it makes no sense to aspire to anything beyond impossibility of the very knowledge of the world
what is possible for us, it will seem that we should that we are most interested in and which we began
give no further thought to this allegedly imper- by thinking we possess or can easily acquire. In
ceptible "reality': Our sensory experiences, past, any case, that would be the only conclusion to
present, and future, will then be thought to be all draw if Descartes's investigation does indeed par-
we are or should be concerned with, and the idea allel the ordinary kinds of assessments we make
of a "reality" lying beyond them necessarily out of of our knowledge in everyday life.
our reach will seem like nothing more than a phi- We saw that I can ask what I really know about
losopher's invention. What a sceptical philoso- the common cold, or whether I really know that
pher would be denying us would then be nothing the witness was in Cleveland on the night in ques-
we could have ordinary commerce with or inter- tion, and that I can go on to discover that I do not
est in anyway. Nothing distressing about our really know what I thought I knew. In such ordi-
ordinary position in the familiar world would nary cases there is no suggestion that what I have
have been revealed by a philosopher who simply discovered is that I lack some special, esoteric
invents or constructs something he calls "reality" thing called "real knowledge", or that I lack knowl-
or "the external world" and then demonstrates edge of some exotic, hitherto-unheard-of domain
that we can have no access to it. That would show called "reality". If I ask what I know about the
nothing wrong with the everyday sensory knowl- common cold, and I come to realize that I do not
edge we seek and think we find in ordinary life really know whether it can be caused by sitting in
and in scientific laboratories, nor would it show a draught or not, the kind of knowledge I discover
that our relation to the ordinary reality that con- I lack is precisely what I was asking about or
cerns us is different from what we originally taking it for granted I had at the outset. I do not
thought it to be. conclude with a shrug that it no longer matters
I think this reaction to the picture of our being because what I now find I lack is only knowledge
somehow imprisoned behind the veil of our own about a special domain called "reality" that was
sensory experiences is very natural and immedi- somehow invented only to serve as the inaccessi-
ately appealing. It is natural and perhaps always ble realm of something called "real knowledge".
advisable for a prisoner to try to make the best of I simply conclude that I don't really know whether
the restricted life behind bars. But however much colds are caused by sitting in draughts or not. If I
more bearable it makes the prospect of life- say in a jury-room on Monday that we can elimi-
imprisonment, it should not lead him to deny the nate the suspect because we know he was in
greater desirability, let alone the existence, of life Cleveland that night, and I then discover by
outside. In so far as the comfort of this response reflection on Tuesday that I don't really know he
to philosophical scepticism depends on such a was in Cleveland that night, what I am denying I
denial it is at the very least premature and is prob- have on Tuesday is the very thing I said on Monday
ably based on misunderstanding. It depends on a that I had.
particular diagnosis or account of how and why There is no suggestion in these and countless
the philosophical argument succeeds in reaching similar everyday cases that somehow in the course
its conclusion. The idea is that the "conclusion" is of our reflections on whether and how we know
reached only by contrivance. The inaccessible something we are inevitably led to change or ele-
"reality" denied to us is said to be simply an arte- vate our conception of knowledge into something
fact of the philosopher's investigation and not else called "real knowledge" which we showed no
something that otherwise should concern us. signs of being interested in at the beginning. Nor
That is partly a claim about how the philosophi- is it plausible to suggest that our ordinary assess-
cal investigation of knowledge works; as such, it ments of knowledge somehow lead us to postu-
needs to be explained and argued for. We can late a "reality" that is simply an artefact of our
draw no consolation from it until we have some inquiries about our knowledge. When we ask
reason to think it might be an accurate account of whether we really know something we are simply
what the philosopher does. So far we have no such asking whether we know that thing. The "really"

signifies that we have had second thoughts on the Descartes is right, we know nothing of such
matter, or that we are subjecting it to more careful things. What we perceive and are in direct sensory
scrutiny, or that knowledge is being contrasted contact with is never a physical object or state of
with something else, but not that we believe in affairs, but only a representation - something that
something called "real knowledge" which is dif- could be just the way it is even if there were no
ferent from or more elevated than the ordinary objects at all of the sort it represents. So if we were
knowledge we are interested in. Knowing some- to settle for the realm of things we could have
thing differs from merely believing it or assuming knowledge about even if Descartes's conclusion
it or taking it for granted or simply being under were correct, we would not be settling for the
the impression that it is true, and so forth, so comfortable world with which we began. We
asking whether we really know something is would have lost all of that, at least as something
asking whether we know it as opposed to, for we can know anything about, and we would be
example, merely believing it or assuming it or restricted to facts about how things seem to us at
taking it for granted or simply being under the the moment rather than how they are.
impression that it is true. It might still be felt that after all nothing is cer-
If that is true of our ordinary assessments of tain in this changing world, so we should not
knowledge, and if Descartes's investigation of his aspire to firm truths about how things are. As
knowledge that he is sitting by the fire with a piece long as we know that all or most of us agree about
of paper in his hand is just like those ordinary how things seem to us, or have seemed to us up
cases, his discovery that he doesn't know in the till now, we might feel we have enough to give our
case he considers will have the same significance social, cultural, and intellectual life as much sta-
as it has in those ordinary cases. And if that exam- bility as we can reasonably expect or need. But
pIe is indeed representative of our knowledge of again this reaction does not really acknowledge
the world around us, the kind of knowledge we the poverty or restrictedness of the position
are shown to lack will be the very kind of knowl- Descartes's sceptical conclusion would leave each
edge we originally thought we had of things like of us in. Strictly speaking, there is no community
our sitting by the fire holding a piece of paper. of acting, experiencing and thinking persons I
Without a demonstration that Descartes's philo- can know anything about if Descartes is correct.
sophical investigation differs from our ordinary Other people, as I understand them, are not
assessments in some way that prevents its nega- simply sensory experiences of mine; they too, if
tive conclusion from having the kind of signifi- they exist, will therefore inhabit the unreachable
cance similar conclusions are rightly taken to world beyond my sensory experiences, along with
have in everyday life, we can derive no consola- the tables and chairs and other things about
tion from the ungrounded idea that the reality which I can know nothing. So at least with respect
from which he shows our knowledge is excluded to what I can know I could not console myself
does not or should not concern us anyway. It is with thoughts of a like-minded community of
the investigation of his everyday knowledge, and perceivers all working together and cheerfully
not merely the fanciful picture of a veil of percep- making do with what a communal veil of percep-
tion, that generates Descartes's negative verdict. tion provides. I would have no more reason to
But even if we did try to console ourselves believe that there are any other people than I have
with the thought that we can settle for what we to believe that I am now sitting in a chair writing.
can know on Descartes's account, how much con- The representations or sensory experiences to
solation could it give us? The position Descartes's which Descartes's conclusion would restrict my
argument says we are in is much worse than what knowledge could be no other than my own sen-
is contemplated in the optimistic response of sory experiences; there could be no communal
merely shrugging off any concern with an imper- knowledge even of the veil of perception itself. If
ceptible "reality". my own sensory experiences do not make it pos-
For one thing, we would not in fact be left with sible for me to know things about the world
what we have always taken to be the familiar around me they do not make it possible for me to
objects of our everyday experience - tables and know even whether there are any other sensory
chairs, trees and flowers, bread and wine. If experiences or any other perceiving beings at all.

The consequences of accepting Descartes's and that there can be no question of "accepting"
conclusion as it is meant to be understood are Descartes's conclusion at all. I have no wish to
truly disastrous. There is no easy way of accom- discourage such a reaction. I would only insist
modating oneself to its profound negative impli- that the alleged absurdity or unintelligibility must
cations. But perhaps by now we have come far be identified and made out. I think that is the
enough to feel that the whole idea is simply only way we can hope to learn whatever there is to
absurd, that ultimately it is not even intelligible, be learned from Descartes's investigation.


It has been argued that the problem in the E. S. Haldane and G. R. T. Ross (2 vols,
completely general form in which I discuss it New York, 1955), vol. I, p. 145. (Hereafter cited
here is new in Descartes, and that nothing as HR.)
exactly similar appears in philosophy before 3 See his Discourse on the Method of Rightly
that time. See M. F. Burnyeat, "Idealism and Conducting Reason and Seeking Truth in the
Greek Philosophy: What Descartes Saw and Sciences in HR, pp. 81 ff.
Berkely Missed", The Philosophical Review 4 See G. E. Moore, "Certainty", this vol., ch. 4.
(1982). 5 A memorable example H. H. Price gave in
2 See the beginning of the first of his Meditations a lecture in 1962. It is my impression that
on First Philosophy in The Philosophical Price was reporting on an actual hallucination
Works of Descartes, edited and translated by of his.

Proof of an External World

G. E. Moore

It seems to me that, so far from its being true, as case, and not merely something which I believed
Kant declares to be his opinion, that there is only but which was by no means certain, or something
one possible proof of the existence of things out- which, though in fact true, I did not know to be so;
side of us, namely the one which he has given, I and (3) unless the conclusion did really follow
can now give a large number of different proofs, from the premiss. But all these three conditions
each of which is a perfectly rigorous proof; and were in fact satisfied by my proof. (1) The premiss
that at many other times I have been in a position which I adduced in proof was quite certainly dif-
to give many others. I can prove now, for instance, ferent from the conclusion, for the conclusion was
that two human hands exist. How? By holding up merely "Two human hands exist at this moment";
my two hands, and saying, as I make a certain ges- but the premiss was something far more specific
ture with the right hand, "Here is one hand", and than this - something which I expressed by show-
adding, as I make a certain gesture with the left, ing you my hands, making certain gestures, and
"and here is another". And if, by doing this, I have saying the words "Here is one hand, and here is
proved ipso facto the existence of external things, another". It is quite obvious that the two were dif-
you will all see that I can also do it now in num- ferent, because it is quite obvious that the conclu-
bers of other ways: there is no need to multiply sion might have been true, even if the premiss had
examples. been false. In asserting the premiss I was asserting
But did I prove just now that two human hands much more than I was asserting in asserting the
were then in existence? I do want to insist that I conclusion. (2) I certainly did at the moment know
did; that the proof which I gave was a perfectly rig- that which I expressed by the combination of cer-
orous one; and that it is perhaps impossible to give tain gestures with saying the words "Here is one
a better or more rigorous proof of anything what- hand and here is another': I knew that there was
ever. Of course, it would not have been a proof one hand in the place indicated by combining a
unless three conditions were satisfied; namely (1) certain gesture with my first utterance of "here"
unless the premiss which I adduced as proof of the and that there was another in the different place
conclusion was different from the conclusion I indicated by combining a certain gesture with my
adduced it to prove; (2) unless the premiss which I second utterance of "here". How absurd it would
adduced was something which I knew to be the be to suggest that I did not know it, but only
believed it, and that perhaps it was not the case!
You might as well suggest that I do not know that I
From G. E. Moore, Philosophical Papers (New York: am now standing up and talking - that perhaps
Collier Books, 1962), pp. 144-8.
after all I'm not, and that it's not quite certain that

1am! And finally (3) it is quite certain that the con- confront him with a satisfactory proof. But by a
clusion did follow from the premiss. This is as cer- person who questions their existence, he certainly
tain as it is that if there is one hand here and means not merely a person who questions
another here now, then it follows that there are two whether any exist at the moment of speaking, but
hands in existence now. a person who questions whether any have ever
My proof, then, of the existence of things existed; and a proof that some have existed in the
outside of us did satisfy three of the conditions past would certainly therefore be relevant to part
necessary for a rigorous proof. Are there any other of what such a person is questioning. How then
conditions necessary for a rigorous proof, such can 1 prove that there have been external objects
that perhaps it did not satisfy one of them? Perhaps in the past? Here is one proof. 1 can say: "I held up
there may be; 1 do not know; but 1 do want to two hands above this desk not very long ago;
emphasise that, so far as 1 can see, we all of us do therefore two hands existed not very long
constantly take proofs of this sort as absolutely ago; therefore at least two external objects have
conclusive proofs of certain conclusions - as existed at some time in the past, QED". This is a
finally settling certain questions, as to which we perfectly good proof, provided 1 know what is
were previously in doubt. Suppose, for instance, it asserted in the premiss. But 1 do know that 1 held
were a question whether there were as many as up two hands above this desk not very long ago.
three misprints on a certain page in a certain book. As a matter of fact, in this case you all know it toO.
A says there are, B is inclined to doubt it. How There's no doubt whatever that 1 did. Therefore
could A prove that he is right? Surely he could 1 have given a perfectly conclusive proof that
prove it by taking the book, turning to the page, external objects have existed in the past; and you
and pointing to three separate places on it, saying will all see at once that, if this is a conclusive
"There's one misprint here, another here, and proof, 1 could have given many others of the same
another here": surely that is a method by which it sort, and could now give many others. But it is
might be proved! Of course, A would not have also quite obvious that this sort of proof differs in
proved, by doing this, that there were at least three important respects from the sort of proof 1 gave
misprints on the page in question, unless it was just now that there were two hands existing then.
certain that there was a misprint in each of the 1 have, then, given two conclusive proofs of the
places to which he pointed. But to say that he existence of external objects. The first was a proof
might prove it in this way, is to say that it might be that two human hands existed at the time when 1
certain that there was. And if such a thing as that gave the proof; the second was a proof that two
could ever be certain, then assuredly it was certain human hands had existed at a time previous to
just now that there was one hand in one of the two that at which 1 gave the proof. These proofs were
places 1 indicated and another in the other. of a different sort in important respects. And 1
1 did, then, just now, give a proof that there pointed out that 1 could have given, then, many
were then external objects; and obviously, if 1 did, other conclusive proofs of both sorts. It is also
1 could then have given many other proofs of the obvious that 1 could give many others of both
same sort that there were external objects then, sorts now. So that, if these are the sort of proof
and could now give many proofs of the same sort that is wanted, nothing is easier than to prove the
that there are external objects now. existence of external objects.
But, if what 1 am asked to do is to prove that But now 1 am perfectly well aware that, in
external objects have existed in the past, then 1 can spite of all that 1 have said, many philosophers
give many different proofs of this also, but proofs will still feel that 1 have not given any satisfactory
which are in important respects of a different sort proof of the point in question. And 1 want briefly,
from those just given. And 1 want to emphasise in conclusion, to say something as to why this dis-
that, when Kant says it is a scandal not to be able satisfaction with my proofs should be felt.
to give a proof of the existence of external objects, One reason why, is, 1 think, this. Some people
a proof of their existence in the past would cer- understand "proof of an external world" as includ-
tainly help to remove the scandal of which he is ing a proof of things which 1 haven't attempted to
speaking. He says that, if it occurs to anyone to prove and haven't proved. It is not quite easy to
question their existence, we ought to be able to say what it is that they want proved - what it is
28 G. E. MOORE

that is such that unless they got a proof of it, they asserting that I am not now dreaming; I have con-
would not say that they had a proof of the exist- clusive evidence that I am awake: but that is
ence of external things; but I can make an a very different thing from being able to prove it.
approach to explaining what they want by saying I could not tell you what all my evidence is; and I
that if I had proved the propositions which I used should require to do this at least, in order to give
as premisses in my two proofs, then they would you a proof.
perhaps admit that I had proved the existence of But another reason why some people would
external things, but, in the absence of such a proof feel dissatisfied with my proofs is, I think, not
(which, of course, I have neither given nor merely that they want a proof of something which
attempted to give), they will say that I have not I haven't proved, but that they think that, if I
given what they mean by a proof of the existence cannot give such extra proofs, then the proofs that
of external things. In other words, they want a I have given are not conclusive proofs at all. And
proof of what I assert now when I hold up my this, I think, is a definite mistake. They would say:
hands and say "Here's one hand and here's "If you cannot prove your premiss that here is one
another"; and, in the other case, they want a proof hand and here is another, then you do not know it.
of what I assert now when I say "I did hold up two But you yourself have admitted that, if you did not
hands above this desk just now". Of course, what know it, then your proof was not conclusive.
they really want is not merely a proof of these two Therefore your proof was not, as you say it was,
propositions, but something like a general state- a conclusive proof." This view that, if I cannot
ment as to how any propositions of this sort may prove such things as these, I do not know them, is,
be proved. This, of course, I haven't given; and I I think, the view that Kant was expressing in the
do not believe it can be given: if this is what is sentence which I quoted at the beginning of this
meant by proof of the existence of external things, lecture, when he implies that so long as we have
I do not believe that any proof of the existence of no proof of the existence of external things, their
external things is possible. Of course, in some existence must be accepted merely on faith. He
cases what might be called a proof of proposi- means to say, I think, that if I cannot prove that
tions which seem like these can be got. If one of there is a hand here, I must accept it merely as a
you suspected that one of my hands was artificial matter of faith - I cannot know it. Such a view,
he might be said to get a proof of my proposition though it has been very common among philoso-
"Here's one hand, and here's another': by coming phers, can, I think, be shown to be wrong - though
up and examining the suspected hand close up, shown only by the use of premisses which are not
perhaps touching and pressing it, and so estab- known to be true, unless we do know of the exist-
lishing that it really was a human hand. But I do ence of external things. I can know things, which I
not believe that any proof is possible in nearly all cannot prove; and among things which I certainly
cases. How am I to prove now that "Here's one did know, even if (as I think) I could not prove
hand, and here's another"? I do not believe I can them, were the premisses of my two proofs.
do it. In order to do it, I should need to prove for I should say, therefore, that those, if any, who are
one thing, as Descartes pointed out, that I am not dissatisfied with these proofs merely on the
now dreaming. But how can I prove that I am ground that I did not know their premisses, have
not? I have, no doubt, conclusive reasons for no good reason for their dissatisfaction.

Four Forms of Scepticism

G. E. Moore

We pass next to the argument: "Descartes's malicious these three: What is meant by saying that it is logically
demon is a logical possibility:' This is obviously quite possible that this percept was produced by a malicious
different from both the two preceding. Russell does demon? Is it true that this is logically possible? And:
not say that any percepts are produced by Descartes's If it is true, does it follow that I don't know for cer-
malicious demon; nor does he mean that it is practi- tain that it was not produced by a malicious
cally or theoretically possible for Descartes's mali- demon?
cious demon to produce in me percepts like this, in Now there are three different things which
the sense in which it is (perhaps) practically possible might be meant by saying that this proposition is
that a conjurer should, and theoretically possible logically possible. The first is that it is not a self-
that a physiologist should by stimulating the optic contradictory proposition. This I readily grant.
nerve. He only says it is a logical possibility. But what But from the mere fact that it is not self-contra-
exactly does this mean? It is, I think, an argument dictory, it certainly does not follow that I don't
which introduces quite new considerations, of know for certain that it is false. This Russell
which I have said nothing so far, and which lead us grants. He holds that I do know for certain to be
to the root of the difference between Russell and me. false, propositions about my percepts which are
I take it that Russell is here asserting that it is logically not self-contradictory. He holds, for instance, that
possible that this particular percept of mine, which I I do know for certain that there is a white visual
think I know to be associated with a percept belong- percept now; and yet the proposition that there
ing to someone else, was in fact produced in me by a isn't is certainly not self-contradictory.
malicious demon when there was no such associ- He must, therefore, in his argument, be using
ated percept: and that, therefore, I cannot know for "logically possible" in some other sense. And one
certain what I think I know. It is, of course, being sense in which it might naturally be used is this:
assumed that, if it was produced by a malicious Not logically incompatible with anything that I
demon, then it follows that it is not associated with know. If, however, he were using it in this sense,
a percept belonging to someone else, in the way in he would be simply begging the question. For the
which I think I know it is: that is how the phrase very thing I am claiming to know is that this
"was produced by a malicious demon" is being used. percept was not produced by a malicious demon:
The questions we have to consider are, then, simply and of course the proposition that it was produced
by a malicious demon is incompatible with the
proposition that it was not.
From G. E. Moore, Philosophical Papers (New York: There remains one sense, which is, I think, the
Collier Books, 1962), pp. 220-2.
sense in which he is actually using it. Namely he is
30 G. E. MOORE

saying: The proposition "This percept was pro- give complete certainty."] That is to say he
duced by a malicious demon" is not logically assumes: (1) My belief or knowledge that this is a
incompatible with anything you know immedi- pencil is, if I do not know it immediately, and if
ately. And if this is what he means, I own that I also the proposition does not follow logically
think Russell is right. This is a matter about from anything that I know immediately, in some
which I suppose many philosophers would disa- sense "based on" an analogical or inductive argu-
gree with us. There are people who suppose that I ment; and (2) What is "based on" an analogical or
do know immediately, in certain cases, such things inductive argument is never certain knowledge,
as: That person is conscious; at least, they use this but only more or less probable belief. And with
language, though whether they mean exactly what regard to these assumptions, it seems to me that
I am here meaning by "know immediately" may the first must be true in some sense or other,
be doubted. I can, however, not help agreeing though it seems to me terribly difficult to say
with Russell that I never do know immediately exactly what the sense is. What I am inclined to
that that person is conscious, nor anything else dispute, therefore, is the second: I am inclined
that is logically incompatible with "This percept to think that what is "based on" an analogical or
was produced by a malicious demon." Where, inductive argument, in the sense in which my
therefore, I differ from him is in supposing that I knowledge or belief that this is a pencil is so,
do know for certain things which I do not know may nevertheless be certain knowledge and not
immediately and which also do not follow logi- merely more or less probable belief.
cally from anything which I do know immediately. What I want, however, finally to emphasize is
This seems to me to be the fundamental ques- this: Russell's view that I do not know for certain
tion at issue in considering my classes (3) and (4) that this is a pencil or that you are conscious
and what distinguishes them from cases (1) and rests, if I am right, on no less than four distinct
(2). I think I do know immediately things about assumptions: (1) That I don't know these things
myself and such things as "There was a sound like immediately; (2) That they don't follow logically
'Russell' a little while ago" - that is, I think that from any thing or things that I do know immedi-
memory is immediate knowledge and that much ately; (3) That, if (1) and (2) are true, my belief in
of my knowledge about myself is immediate. But or knowledge of them must be "based on an ana-
I cannot help agreeing with Russell that I never logical or inductive argument"; and (4) That what is
know immediately such a thing as "That person is so based cannot be certain knowledge. And what
conscious" or "This is a pencil," and that also the I can't help asking myself is this: Is it, in fact, as
truth of such propositions never follows logically certain that all these four assumptions are true, as
from anything which I do know immediately, and that I do know that this is a pencil and that you are
yet I think that I do know such things for certain. conscious? I cannot help answering: It seems to
Has he any argument for his view that if their me more certain that I do know that this is a pencil
falsehood is logically possible (i.e. if I do not know and that you are conscious, than that any single
immediately anything logically incompatible with one of these four assumptions is true, let alone all
their falsehood) then I do not know them for cer- four. That is to say, though, as I have said, I agree
tain? This is a thing which he certainly constantly with Russell that (1), (2) and (3) are true; yet of
assumes; but I cannot find that he anywhere gives no one even of these three do I feel as certain as
any distinct arguments for it. that I do know for certain that this is a pencil. Nay
So far as I can gather, his reasons for holding it more: I do not think it is rational to be as certain
are the two assumptions which he expresses when of anyone of these four propositions, as of the
he says: "If (I am to reject the view that my life is proposition that I do know that this is a pencil.
one long dream) I must do so on the basis of an And how on earth is it to be decided which of the
analogical or inductive argument, which cannot two things it is rational to be most certain of?


Bertrand Russell, An Outline of Philosophy

(Allen & Unwin: London, 1927), p. 218.


G. E. Moore

Suppose I say: "I know for certain that I am standing it certainly would not follow that I am not stand-
up; it is absolutely certain that I am; there is not ing up; for it is certainly logically possible that a
the smallest chance that I am not:' Many philoso- man should be fast asleep and dreaming, while he
phers would say: "You are wrong: you do not is standing up and not lying down. It is therefore
know that you are standing up; it is not absolutely logically possible that I should both be standing
certain that you are; there is some chance, though up and also at the same time dreaming that I am;
perhaps only a very small one, that you are not." just as the story, about a well-known Duke of
And one argument which has been used as an Devonshire, that he once dreamt that he was
argument in favour of saying this, is an argument speaking in the House of Lords and, when he
in the course of which the philosopher who used woke up, found that he was speaking in the House
it would assert: "You do not know for certain that of Lords, is certainly logically possible. And if, as
you are not dreaming; it is not absolutely certain is commonly assumed, when I am dreaming that I
that you are not; there is some chance, though am standing up it may also be correct to say that I am
perhaps only a very small one, that you are." And thinking that I am standing up, then it follows
from this, that I do not know for certain that I am that the hypothesis that I am now dreaming is
not dreaming, it is supposed to follow that I do quite consistent with the hypothesis that I am
not know for certain that I am standing up. It is both thinking that I am standing up and also
argued: If it is not certain that you are not dream- actually standing up. And hence, if as seems to me
ing, then it is not certain that you are standing up. to be certainly the case and as this argument
And that if I don't know that I'm not dreaming, assumes, from the hypothesis that I am now
I also don't know that I'm not sitting down, dreaming it would follow that I don't know that I
I don't feel at all inclined to dispute. From the am standing up, there follows a point which is of
hypothesis that I am dreaming, it would, I think, great importance with regard to our use of the
certainly follow that I don't know that I am stand- word "knowledge" and therefore also of the word
ing up; though I have never seen the matter argued, "certainly" - a point which has been made quite
and though it is not at all clear to me how it is to be conclusively more than once by Russell, namely
proved that it would follow. But, on the other that from the conjunction of the two facts that a
hand, from the hypothesis that I am dreaming, man thinks that a given proposition p is true, and
that p is in fact true, it does not follow that the
man in question knows that p is true: in order that
From G. E. Moore, Philosophical Papers (New York: I may be justified in saying that I know that I am
Collier Books, 1962), pp. 240-6. standing up, something more is required than the
32 G. E. MOORE

mere conjunction of the two facts that I both can give better reasons for asserting that I don't
think I am and actually am - as Russell has know that I'm not dreaming, than I can give for
expressed it, true belief is not identical with asserting that I do know that I am standing up.
knowledge; and I think we may further add that What reasons can be given for saying that I
even from the conjunction of the two facts that I don't know for certain that I'm not at this moment
feel certain that I am and that I actually am it dreaming?
would not follow that I know that I am, nor there- I do not think that I have ever seen clearly
fore that it is certain that I am. As regards the stated any argument which is supposed to show
argument drawn from the fact that a man who this. But I am going to try to state, as clearly as I
dreams that he is standing up and happens at the can, the premisses and the reasonings from them,
moment actually to be standing up will neverthe- which I think have led so many philosophers to
less not know that he is standing up, it should suppose that I really cannot now know for certain
indeed be noted that from the fact that a man is that I am not dreaming.
dreaming that he is standing up, it certainly does I said, you may remember, in talking of the
not follow that he thinks he is standing up; since it seven assertions with which I opened this lecture,
does sometimes happen in a dream that we think that I had "the evidence of my senses" for them,
that it is a dream, and a man who thought this though I also said that I didn't think this was the
certainly might, although he was dreaming that only evidence I had for them, nor that this by itself
he was standing up, yet think that he was not, was necessarily conclusive evidence. Now if I had
although he could not know that he was not. It is then "the evidence of my senses" in favour of the
not therefore the case, as might be hastily assumed, proposition that I was standing up, I certainly have
that, if I dream that I am standing up at a time now the evidence of my senses in favour of the
when I am in fact lying down, I am necessarily proposition that I am standing up, even though
deceived: I should be deceived only if I thought I this may not be all the evidence that I have, and
was standing when I wasn't; and I may dream that may not be conclusive. But have I, in fact, the evi-
I am, without thinking that I am. It certainly does, dence of my senses at all in favour of this proposi-
however, often happen that we do dream that so- tion? One thing seems to me to be quite clear
and-so is the case, without at the time thinking about our use of this phrase, namely, that, if a man
that we are only dreaming; and in such cases, at a given time is only dreaming that he is standing
I think we may perhaps be said to think that what up, then it follows that he has not at that time the
we dream is the case is the case, and to be deceived evidence of his senses in favour of that proposi-
if it is not the case; and therefore also, in such tion: to say "Jones last night was only dreaming
cases, if what we dream to be the case happens that he was standing up, and yet all the time he
also to be the case, we may be said to be thinking had the evidence of his senses that he was" is to say
truly that it is the case, although we certainly do something self-contradictory. But those philoso-
not know that it is. phers who say it is possible that I am now dream-
I agree, therefore, with that part of this argu- ing, certainly mean to say also that it is possible
ment which asserts that if I don't know now that that I am only dreaming that I am standing up; and
I'm not dreaming, it follows that I don't know that this view, we now see, entails that it is possible that
I am standing up, even if I both actually am and I have not the evidence of my senses that I am.
think that I am. But this first part of the argument If, therefore, they are right, it follows that it is not
is a consideration which cuts both ways. For, if it certain even that I have the evidence of my senses
is true, it follows that it is also true that if I do that I am; it follows that it is not certain that I have
know that I am standing up, then I do know that the evidence of my senses for anything at all. If,
I am not dreaming. I can therefore just as well therefore, I were to say now, that I certainly have
argue: since I do know that I'm standing up, it fol- the evidence of my senses in favour of the propo-
lows that I do know that I'm not dreaming; as my sition that I am standing up, even if it's not certain
opponent can argue: since you don't know that that I am standing up, I should be begging the very
you're not dreaming, it follows that you don't question now at issue. For if it is not certain that I
know that you're standing up. The one argument am not dreaming, it is not certain that I even have
is just as good as the other, unless my opponent the evidence of my senses that I am standing up.

But, now, even if it is not certain that I have at he does not know that he is not dreaming, can he
this moment the evidence of my senses for any- possibly know that he is not only dreaming that
thing at all, it is quite certain that I either have the dreams have occurred? Can he possibly know
evidence of my senses that I am standing up or therefore that dreams have occurred? I do not
have an experience which is very like having the think that he can; and therefore I think that
evidence of my senses that I am standing up. If I anyone who uses this premiss and also asserts the
am dreaming, this experience consists in having conclusion that nobody ever knows that he is not
dream-images which are at least very like the sen- dreaming, is guilty of an inconsistency. By using
sations I should be having if I were awake and had this premiss he implies that he himself knows that
the sensations, the having of which would consti- dreams have occurred; while, if his conclusion is
tute "having the evidence of my senses" that I am true, it follows that he himself does not know that
standing up. Let us use the expression "sensory he is not dreaming, and therefore does not know
experience;' in such a way that this experience that he is not only dreaming that dreams have
which I certainly am having will be a "sensory occurred.
experience," whether or not it merely consists in However, I admit that the premiss is true. Let
the having of dream-images. If we use the expres- us now try to see by what sort of reasoning it
sion "sensory experience" in this way, we can say, might be thought that we could get from it to the
I think, that, if it is not certain that I am not conclusion.
dreaming now, then it is not certain that all the I do not see how we can get forward in that
sensory experiences I am now having are not direction at all, unless we first take the following
mere dream-images. huge step, unless we say, namely: since there have
What then are the premisses and the reason- been dream-images similar in important respects
ings which would lead so many philosophers to to some of the sensory experiences I am now
think that all the sensory experiences I am having having, it is logically possible that there should be
now may be mere dream-images - that I do not dream-images exactly like all the sensory experi-
know for certain that they are not? ences I am now having, and logically possible,
So far as I can see, one premiss which they therefore, that all the sensory experiences I am
would certainly use would be this: "Some at least now having are mere dream-images. And it might
of the sensory experiences which you are having be thought that the validity of this step could be
now are similar in important respects to dream- supported to some extent by appeal to matters of
images which actually have occurred in dreams:' fact, though only, of course, at the cost of the same
This seems a very harmless premiss, and I am sort of inconsistency which I have just pointed
quite willing to admit that it is true. But I think out. It might be said, for instance, that some
there is a very serious objection to the procedure people have had dream-images which were exactly
of using it as a premiss in favour of the derived like sensory experiences which they had when
conclusion. For a philosopher who does use it as a they were awake, and that therefore it must be
premiss, is, I think, in fact implying, though he logically possible to have a dream-image exactly
does not expressly say, that he himself knows it to like a sensory experience which is not a dream-
be true. He is implying therefore that he himself image. And then it may be said: If it is logically
knows that dreams have occurred. And, of course, possible for some dream-images to be exactly like
I think he would be right. All the philosophers I sensory experiences which are not dream-images,
have ever met or heard of certainly did know that surely it must be logically possible for all the
dreams have occurred: we all know that dreams dream-images occurring in a dream at a given
have occurred. But can he consistently combine time to be exactly like sensory experiences which
this proposition that he knows that dreams have are not dream-images, and logically possible also
occurred, with his conclusion that he does not for all the sensory experiences which a man has at
know that he is not dreaming? Can anybody pos- a given time when he is awake to be exactly like all
sibly know that dreams have occurred, if, at the the dream-images which he himself or another
time, he does not himself know that he is not man had in a dream at another time.
dreaming? If he is dreaming, it may be that he is Now I cannot see my way to deny that it is
only dreaming that dreams have occurred; and if logically possible that all the sensory experiences
34 G. E. MOORE

I am having now should be mere dream-images. I know nobody ever has, and I don't know how
And if this is logically possible, and if further the anybody ever could. And so long as this is not
sensory experiences I am having now were the done my argument, "I know that I am standing
only experiences I am having, I do not see how up, and therefore I know that I am not dream-
I could possibly know for certain that I am not ing," remains at least as good as his, "You don't
dreaming. know that you are not dreaming, and therefore
But the conjunction of my memories of the don't know that you are standing up." And I
immediate past with these sensory experiences don't think I've ever seen an argument expressly
may be sufficient to enable me to know that I am directed to show that it is not.
not dreaming. I say it may be. But what if our One final point should be made clear. It is
sceptical philosopher says: It is not sufficient; certainly logically possible that I should have been
and offers as an argument to prove that it is not, dreaming now; I might have been dreaming now;
this: It is logically possible both that you should and therefore the proposition that I am dreaming
be having all the sensory experiences you are now is not self-contradictory. But what I am in
having, and also that you should be remember- doubt of is whether it is logically possible that I
ing what you do remember, and yet should be should both be having all the sensory experiences
dreaming. If this is logically possible, then I don't and the memories that I have and yet be dream-
see how to deny that I cannot possibly know for ing. The conjunction of the proposition that I
certain that I am not dreaming: I do not see that have these sense experiences and memories with
I possibly could. But can any reason be given the proposition that I am dreaming does seem to
for saying that it is logically possible? So far as me to be very likely self-contradictory.

How a Pyrrhonian Skeptic

Might Respond to
Academic Skepticism

Peter Klein

How much do we know? My answer is that we do reject at the outset. It is this: any argument for
not know what the extent of our knowledge is. the claim that reasoning can settle matters will,
But since that answer is not immediately evident of necessity, beg the question because one is
it will require us to employ our reasoning. Thus, employing the very capacity that is at issue in the
the question really becomes this: can our reason- argument.
ing ever give a definitive reply to the question Now, some might respond to that argument
about the extent of our knowledge? And that is by saying that some circular reasoning is permis-
just a specific instance of the general question: is sible - especially if it is logically impossible to
our reasoning able to settle anything, where some avoid it. 1 But I believe all circular reasoning to a
claim is settled by reasoning just in case no further disputed conclusion that has no warrant aside
reasons are required to make the proposition from that provided by the argument is fallacious
completely justified? It is crucial to note that in (more about this later). Thus, if "reasoning can
the way in which I will be using "completely justi- settle matters" were undisputed and had some
fied"; a proposition could be completely justified prima facie warrant not dependent upon an argu-
and false. Hence, I am not asking whether reason- ment (or arguments), then perhaps the fact that it
ing is infallible. In addition, a proposition could cohered with other propositions could raise its
be completely justified and defeasible. Hence, I warrant. But I take it that neither of those condi-
am not asking whether reasoning can produce tions obtains. Since the pre-Socratics, the ability
indefeasible justifications. The question is whether of reasoning to settle matters has been contested,
reasoning - the process of producing reasons for and whatever warrant a favourable assessment
our beliefs - is ever such that further, as yet of reasoning has derives from an argument (or
unused, reasons cannot be legitimately required. arguments). Thus, if the argument(s) for the
Although I will be arguing that reasoning claim that reasoning can settle matters employed
cannot settle anything, there is a rather quick and that very proposition as a premise, that argument
dirty argument to that same conclusion that (or those arguments) could provide no basis for
might seem obviously correct which I wish to thinking that our methods of arriving at beliefs
can settle matters.2
But why should one think that all arguments
Originally published in Steven Loper (ed.), The to the conclusion that good reasoning is reliable
Skeptics: Contemporary Essays (Aldershot: Ashgate, must employ that proposition as a premise? Here
2003), pp. 75-94. is an argument that does not do that:

Good reasoning satisfies conditions C. that reasoning can settle matters, pessimism or
2 Anything satisfying conditions C can settle dogmatism will be the likely result. Pessimism, if
matters. we believe that our goal hasn't yet been satisfied;
3 Therefore, good reasoning can settle matters. dogmatism, if we believe that our goal has been
reached because we might then refuse to inquire
Is that argument circular? No. No premise further thinking that only misleading new
employs the conclusion. And I can see no reason information could be uncovered. 3 But if we set
why a sub-argument for either premise (1) or for what I think is the only realistic goal, namely pro-
premise (2) must employ (3) in one of its premises visionally justified belief, that is belief in a propo-
and so on. I think that there might be such an sition that, as far as we have reasoned satisfies (i)
argument with true premises that can provide us and (ii) above, we can - and at least sometimes
with some reasons for thinking that reasoning will - recognize that further inquiry is always
can settle matters (but those reasons would not appropriate.
settle whether reasoning settles matters). Thus, this chapter can be seen as a defence of a
So, what would lead anyone to think that such form ofPyrrhonism (named after Pyrrho, c.300 Be)
arguments must be circular? The answer, I which endorses neither the claim that we have
believe, is that any prudent person who believes knowledge nor the claim that we do not
(3) will employ what he/she takes to be good have knowledge. This must be carefully distin-
reasoning in fashioning the argument for (3). guished from the more common form of skepti-
But doing so does not commit the fallacy of cir- cism that many, if not most, contemporary
cular reasoning. Indeed, doing so makes one's philosophers find interesting primarily because it
practices consistent with one's beliefs. As we will seems to them to be both highly implausible and
see, satisfying the belief/practice consistency perniciously difficult to reject once the argument
requirement is a problem for the foundationalist for it is investigated. That form of skepticism has
but not for a type of skeptic - the Pyrrhonian been called "Academic skepticism" because it was
type. My point here is that were I not to use what endorsed by members of the Late Academy
I took to be good reasoning in arguing for (3), I founded by Plato, "Cartesian skepticism" because
would legitimately be accused of not practising of the arguments investigated by Descartes and
what I preach. his critics in the mid-seventeenth century, and
Nevertheless, I do believe (but am not pre- "switched world" skepticism by contemporary
pared to say that I know) that our reasoning philosophers because it involves imaging oneself
cannot settle anything, including the question to be in some possible world that is both vastly
about the extent of our knowledge. Pessimism, different from the actual world and at the same
however, is not the proper response to that assess- time absolutely indistinguishable (at least by us)
ment of the power of reasoning. I value reason- from the actual world. I will most often use
ing, as I think we all do. (The "we" in the previous "Academic skepticism" but in order to avoid wea-
sentence means "adult human beings".) What we risome repetition, I will occasionally refer to the
value is having good enough reasons for our same view with one of the other labels. Its central
(actual) beliefs so that it is (i) more reasonable to claim is that we do not (in fact, cannot) have
hold them than to withhold them and (ii) more knowledge or any type of justified belief - even
reasonable to hold them than to hold any con- provisionally justified belief. I will examine the
trary propositions. We might not value having standard argument for Academic skepticism from
such good reasons above all other things - like the Pyrrhonian perspective in order to illustrate
faith, or the pursuit of evil, or the satisfaction of my general claim that reasoning cannot settle
our appetites. But even the religious, the wicked matters.
or the hedonist value reasoning instrumentally
because they want their beliefs to be efficacious
and they believe that reasoning will assist them in Academic Skepticism
achieving that goal.
Wanting good enough reasons is one thing, Here is a way of stating the standard argument for
but if we begin inquiry with the hope or expectation Academic skepticism: 4

If a person, say S, is justified (to some positive underlying premise 1. That principle, called the
degree, d) in believing that there is a table closure principle, goes like this:
before her, then S is justified (to degree d) in
believing that she is not in one of the skeptical Closure principle: if someone, say S, is justified
scenarios in which there is no table but it (to any positive degree, d) in believing some
appears just as though there were one. proposition, say p, and if p strictly implies
2 S is never justified (to degree d) in believing another proposition, say q, then S is justified
that she is not in one of the skeptical scenarios in believing (to degree d) that q.
in which there is no table but it appears just as
though there were one. The issue is: does closure hold for justified belief?
Closure certainly does hold for some properties,
Therefore, S is never justified (to degree d) III for example truth. If p is true and it strictly implies
believing that there is a table before her. q, then q is true. It just as clearly does not hold for
other properties. If p is a belief of mine, and p
This is deeply puzzling because it appears that the strictly implies q, it does not follow that q is a belief
premises are true, that the argument is valid (that of mine. For I might fail to see the implication or I
is, it is not possible for the premises to be true and simply might be epistemically perverse or I might
the conclusion false) but, at the same time, the be "wired" incorrectly (from birth or as the result
conclusion appears false. Further, it seems that of an injury). I might, for example, believe all of
there are only three possible responses: (1) deny at the axioms of Euclidean plane geometry, but fail to
least one premise of the argument; (2) deny that believe (or perhaps even refuse to believe) that the
the argument is valid; (3) accept the conclusion. 5 exterior angle of a triangle is equivalent to the sum
None of those options seems initially promising. of the two opposite interior angles.
The belief that we have no knowledge seems Since closure does not hold for belief, it prob-
preposterous and the argument certainly seems ably doesn't hold for justified belief when that
valid. Thus, the strategy of choice for rejecting entails that S actually has the belief. 6 In addition,
Academic skepticism has been to deny at least one since a necessary truth is entailed by every propo-
of the premises. But the prospects of finding a sition, if S were justified in believing any proposi-
basis for rejecting a premise are dim because, on tion, then S would be justified in believing every
close inspection, the arguments for doing so seem necessary truth. But these are matters of detail
to rest on assumptions that are both unmotivated and the principle can be repaired to account for
and ones which the Academic skeptic should these minor problems. We could, for example,
reject. There are many types of those arguments, restrict the range of the propositions justifiably
but I will consider only one type in order merely believed to contingent ones, and we could restrict
to illustrate my point (as opposed to demonstrat- the entailments to known ones, and we could
ing it). I chose this argument against Academic stipulate that S could be justified in believing that
skepticism because it has struck many as the most p without actually believing that p. The real issue
plausible and also because investigating it will is this: does closure hold for what we are entitled
prove to be very useful later. to believe (even if we don't, in fact, believe it)?
To unbag the cat now: I do not think that there It certainly seems that it does. For ifI am enti-
is a good response available within the three alter- tled to believe p and p strictly implies q, then how
natives just mentioned but I will propose a fourth could I fail to be entitled to believe q? If, for exam-
alternative response to Academic skepticism that ple, I am justified in believing that today is
employs the general considerations about the Wednesday, then I must be justified in believing
limits of reasoning which I will be exploring. But, that it is not Thursday. Nevertheless, the principle
for now, let us focus on what I think is the most has been challenged. Consider this much dis-
plausible argument against Cartesian skepticism cussed counter-example to the closure principle
that can be given within the three options listed developed by Fred Dretske:
It is an argument based upon some sup- something's being a zebra implies that it is not a
posed counter-examples to the general principle mule ... cleverly disguised by the zoo authorities

to look like a zebra. Do you know that these ani- This pattern illustrates the constraint on closure
mals are not mules cleverly disguised? If you are imposed by Dretske, namely that whenever p
tempted to say "Yes" to this question, think a entails q, the adequate evidence, e, for p is the very
moment about what reasons you have, what same evidence that is adequate for q.
evidence you can produce in favor of this claim. No doubt this constraint sometimes correctly
The evidence you had for thinking them zebras portrays the relevant evidential relationships
has been effectively neutralized, since it does not when some proposition, p, entails some other
count toward their not being mules cleverly proposition, q. For example, suppose I have ade-
disguised to look like zebras. (1970, pp. 1015-16) quate evidence that Anne has two brothers, then
it would seem that the very same evidence would
Dretske is speaking of "knowledge" rather than be adequate for believing that Anne has at least
beliefs to which one is entitled, but that seems irrel- one brother. But the Academic skeptic would
evant since the issue concerns the supposed lack of (or at least should) point out that closure need
sufficient evidence or reasons for the claim that the not require that type of evidence path in all cases
animal is not a cleverly disguised mule. 7 In other in which one proposition entails another.
words, Dretske grants that there is an adequate There are two other possibilities for instantiat-
source of justification for the claim that the animal ing closure that can be depicted as follows:
is a zebra, but he claims that the adequate source of
evidence that you have for identifying the animals
Pattern 2 ... ReRp ... Rq
as zebras is not an adequate source for determining
Pattern 3 ... Re (where e includes q) Rp
that they are not cleverly disguised mules.
The crucial thing to note about this proposed
counter-example is that it works only if the clo- In Pattern 2 cases there is some adequate evidence, e,
sure principle entails that the very same evidence for p; and p, itself, is the adequate evidence for q,
that justifies S in believing that the animals are since p strictly implies q. For example, if I have ade-
zebras must justify S in believing that they are not quate evidence for believing that 2 is a prime
cleverly disguised mules because, it is presumed, number, I can use that proposition as an adequate
that is the only evidence that we can be sure S has. reason for believing that there is at least one even
To generalize, the purported counter-example prime. In Pattern 3 cases the order of the evidence is
depends upon the assumption that the closure of reversed because q serves as part of the evidence
justified belief depends upon it being the case for p. For example, I am justified in believing that
that the very same evidence, e, that justifies S in water is present if I am justified in believing
believing the entailing proposition, p, also justi- that there is present a clear, odourless, watery-
fies S in believing the entailed proposition q. tasting and watery-looking fluid at STP. This
Thus, letting "xRy" mean that x is an adequate pattern is typical of abductive inferences.
reason for y, the counter-example depends upon Thus, showing that there is no Pattern 1 type
assuming that if closure holds between p and q, evidence path available to S in the zebra-in-the-
then the evidence "path" must look like this: zoo case is not sufficient to show that closure fails.
Indeed, I would suggest that the animals looking
Pattern 1 ... Rp like zebras in a pen marked "zebras" are, ceteris
... Re paribus, adequate evidence to provisionally jus-
~ ... Rq tify the claim that they are zebras; and once S is
entitled to believe that the animals are zebras,
Evidence paths specify what propositions serve as S can reasonably deduce from that proposition
good enough reasons, ceteris paribus, for believ- that they are not cleverly disguised mules. That is,
ing other propositions. So, in Pattern 1 type cases, S can employ an evidence path like that depicted
if S has good enough reasons for believing the in Pattern 2. 8 Or alternatively, if S had some reason
proposition e, then S is entitled to "take" the evi- to think that the animals were cleverly disguised
dence path to proposition p, and S is entitled to mules, then S would have to eliminate that possi-
take the path to proposition q. So if S can get to bility before she could justifiably believe that they
point e on the path, S can get to points p and q. are zebras. In other words, in that case S would

have to employ an evidence path like the one of inquiry is subject to actual or potential dispute
depicted in Pattern 3. and that reasoning is employed to resolve the dis-
I think it is clear that this alleged refutation of pute. The issue before us then is whether there is
Academic skepticism based upon the rejection of a mode of reasoning that can settle matters about
closure rests upon a premise that requires further which there is some dispute. Of those modes,
reasons to support it, namely the premise that the Sextus writes:
appropriate evidential relationship between "the
animals are zebras" and "the animals are not clev- The Mode based upon regress ad infinitum is
erly disguised mules" is that depicted in Pattern 1. that whereby we assert that the thing adduced as
There are other patterns of reasoning that instan- a proof of the matter proposed needs a further
tiate closure and until some reason is given for proof, and this again another, and so on ad infini-
thinking that the appropriate pattern in this case tum, so that the consequence is suspension [of
is Pattern I, reasoning would have failed to settle assent], as we possess no starting-point for our
argument ... We have the Mode based upon
whether closure should be rejected.
hypothesis when the Dogmatists, being forced to
Thus, one of the purposes of exploring this
recede ad infinitum, take as their starting-point
argument against Academic skepticism has been
something which they do not establish but claim
fulfilled, namely to illustrate the general claim I
to assume as granted simply and without dem-
will be defending shortly that our reasoning onstration. The Mode of circular reasoning is the
cannot settle matters. The other purpose was to form used when the proof itself which ought to
gesture in the direction of a general claim that the establish the matter of inquiry requires confir-
prospects are dim for the success of anyone of the mation derived from the matter; in this case,
three alternative responses to the argument for being unable to assume either in order to estab-
Academic skepticism mentioned above. I have lish the other, we suspend judgement about both.
certainly not demonstrated that there is no way to (1993, I, pp. 166-9)
respond to the Academic skeptic within the three
alternatives. But I have shown that one of the I will call the first account of the normative con-
better known responses is not compelling. ditions required for complete justification "infin-
So, here is what remains for me to do: itism".10 Today we commonly refer to the second
account as "foundationalism". Finally, I will refer
Argue that reasoning, in general, cannot settle to the third possibility as "coherentism" - but
matters, but that provisionally justified belief some important distinctions between forms of
is still possible. coherentism will be discussed below.
2 Apply that general conclusion to the argu- The regress problem, then, can be stated briefly
ments for academic skepticism in order to in this way: there is a trilemma facing all who
delineate the fourth alternative response men- attempt to use reasoning to settle matters. Either
tioned earlier. foundationalism, coherentism or infinitism is the
appropriate method of responding to the regress
of reasons. Foundationalism appears to advocate
Pyrrhonian Skepticism a process of reasoning that relies upon arbitrary
propositions at the base. (What makes a proposi-
My belief that reasoning cannot settle matters can tion arbitrary will be discussed later.) Coherentism
be traced to a famous passage in Sextus Empiricus's is nothing but a thinly disguised form of circular
Outlines of Pyrrhonism in the chapter called "The reasoning. Finally, infinitism advocates a process
Five Modes" in which he discusses the regress of justification that could never be completed.
problem. Although the chapter title mentions five Put another way: there are only three norma-
modes, two of them repeat those found else- tive constraints that could apply to any instance of
where. 9 They are the modes of discrepancy and reasoning. For either the process of producing rea-
relativity and are important here because they sons properly stops at foundational beliefs or it
provide the background for understanding the doesn't. If it does, then foundationalism is correct.
description of the three modes of reasoning. If it doesn't, then either reasoning is properly cir-
Specifically, it is presumed that the relevant object cular, or it is properly infinite and non-repeating. ll

There are no other possibilities. 12 Thus, if none of between two people: the Foundationalist, Fred,
these forms of reasoning can settle matters, no and the Pyrrhonian Skeptic, Sally. Fred begins
form can. by saying that he believes that p. He might say
My view of the matter is that (1) the hinted -at something quite strong like "I know that p" or "p
arguments of the Pyrrhonians against founda- is certainly true" or "I have conclusive reasons
tiona/ism and coherentism, when properly fleshed for p", or he might just say "p is true" or even just
out, do render plausible the claim that those forms "p" with the appropriate gusto. The Pyrrhonian
of reasoning are inherently unacceptable models Skeptic, Sally, asks Fred-the-foundationalist why
of good reasoning because they cannot provide he believes that p is true. Fred gives his reason
the basis for any type of rational practice leading for believing that p, say q. Again, Sally asks Fred
to the acceptance of beliefs. But (2) infinitism, why he believes that q is true. Fred replies. This
when properly understood, appears acceptable goes on a while. Finally, Fred (being a founda-
and can lead to provisional justification. tionalist) replies by citing what he takes to be a
So, I want now to take up foundationalism basic proposition, b.
and coherentism and provide some reasons for Sally then asks Fred, "Why do you think b is
thinking that they cannot provide a good model true?" Fred, being a self-conscious foundational-
for reasoning, where reasoning is understood ist, replies that b is properly basic and has some
simply to be the process of producing reasons for warrant that does not depend upon any further
our beliefs. Then, we will turn to infinitism. reason for thinking b is true. 15 To use our termi-
Finally, I want to apply the lessons learned in the nology, Fred is claiming that b has some autono-
general discussion of reasoning to the problem mous bit of warrant. Sally replies as follows, "But
with which we started, namely the standard argu- Fred, what I asked you was 'What makes you think
ment for Academic skepticism, in order to explain b is true?' and you replied, in part, by claiming
the fourth alternative response, mentioned above, that b is a basic proposition. So you must think
to that form of skepticism. that because a proposition is basic there is some
positive likelihood, however small, that it is true.
Right? That is, you must think that propositions
Foundationalism possessing the autonomous bit of warrant are
more likely to be true than they would be were
Foundationalism comes in many forms. But all they not to possess that autonomous bit of war-
forms hold that all propositions are either rant. Why do you think that possession of auton-
basic propositions or non-basic propositions and omous warrant is linked in any way with truth?"
no proposition is both. Basic propositions have We have come to the crucial point in the dis-
some autonomous bit of warrant that does not cussion. For Fred faces a dilemma. Either Fred
depend (at all) upon the warrant of any other will give a reason for thinking that the possession
proposition. 13 Non-basic propositions depend of autonomous warrant is at least somewhat
(directly or indirectly) upon basic propositions truth-conducive or he won't.
for all of their warrant. Consider the first alternative. If Fred has a
I do not believe that this account of the struc- reason for thinking that propositions possessing
ture of reasons can provide a model of reasoning the autonomous bit of warrant are, in virtue of
that can be rationally practised. My discussion of that fact, likely to be true (even to some small
this issue will be reminiscent of Laurence extent), then the regress has not actually stopped,
BonJour's (1978) rejection of foundationalism for Fred has a reason for thinking that b is true.
but unlike his argument, I am not claiming (here) Fred has given up his foundational ism in order to
that this account of the structure of reasons is satisfy a perfectly reasonable question, namely
false. 14 My claim is that a foundationalist cannot "Do you think the possession of autonomous
rationally practise his foundationalism because it warrant is linked to truth?"
inevitably leads to arbitrariness, that is asserting a Now, consider the second alternative and sup-
proposition for no reason at all. pose that Fred does not have a reason for thinking
To see that foundationalism cannot provide a that b's possession of the autonomous bit of war-
rational model of reasoning consider a discussion rant makes it at all likely that b, or any other basic

proposition, is true. Then Fred ought (rationally) Now consider a more traditional brand of
to give up assenting to all basic and non-basic foundationalism and suppose that Fred offers a
propositions.1 6 After all, Fred has no reason for first-person introspective report as the basic
thinking that the basic ones are (somewhat likely proposition, for example "I seem to remember
to be) true by virtue of whatever he thinks makes that there is a football game on TV today". When
them basic and, being a foundationalist, he asked why he thinks that it is true that he seems to
believes that without the warrant provided by the remember that there is a game, Fred could say
basic propositions, the non-basic ones are not that he has no reasons for thinking it is true - he
warranted. He has been forced to admit that just does think it is true. Arbitrariness looms.
accepting basic propositions and everything that What is much more likely is that Fred will come
depends upon them is arbitrary - meaning by up with a story about how he acquires knowledge
arbitrary that there are no better reasons for of his memories - a story told to get Sally to
thinking that they are true than for thinking that see why introspection delivers propositions that
they are falseY are (at least) likely to be true. It could be a rela-
Consider an example: suppose that it is argued, tively straightforward story about our privileged
along contextualist lines, that some propositions access to certain kinds of our states, for example
just don't need to be justified - what makes a certain kinds of mental states, as contrasted with
proposition properly basic is some fact about the our ability to gain knowledge of other of our
context - perhaps that it is accepted by some states. Privileged access, it could be claimed, is
specified group of people. 18 just that sort of access such that the content of
First, I don't think that is a plausible charac- what is delivered is very likely to be true. Or the
terization of the property that could possibly story that Fred could tell could be a relatively
make a proposition basic. 19 I grant that on many complex one - perhaps even one that Fred thinks
occasions the foundationalist will not be chal- contains a priori propositions - about the mean-
lenged to provide a reason for the offered basic ing of some "methodologically basic" words and
proposition - perhaps because everyone in the the conditions for their application which guar-
relevant context believes it and accepts it as a antee that they are "true in the main':zo The con-
reason for further beliefs. But it is crucial to note tent of the story is not crucial here. What is crucial
that the unchallenged stopping points could is that Fred is giving his reasons for thinking that
include a wide variety of propositions. Suppose the propositions of a certain sort are likely to be true.
issue at hand is whether there is an American foot- Thus, in order to avoid arbitrariness, Fred has
ball game on TV today. The response, "Today is offered reasons for thinking that introspection
New Year's Day" might stop the conversation. reports or propositions about methodologically
Similarly, "The newspaper said so" or "My mother basic items are likely to be true. In other words,
told me" could all be conversation-stoppers. But are the regress continues.
they basic in the sense required by the foundation- It is crucial to recall that I am not claiming that
alist? I doubt it. They do not have any autonomous foundational ism is false. Perhaps some proposi-
warrant. For if I didn't believe that I was correctly tions do have autonomous warrant which is
reading the calendar, or that I am correctly remem- truth-conducive and all other propositions
bering the newspaper story, or that I understood depend for some of their warrant upon those
what my mother said, the conversation-stopping basic propositions (although I doubt it). What I
propositions would possess no warrant at all. am claiming is that there is a deep irrationality in
Second, and more to the point, even if contex- being a practising self-conscious foundationalist.
tualism correctly identified what makes a propo- If Fred remains true to his foundationalism, he
sition basic, the crucial point here is that the will not provide a reason for thinking that the
contextualist response does not serve to stop the basic proposition, b, is true unless that reason
regress. For the foundational is tic contextualist ultimately depends upon other basic proposi-
will still be asked this: does the fact that a propo- tions. But basic propositions are supposed to have
sition is a conversation-stopper give anyone any some warrant that does not depend upon another
reason for thinking that it is true? And Fred's proposition being warranted. So, the question to
dilemma returns. Fred can be put this way: on the assumption that

you cannot appeal to any other proposition, do did not rain last night. In this fashion anything
you still think b is true? Fred not only won't have could be justified - too simply! It is ultimately
any such reason for thinking b is true, given that arbitrary which set of mutually probability
constraint, he cannot have one (if he remains true enhancing propositions we believe because there
to his foundationalism). Arbitrariness is inevita- is no basis for preferring one over the other.
ble. Of course, foundationalists typically realize The warrant -transfer coherentist could attempt
this and, in order to avoid arbitrariness, tell some to reply to this objection by claiming that there is
story that, if true, would provide a reason for some property in one of the two competing cir-
thinking basic propositions are at least somewhat cles that is not present in the other and the pres-
likely to be true. But then the regress of reasons ence of that property makes one and only one of
has continued. Foundationalism, then, cannot the circles properly circular. For example, in one
provide a good model for reasoning since, when and only one of the circles are there propositions
practised, it endorses arbitrariness. that we actually believe, or perhaps believe spon-
taneously (BonJour 1985). More generally, the
coherentist could claim that all and only circles
Coherentism with some property, P, have some initial plausi-
bility. But then it is clear that the warrant-transfer
Let us now turn to coherentism. This section can coherentist has adopted a form of foundational-
be much shorter because we can apply some of ism because she is now claiming that all and only
the lessons learned in the discussion of founda- the propositions in circles with P have the auton-
tionalism. omous bit of warrant. And, all that we have said
At its base, coherentism holds that there are about the dilemma facing the foundationalist
no propositions with autonomous warrant. But transfers immediately. Is the possession of P
it is important to note that coherentism comes in truth-conducive or not? If it is ... well, you can
two forms. What I choose to call the "warrant- see how that would go.
transfer form" responds to the regress problem So much for the warrant-transfer version of
by suggesting that the propositions are arranged coherentism. The second form of coherentism,
in a circle and that warrant is transferred within what we can call the "warrant-emergent form",
the circle - just as basketball players standing in a does not imagine the circle as consisting of
circle pass the ball from one player to another.21 propositions that transfer their warrant from
I could, for example, reason that it rained last one proposition to another. Rather warrant for
night by calling forth my belief that there is water each proposition in the circle depends upon the
on the grass and I could reason that there is water fact that they are mutually probability enhanc-
(as opposed to some other liquid, say glycerin, ing. Coherence itself is the property by virtue of
that looks like water) on the grass by calling forth which each member of the set of propositions
my belief that it rained last night. has warrant. Warrant emerges all at once, so to
At the beginning of the chapter, I claimed that speak, from the web-like structure of the propo-
all circular reasoning in which the contested con- sitions. The coherentist can then argue that the
clusion was employed as a premise and for which fact that the propositions cohere provides each
no warrant existed beyond what was available to of them with some prima facie credibility.
be transferred via the argument was fallacious. This might initially seem to be a more plausi-
That is the model of reasoning embedded in the ble view since it avoids the circularity charge. But,
warrant-transfer form of coherentism.1t seems to aside from the fact that there are, again, just too
me that Aristotle explained why it is fallacious. many competing circles that are coherent, there is
As he put it: this is a "simple way of proving any- one, by now very familiar, problem with this
thing" (1994: bk I, ch. 3, 73a5). The propositions alternative. It is crucial to note that the coherentist
in the circle might be mutually probability is now explicitly assigning some initial positive
enhancing, but the point is that we could just as warrant to all of the individual propositions in a
well have circular reasoning to the conclusion set of coherent propositions that does not depend
that it did not rain last night because the liquid is upon the warrant of any other proposition in the
not water and the liquid is not water because it set. In other words, he is assigning to them what

we have called the autonomous bit of warrant. one proposition over its competitors and provi-
Thus, this coherentist has, once again, endorsed a sionally justified in believing it rather than
form of foundationalism and, once again, the withholding belief?
dilemma facing the foundationalist returns. First, it is important to note that the infinitist
Let me sum up where we are at this point. can rationally practise what she thinks is the cor-
There seems no way for the foundationalist or the rect solution to the Pyrrhonian trilemma even
coherentist to avoid arbitrariness and at the same though the process of justifying a proposition is
time stop the regress. It is now time to consider never completed. When needed, the infinitist can
what happens if the regress is unavoidable. always seek a further reason. Contrast this with,
say, the foundationalist who must produce a
reason for which no further reason can be given -
Infinitism even when sincerely requestedY
Second, infinitism provides a good model for
Infinitism is the view that the answer to the provisionalism. Here's how it goes. The infinitist
regress problem is that the regress never properly finds a reason, q, for her belief, p. She would not
ends. There is always another reason, one that has think that it is settled whether p is true because
not already been employed, that can legitimately she knows that she will never complete the pro-
be required for each reason that is given for a cess of providing the infinite set of reasons for p
belief. Only if there is an infinite set of non- (if there is such a set). However, if she does locate
repeating reasons available for a belief is it fully a reason for p and she doesn't have an equally
justifiable. good or better reason against it, it would be more
There is an obvious objection to this form of reasonable to believe p than to deny p or withhold
reasoning as a method for settling what we should p because she does have a reason for believing it.
believe. Here's a close paraphrase of the objection Indeed, on many occasions perhaps we can't help
as put by Jonathan Dancy: but believe p - at least to some extent - if we have
better reasons for it than against it. But we can
Suppose that all justification is inferential. We assess the epistemic situation and, as infinitists,
justify belief A by appeal to belief B, and belief come to recognize that we ought not to think it is
B by appeal to C. The result is that A is justified settled that p, even though it is more reasonable
only if Band C are. Justification by inference is to believe p than to deny p or to withhold p.
conditional justification only; Ns justification
is conditional upon the justification ofB and C.
But if all justification is conditional in this A Clarification and Partial Defence
sense, then nothing is actually non-conditionally of Infinitism
justified. (1985, p. 55)22
Nevertheless, it is one thing to claim that infin-
My response is that Dancy is absolutely right: itism can provide an acceptable account of
infinitism does not sanction non-conditional jus- rational belief and another to claim that infin-
tification. But that is quite different from the itism is true. This is not the place for a full-blown
objections we discovered to foundationalism and defence of infinitism.24 But I would like to con-
coherentism. There we found that those models sider one reason that has been offered for reject-
of reasoning were unacceptable because they ing it because doing so will help to clarify
endorsed arbitrariness or circularity. We have just infinitism.
seen that infinitism is not able to provide an The worry is simply this: how could I have an
account of a type of reasoning that would settle infinite number of beliefs? I have a finite mind.
matters because each belief in the set of offered Here is how John Williams puts this worry:
reasons is only provisionally justified. But that
does not lead to the conclusion that infinitism is The [proposed] regress of justification of S's
unable to be practised rationally. belief that p would certainly require that he holds
So the question becomes this: can the practis- an infinite number of beliefs. This is psychologi-
ing infinitist be provisionally justified in believing cally, if not logically, impossible. If a man can

believe an infinite number of things, then there normative conditions required for full justifica-
seems to be no reason why he cannot know an tion. And each could have an optimistic or pes-
infinite number of things. Both possibilities con- simistic form. Thus, it would not be an objection
tradict the common intuition that the human to infinitism to claim that there is no such infi-
mind is finite. Only God could entertain an infi- nite set of available propositions. Pessimistic
nite number of beliefs. But surely God is not the infinitism is an available option.
only justified believer. (1978, pp. 311-12) Perhaps, though, the worry here is that infin-
itism comes in only the pessimistic form. If that
I think this worry (or perhaps set of worries) can were true, it would not constitute an epistemic
be resolved by clarifying what infinitism claims. It reason for rejecting infinitism. For that which we
is crucial to remember that infinitism is not a have reason to believe true is sometimes quite
form of dogmatism. It acknowledges that we do discouraging. If pessimistic infinitism were the
not ever have fully justified beliefs - perhaps that only reasonable alternative, we might strongly
epistemic state is available only to a being that wish it to be otherwise and so, perhaps, it would
could consciously and simultaneously entertain an be better, in some sense, were infinitism false.
infinite number of beliefs. But the issue here is But, second, I think the worry that infinitism
not whether we can be fully justified, it is whether comes in only the pessimistic form misconstrues
we can have provisionally justified beliefs. what, in general, is required for a belief to be
Nevertheless, there is a deep worry here that available and what, in particular, infinitism
does not depend essentially upon how many or requires for beliefs to be available. Generally,
how few conscious states, that is non-overlapping beliefs are dispositions to sincerely assert some-
temporal states, humans can occupy during a thing under the appropriate conditions. We can
finite time period. The worry is this: there is no have those dispositions even if we have never con-
reason to believe that there is an infinite number sciously entertained the proposition. For example,
of propositions available to us that could serve as I think that we all believe that pears don't nor-
reasons for our beliefs. mally grow on apple trees, that 61 + 346 = 407,
The response to that worry is twofold. First, and that Chicago is east of every city in California,
like foundationalism and coherentism, infinitism but most of us have never before considered those
comes in two varieties - an optimistic and pessi- propositions. Thus, we might very well have an
mistic form. What both varieties of infinitism infinite number of beliefs even though we will
have in common is the belief that the normative never consciously entertain an infinite number of
conditions for full or complete justification propositions.
include the existence of an infinite series of non- Equally, if not more, important is the fact that
repeating reasons available to us for our beliefs. we have the capacity to develop new reasons for
The optimistic form goes on to claim that in the our beliefs when we are called upon to do so. For
required sense there are such reasons available. example, at a certain point in human history we
The pessimistic variety says that there are no such did not believe that diseases were caused by
reasons available. 25 microscopic organisms. Nevertheless, we had the
Consider the parallel with foundationalism. capacity to form that belief. Of course, we needed
A foundationalist holds that the normative con- new experiences, insight and perhaps a certain
ditions of complete justification require that all amount of luck in order to form it. But the new
of our non-foundational beliefs rest on some basic belief was formed. Thus, beliefs might be availa-
beliefs with autonomous warrant. An optimistic ble to us in the requisite sense even though we do
foundationalist - a Cartesian, for example - could not have them.
claim that there are such foundational beliefs. Infinitism requires only that there be an infi-
A pessimistic foundationalist - a Humean, for nite set of distinct propositions each member of
example - could claim that (at least for many of which we have the capacity to legitimately call
our important beliefs) there is no such set of basic forth as reasons for our beliefs. 26 It does not
beliefs. require that we have already formed the beliefs
Infinitism is like foundationalism and co her- with those propositions as their contents.
entism because all three are theories about the Optimistic infinitism says that there is such a set.

Pessimistic infinitism says that at some point we assumption. If that were true, the argument
will run out of such available reasons. It predicts would give us no good reason for accepting that
that we will hit a permanent brick wall of ulti- form of skepticism.
mate arbitrary beliefs or we will have to employ a Recall the three patterns of reasoning exhibit-
reason that has already appeared in the path of ing closure:
reasons. History suggests to me - but of course it
does not fully justify for me - that when we need Pattern 1 ~ ... Rp
new reasons for our beliefs we can find them.
... Re
No belief is ever fully justified for any person.
~ ... Rq
The process of justifying a proposition is never
completed. That is a consequence of infinitism.
But that is not because there is no infinite set of Pattern 2 ... ReRp ... Rq
propositions available that could serve as good Pattern 3 ... Re (where e includes q) Rp
reasons for our beliefs. Rather, no belief is fully
justified because at no point in time will we have And recall the standard argument for Academic
completed the process of justifying our beliefs. skepticism:
All justification is provisional. And as mentioned
at the very beginning of this paper, that's a good If a person, say S, is justified (to some positive
thing to recognize since it provides a basis for degree, d) in believing that there is a table
avoiding pessimism and dogmatism. before her, then S is justified (to degree d) in
believing that she is not in one of the skeptical
scenarios in which there is no table but it
Academic Skepticism Reconsidered appears just as though there were one.
2 S is never justified (to degree d) in believing
Now, before concluding, I want to return to the that she is not in one of the skeptical scenarios
puzzle with which we began, namely the argu- in which there is no table but it appears just as
ment for Academic skepticism, and test what we though there were one.
have learned about reasoning. Does the standard
argument beg the question or depend upon an Therefore, S is never justified (to degree d) In
arbitrary assumption for which there are no better believing that there is a table before her.
reasons for believing than there are for denying?
Recall where we left off. We saw that the argument Now, suppose that the Academic skeptic thinks
for Academic skepticism looked pretty good: the that closure regarding justification holds between
premises seemed true and the argument seemed "there is a table before me" and "I am not in one
valid. Nevertheless, the conclusion seemed false. of the skeptical scenarios in which there is no
We also saw that one argument against premise 1 table but it appears just as though there were one"
ended with an arbitrary assumption, namely that because the requisite evidential path exemplifies
the closure of justified beliefs depended upon the Pattern 1. That is, he holds that the very same evi-
- claim that all reasoning sanctioned by closure was dence that is adequate for arriving at the proposi-
like that depicted in Pattern 1. Thus, we discov- tion that there is a table is adequate for denying
ered one important instance of the general the skeptical hypothesis. Grant that premise 1 is
Pyrrhonian claim that arguments that end, end true for that reason. But, now, when the Academic
either arbitrarily or commit the fallacy of begging skeptic argues for premise 2 - as surely he must
the question. since it is not immediately evident - the sub-
But there is another lesson here related to our argument for premise 2 must be good enough to
discovery of the various patterns of reasoning establish that there is no evidence adequate to
that instantiate closure. For a careful examination justify the proposition that there is a table before
of them reveals the fourth alternative, mentioned him because the argument must be good enough
earlier, for appraising the standard argument for to show that S cannot arrive at point e on the evi-
Academic skepticism, namely that it, too, either dence path. For if S were able to do that, S would
begs the question or is based upon an arbitrary be able to arrive at the denial of the skeptical

hypothesis. Hence, the argument for premise 2 possible alternative before we arrive at the one
would be sufficient to show that the conclusion is that is acceptable? Return to the zebra-in-the-zoo
true and the argument employing premises 1 and case. Would we have to eliminate the hypothesis
2 begs the question since the argument for that the zebra-like looking things are cleverly dis-
premise 2 alone establishes the conclusion. The guised aliens or very cleverly disguised members
standard argument does not work. of the long lost tribe of Israel before arriving at
The situation vis avis this version of the argu- the proposition that they are zebras? I doubt it.
ment for Academic skepticism is similar to this So, why should we have to eliminate the skeptical
one for God's existence: hypothesis before arriving at the proposition that
there is a table before us? Is the skeptical hypoth-
The Bible says "God exists". esis prima facie plausible? No. Is there some evi-
2 Whatever the Bible says is true. dence that it is true? No. In fact, the requirement
that we eliminate every contrary hypothesis to p
Therefore, God exists. before we are entitled to believe p has the conse-
quence that we must have entailing evidence for
As stated, this argument doesn't beg the question. p - and surely that is too strong a requirement for
Similarly, as stated, the standard argument for being justified in believing a contingent, empiri-
Cartesian skepticism does not beg the question. cal claim such as "there are zebras in the pen".28
But if the argument for premise 2 in the argu- So, this argument for Academic skepticism rests
ment for God's existence were that the Bible was upon an arbitrary assumption that we must
written by God and whatever God writes is true, eliminate the skeptical hypothesis before we are
then the argument would beg the question justified in believing any contingent, empirical
because the sub-premises employed in the argu- proposition.
ment for premise 2 imply the conclusion. My Thus, on careful inspection, this finite argu-
claim is that if the Cartesian skeptic thinks ment -like all such arguments, if the Pyrrhonian
premise 1 in his argument is true because the is correct - either ends in an arbitrary assump- .
appropriate evidential relationships are depicted tion or begs the question. Thus, the fourth alter-
by Pattern 1, he will be forced to beg the question native for rejecting the argument for Academic
when he gives his argument for premise 2. skepticism involves conceding that even if all of
I think this point becomes clearer when we the premises are true and even if it is valid, the
look at Pattern 2. If the Cartesian skeptic thinks argument, at least so far, gives us no good reason
that closure holds in this case because one must for accepting the conclusion.
arrive on the inference path at the proposition
that there is a table before one arrives at the denial
of the skeptical hypothesis, then in arguing for the Conclusion
second premise the skeptic must show that we
cannot arrive at the proposition that there is a We have come a long way and covered a lot of
table, because if we did, we could get to the denial ground. To sum up, I have argued for the follow-
of the skeptical hypothesis (since that is what this ing two main points:
instantiation of closure maintains). But if the sub-
argument for premise 2 shows that we can't arrive Reasoning cannot settle matters, but it can
at the proposition that there is a table, then that provide provisional justification. Further
sub-argument already establishes the conclusion. inquiry is always in order.
Finally, suppose the skeptic thinks that closure 2 The standard argument for Academic skepti-
holds because the evidential relationship is cism with the conclusion that reasoning
depicted by Pattern 3. Here the skeptic is claiming cannot produce any type of justified belief,
that we must first eliminate the skeptical hypoth- including provisionally justified belief, has
esis in order to arrive at the proposition that there been shown to be an instance of the general
is a table before usY That evidential prerequisite constraints that Pyrrhonists believe apply to all
is not immediately evident and, hence, requires finite arguments. They either rest on arbitrary
some reasons. Would we have to eliminate every assumptions or they beg the question.

One important caveat: I have argued that no some of the, as yet, unsupported suppositions
belief is unconditionally justified. But even that in this chapter. And maybe I begged the ques-
conclusion has to be taken only provisionally - tion. Prudence requires that we view the two
if I am right. My reasoning here has been finite main points of the chapter as only provisionally
in length. Maybe there aren't good reasons for justified.


I want to thank Anne Ashbaugh, Michael 4 I chose to put the argument in terms of degrees
Bergmann, Claudio de Almeida, Mylan Engel, of justified belief rather than knowledge for
Michael Huemer, Michael Lynch, Steven Luper, two reasons: 1) the argument for Academic
Kenton Machina, Stephen Maitzen, Robert skepticism about knowledge usually depends
Martin, Brian McLaughlin, George Pappas, John upon assuming that knowledge is at least ade-
Post, Bruce Russell and Thomas Vinci for their quately justified true belief and it is the sup-
help with various aspects of this chapter. Earlier posed lack of fulfilment of the justification
versions of some ancestor parts of it were condition that leads to the denial of knowledge,
presented in March 1999, at the Mississippi and 2) it is the power of reasoning to justify
Philosophical Association and in June 1999, at beliefs that is the primary concern of this chap-
the Bled Conference in Epistemology, and at the ter and, consequently, even if knowledge did
Rutgers Summer Institute for Minority Students not entail adequately justified beliefs, if there
in Philosophy, July 2000, at Acadia University, were a good argument showing that our beliefs
September 2000, and at the Illinois Philosophical were never justified to any positive degree, my
Association, November 2000. The discussions claim that reasoning can make a belief provi-
that followed helped me in refashioning the argu- sionally justified would have been undercut.
ments in this version. In addition various parts 5 Hume claimed that the only way to deal with
of this chapter draw on Klein (1999), (2000b) skepticism was simply not to think about it.
and (2000a). That might work for some, but not for me -
1 This is the view developed in Lehrer (1997) and it didn't work for Hume either!
and (2000, esp. pp. 142-4). 6 Strictly speaking, it could be that beliefs which
2 I think that the difficult task in providing are justified are closed under entailment
an account of what makes circular reason- because justified beliefs are a subset of beliefs
ing fallacious is to spell out clearly what is and the subset could be closed while beliefs are
meant by the conclusion being "employed in" not. But the same objections to the closure of
a premise. beliefs, simpliciter, seems to apply to justified
3 The only way out of this predicament is to beliefs as well.
withhold belief about whether we have been 7 Robert Audi gives some other examples in
successful in settling an issue by the employ- which it appears that one can have sufficient
ment of reasoning. But note that I am not evidence for the entailing proposition but not
claiming that if we believe that we know for the entailed one (1988, esp. pp. 77 ff.).
something, we will, in fact, ignore contrary I discuss these examples in some detail in Klein
evidence if (or when) it appears. At that (1995, pp. 213-36). Briefly, I think these
point we could both lose our belief that we counter-examples fail for the same reasons
know and acknowledge the contrary evi- that Dretske's proposed counter-examples fail.
dence. Thus, there is no real "Kripke prob- 8 I have argued for that in Klein (1984), in Klein
lem" - the alleged consequence of believing (1995) and in Klein (2000a).
that we know, namely that we become per- 9 The modes of relativity and discrepancy reca-
manently (and perhaps even rationally) con- pitulate passages in Sextus's chapter
vinced of the truth of what we believe we "Concerning the Ten Modes" in Sextus (1933).
know. I have discussed this elsewhere; see Sextus attributes this formulation of the
Klein (1984). modes to Aenesidemus.

10 The term "infinitism" is not original to me. think that every member of the set of basic
To the best of my knowledge, the first use of propositions is true?"
a related term is in Moser (1984), in which 16 There is one move Fred might make here.
he speaks of"epistemic infinitism': Also, Post Steven Luper suggests that it is rational to
(1987, p. 91) refers to a position similar to accept foundational beliefs even though
the one I am defending as the "infinitist's they cannot be supported by reasons. Here
claim". is a close paraphrase of his argument. The
11 The reason for the "non-repeating" condi- epistemic goal is to acquire a complete and
tion is that were the propositions to repeat, accurate picture of the world. Granted, at
the result would be a form of coherentism - base our reasons are arbitrary but "an
infinitely long circles. injunction against believing anything ...
12 Strictly speaking, there is a fourth possibility, would obviously make it impossible for us
namely that there are foundational proposi- to achieve the goal of arriving at a complete
tions and that there are an infinite number and accurate understanding of what is the
of propositions between the foundational case ... Indeed, given that our ultimate
one and the one for which reasons are ini- beliefs are arbitrary, it is rational to adopt
tially being sought. Interestingly, such a management principles that allow us to
hybrid view might be indistinguishable in retain these foundational yet arbitrary views,
practice from infinitism and, hence, not sub- since the alternative is to simply give up on
ject to the "foundationalist's dilemma" to be the attempt to achieve the epistemic goal"
discussed later. Thus, I think for our pur- (Luper-Foy 1990, p. 45). Briefly, the claim is
poses we can treat this as a form of infinitism that since the goal of an epistemic agent is to
or an acceptable form of foundationalism acquire a complete and accurate picture of
because it is the ability of the three patterns the world, accepting a basic, though arbi-
of reasoning (foundationalism, coherentism trary, reason is rational since if one did not
and infinitism) to provide a basis for rational accept it, there would be no possibility of
practice that is the criterion which will deter- attaining the goal. It is "rational to do and
mine whether any of the patterns is accept- believe things without reason" (p. 40)
able. As far as I know this possibility has because if we did not, we could not attain
never been explored; it might be worthwhile our goal.
to do so. But I don't think Fred can employ this
l3 I put it that way in order to make clear that line of reasoning. For if Fred's basic beliefs
foundationalism can embrace some aspects are arbitrary, that is if there is no available
of coherentism. Propositions with only min- reason for thinking that they are even some-
imal justification can mount up, so to speak, what likely to be true, then Fred, being a
by gaining extra credibility. Thus, the defini- foundationalist, would have no reason for
tion of foundationalism includes both weak thinking that any of the non-basic proposI-
and strong foundational ism as characterized tions are true either. If, ultimately, it is
in BonJour (1978, pp. 1-l3). rational to accept some "basic" proposition,
14 In other places I have argued that founda- b, for prudential considerations, then no
tionalism is false; see Klein (1999). epistemic warrant can be transferred to the
15 That is, there is no reason that can be given non-basic ones from the basic ones. Thus, if
that does not ultimately depend upon other the non-basic propositions have any epis-
basic propositions. There could be reasons temic warrant at all, it must arise completely
for believing the basic ones for they could from some source other than the basic
cohere with other propositions and coher- beliefs. And that view isn't foundationalism.
ence could add some degree of warrant. But Coherence naturally suggests itself. But as we
Fred would see that Sally would then just ask will see soon, that solution to the regress
about the set of basic propositions. In other problem can't provide a model for rational
words, she would ask, "What makes you practice.

17 I want to thank Steven Luper for his com- 165-91). In addition, BonJour (1978) dis-
ments that forced me to be clearer about tinguishes between linear and non-linear
what counts as an arbitrarily adopted propo- coherentism. That distinction parallels the
sition. I take it that a proposition is arbitrar- one here between warrant-transfer coher-
ily accepted just in case there is no better entism and warrant-emergent coherentism.
reason for accepting it than denying it, but it 22 The original passage is as follows:
is accepted anyway. Thus, in the argument
considered earlier against closure (and hence Suppose that all justification is inferential.
against Academic skepticism) accepting the When we justify belief A by appeal to belief B
premise that closure entails Pattern 1 type and C, we have not yet shown A to be justified.
evidential structures is arbitrary because We have only shown that it is justified if Band
there are better reasons for denying that than C are. Justification by inference is conditional
there are for accepting it (and hence there justification only; !\s justification is condi-
are no better reasons for accepting than tional upon the justification of Band C. But if
denying it). Also, when the infinitist stops all justification is conditional in this sense, then
giving her reasons for beliefs (as she must), nothing can be shown to be actually non-
her last reason given need not be arbitrary conditionally justified.
because she might have better reasons for
believing it than denying it - although she I modified the passage to avoid what seems to
hasn't yet given them. me to be an unfortunate level confusion that
18 See, for example, Cohen (1987, 1988), Lewis conflates one's being justified with showing
(1979, 1996), Wittgenstein (1977) and DeRose that one is justified. I also changed the pas-
(1995, 1992). There are also hints at such a sage to make clear that the alleged problem is
view in Aristotle (Metaphysics, 1006a-1O 11 b). with the infinite chain, per se, as opposed to
19 I have considered (and rejected) the contex- the chain diverging because more than one
tualist response to Academic skepticism in proposition is offered as a reason for a belief.
Klein (2000a). 23 Once again, I am indebted to Steven Luper
20 See Donald Davidson (1986) for this type of for this point.
defence of the claim that certain methodo- 24 See Klein (1999) for a fuller defence.
logically basic propositions must be "true in 25 Of course there is a third option - neutral-
the main': Davidson, of course, was defend- ity, that is being neither optimistic nor pes-
ing what he took to be a brand of coherent- simistic. The reasons presented here for
ism. But as we will soon see, some forms of thinking that infinitism is not inherently
coherentism are really nothing but founda- pessimistic can be applied to the neutral
tionalism in disguise. view as well.
21 I do not think the expressions "warrant- 26 I point to various alternative accounts of
transfer coherent ism" or "warrant-emer- what would make a proposition one which is
gent coherentism" are original to me. Nor is correctly called forth in Klein (1999). There I
the distinction between the two types of call such propositions "objectively available"
coherentism. But I do not recall where I first as reasons.
ran across the use of those expressions or 27 Keith Lehrer makes that claim on behalf of
the discussion of these issues. Ernest Sosa the skeptic in Lehrer (1971, pp. 292-4).
does distinguish between various forms of Lehrer no longer accepts this argument
coherentism. In addition, he argues that (I think); see Lehrer (1974, pp. 238-40) and
what I call the "warrant-emergent" form is Lehrer (2000, pp. 132-7).
actually a form of what he calls "formal 28 Here is the argument: both (-p & q) and
foundational ism". Thus, the claim that some (-p & -q) are contraries of p. If the denials of
forms of coherent ism are actually forms of both are required to be in the evidence for p,
foundationalism is not original to me; see then that evidence entails p because
Sosa (1980), reprinted in Sosa (1991, pp. H -p&q), -( -p&-q)} entails p.


Aristotle. 1994. Posterior Analytics, trans. 1. Barnes, Lehrer, Keith. 1971. "Why Not Skepticism?;'
2nd edn (Oxford: Clarendon Press). Philosophical Forum, pp. 206-98.
Audi, Robert. 1988. Belief, Justification, and - - . 1974. Knowledge (Oxford: Oxford University
Knowledge (Belmont: Wadsworth). Press).
BonJour, Lawrence. 1978. "Can Empirical - - . 1997. SelfTrust: A Study ofReason, Knowledge,
Knowledge Have a Foundation?;' American and Autonomy (Oxford: Clarendon Press).
Philosophical Quarterly 15, pp. 1-13. - - . 2000. Theory of Knowledge, 2nd edn
- - . 1985. The Structure of Empirical Knowledge (Boulder, CO: Westview Press).
(Cambridge, MA: Harvard University Press). Lewis, David. 1979. "Scorekeeping in a Language
Cohen, Stewart. 1987. "Knowledge, Context, and Game;' Journal of Philosophical Logic 8,
Social Standards:' Synthese 73, pp. 3-26. pp.339-59.
- - . 1988. "How to be a Fallibilist." Philosophical - - . 1996. "Elusive Knowledge;' Australasian
Perspectives 2, pp. 91-123. Journal of Philosophy 74, pp. 549-67.
Dancy, Jonathan. 1985. Introduction to Contem- Luper-Foy, Steven. 1990. "Arbitrary Reasons;' in
porary Epistemology (Oxford: Blackwell). M. Roth and G. Ross (eds) Doubting:
Davidson, Donald. 1986. "A Coherence Theory of Contemporary Perspectives on Skepticism
Truth and Knowledge," in Henrich 1983. (Dordrecht: Kluwer), pp. 39-55.
DeRose, Keith. 1992. "Contextualism and Moser, Paul. 1984. "A Defense of Epistemic
Knowledge Attributions;' Philosophy and Intuitionism;' Metaphilosophy 15, 3, pp. 196-204.
Phenomenological Research 52, pp. 913-29. Post, John. 1987. The Faces of Existence (Ithaca,
- - . 1995. "Solving the Skeptical Problem;' NY: Cornell University Press).
Philosophical Review 104, pp. 1-52. Sextus Empiricus. 1933. Outlines of Pyrrhonism,
Dretske, Fred. 1970. "Epistemic Operators;' trans. R. G. Bury (London: W. Heinemann,
Journal of Philosophy 67, pp. 1007-23. Loeb Classical Library).
Klein, Peter. 1995. "Skepticism and Closure: Why Sosa, Ernest. 1980. "The Raft and the Pyramid,"
the Evil Genius Argument Fails;' Philosophical Midwest Studies in Philosophy 5, pp. 3-25.
Topics 23, pp. 213-36. - - . 1991. Knowledge in Perspective (Cambridge:
--.1999. "Human Knowledge and the Infinite Cambridge University Press).
Regress of Reasons;' Philosophical Perspectives Williams, John. 1978. "Justified Belief and the
13, Epistemology, pp. 297-326. Infinite Regress Argument;' American
- - . 2000a. "Contextualism and the Real Nature Philosophical Quarterly 15, 4, pp. 311-12.
of Academic Skepticism," Philosophical Issues Wittgenstein, Ludwig. 1977. On Certainty
10, pp. 108-16. (Oxford: Blackwell).
- - . 2000b. "The Failures of Dogmatism and
a New Pyrrhonism," Acta Analytica 15, 24,

Epistemological Realism

Michael Williams

Generality and Epistemic Priority In trying to explain how what might otherwise
seem to be truisms take on a surprising signifi-
Although a defender of the naturalness of sceptical cance, it is natural to look first to the traditional
doubts must hold that foundationalism is a by- epistemologist's aim of assessing the totality of
product of scepticism, not a presupposition, so our knowledge of the world. Because he wants to
far we have seen nothing to suggest that the case explain how we are able to know anything at all
for scepticism can be understood apart from the about the external world, his plan is to assess all
doctrine of the priority of experiential knowledge such knowledge, all at once. But surely, the argu-
over knowledge of the world. This result would ment now goes, if we are to understand how it is
not be decisive if this essential doctrine could possible for us to know anything at all about
itself be derived from the truistic elements in the external reality, we must trace that knowledge to
sceptic's arguments. But we have seen nothing to knowledge we should still have even if we knew
suggest this either. On the contrary, everything nothing about the world. No explanation of how
points the other way. we come to have knowledge of the world that
This leaves one option: to see how the truistic depended on our already having some would
elements in the sceptic's arguments take on scep- show the required generality: it would not be an
tical significance, we must look to the distinctive explanation of how we have any such knowledge.
character of the traditional epistemological project. But this is as good as to say that, once we accept
The sceptic (or traditional epistemologist) must the legitimacy of the epistemologist's question -
argue that, in the context of a distinctively philo- and we have seen no reason to suppose that it is
sophical investigation of our knowledge of the unintelligible - we must also accept the priority
world, the crucial ideas about epistemic priority of experiential knowledge, since experiential
are forced on us by our ordinary understanding of knowledge is what remains when knowledge of
knowledge or justification. If he can do so, he will the world is set aside.
have rebutted the charge that he simply takes This is Stroud's view, which explains why he
them for granted. thinks that the diagnosis of scepticism that traces
it to foundationalism gets things upside down.
According to Stroud:

Originally published in M. Williams, Unnatural Doubts What we seek in the philosophical theory of
(Oxford: Blackwell Publishers, 1991), pp. 83-93,101-19, knowledge is an account that is completely gen-
121-4,129-39. eral in several respects. We want to understand

how any knowledge at all is possible - how knowledge of the world, an otherwise innocuous
anything we currently accept amounts to knowl- claim gives the sceptic what he needs.
edge. Or less ambitiously, we want to understand Acceptance of the totality condition on a
with complete generality how we come to know properly philosophical understanding of our
anything in a certain specified domain. I knowledge of the world is also the deep source of
the epistemologist's dilemma, for the dilemma
It is the distinctively philosophical goal of under- springs from a fatal interaction of the totality
standing certain kinds of knowledge with "com- condition with the objectivity requirement. This
plete generality" that leads to attempts to ground is the requirement that the knowledge we want to
knowledge of a given kind on some "epistemo- explain is knowledge of an objective world, a
logically prior" kind of knowledge, and the reason world that is the way it is independently of how it
is that no other strategy will yield the right kind appears to us to be or what we are inclined to
of generality. Unfortunately, the lesson of scepti- believe about it. Now, as we have seen, the totality
cism seems to be that such attempts are bound to condition requires us to try to trace our knowl-
fail, so that there is no hope of understanding edge of the world to something more fundamen-
human knowledge in general. tal, which can only be experiential data. But, as a
We can characterize the unusual generality of sceptical argument along Ayer's lines reveals, it is
the traditional epistemological undertaking by impossible to explain how such data could ever
saying that the traditional epistemologist imposes function as evidence. They cannot be linked
a totality condition on a properly philosophical empirically with any facts about the world for, in
understanding of our knowledge of the world. accepting such linkage, we would be crediting
Acceptance of this condition, I believe, is what ourselves with knowledge of the world, in viola-
lies behind the feeling that arguments concern- tion of the totality condition. On the other hand,
ing conceptual points are unfair to the sceptic. conceptual connections between experiential
Purely conceptual points - the neutrality of expe- data and worldly fact seem to be ruled out by the
rience or the "non-dreaming" implication of familiar thought-experiments that the sceptic
ordinary perceptual knowledge - have no intrinsic appeals to to establish the neutrality and auton-
epistemological significance. Moreover, since such omy of experience. And if, in a desperate attempt
sceptical significance as they possess depends to avoid scepticism, we insist on such connec-
entirely on a tacit commitment to the priority of tions, we make the way the world is depend on
experiential knowledge over knowledge of the how it appears to us, in violation of the objectiv-
world, they themselves give no grounds for ity requirement. Accordingly, in the context of the
accepting any such general relation of epistemo- attempt to assess the totality of our knowledge of
logical priority. But perhaps they do not have the world, it seems impossible either to respect or
to. Perhaps the very nature of epistemological violate the objectivity requirement: whatever we
investigation forces us to recognize that relation; do looks like succumbing to the sceptic.
and once it is recognized, the sceptic's truistic Nevertheless, although the epistemologist's
conceptual points are all he needs to reach his dilemma arises from the interaction of the total-
conclusion. ity condition and the objectivity requirement,
> For example, one might argue that the (truis- I take the totality condition to be fundamental.
tic) claim that my knowing (perceptually) that P Many philosophers would disagree, for they see
implies my knowing that I am not dreaming that the objectivity requirement, with its commitment
P is not equivalent to the claim the sceptic must to a "realistic" view of truth, as the deep source of
assimilate it to: that my knowing that P requires sceptical problems. But it is not clear, to me at
my being able to rule out the possibility that I am least, that the objectivity requirement, any more
dreaming that P independently of my knowledge than its relative the neutrality of experience, has
that P (or indeed anything like it). But the sug- any particular sceptical potential outside the
gestion now is that the totality condition, rather context of an assessment of worldy knowledge
than the non-dreaming condition alone, is what governed by the totality condition.
imposes the crucial restriction. So, in the context I say that the totality condition is fundamen-
of the traditional attempt to understand our tal. More strictly, however, what is fundamental

is the attempt to conduct an assessment of our when he tries to connect his "naturalized" episte-
knowledge of the world in the light of that condi- mology with traditional sceptical problems.
tion. If the priority of experiential knowledge It seems, then, that something very like foun-
over knowledge of the world is implicit in the tra- dationalism falls out of a methodological con-
ditional epistemological project, this is not solely straint on a properly philosophical examination
on account of that project's unusual generality. of knowledge of the world. So we have, apparently,
Also crucial is the kind of understanding it sug- found what we were looking for: a defence of the
gests we seek. As Quine has argued, if all we want claim that foundationalism is a by-product of
is some kind of causal or developmental account scepticism, not a presupposition. When this pos-
of the emergence of our knowledge of the objec- sibility was first mooted, I suggested that it would
tive world, there is nothing viciously circular in have to turn out that scepticism and foundation-
our appealing to what we now know about the alism have a common root. We have now located
world in an explanation of how we came to be in that common root in the attempt to gain a certain
our current position. 2 And where there is no kind of understanding of our knowledge of the
threat of circularity, there is no pressure to accede world. In effect, we have glossed Hume's thought
to a general doctrine of epistemic priority. that we set foot on the road to scepticism as soon
As Quine is of course well aware, traditional as we ask distinctively philosophical questions
epistemology is under pressure to accept such a about knowledge. True, this will not yield a defence
doctrine because it seeks a different kind of of the naturalness of sceptical doubts unless, as
understanding. Its aim is to explain how it is that Hume thought, that form of questioning is itself
our beliefs about the world amount to knowl- fully natural. However, even on this point, the scep-
edge. Thus when Stroud says that what we want tic has strengthened his position. It is hard to see
from a theory of knowledge is an account of how how there could be anything unintelligible in what
our knowledge of the world emerges out of some- seems only to be an attempt to understand knowl-
thing that is not our knowledge of the world,3 edge in an unusually general way, so the prospects
he does not mean that we want an explanation for a convincing therapeutic diagnosis of scepti-
of how our current way of looking at things cism seem bleak. But it is not obvious offhand that
developed out of some previous way: i.e. out of the prospects for a satisfactory theoretical diagno-
knowledge (or what our ancestors thought of as sis are any brighter, for how can mere generality
knowledge) that is not ours. This is a task for entail extensive theoretical commitments?
historians and anthropologists. Nor is he think- This is not all. Suppose that we agree that the
ing of an account of how our knowledge emerges traditional epistemological project leads inevita-
out of something that is not our knowledge. bly to the conclusion either that we have no
Quine's idea of a naturalized epistemology is a knowledge or that, if we do, we will never under-
gesture in this direction, for it is supposed to issue stand how we do; and suppose we insist that, since
in a causal explanation of how our interactions this is its outcome, it must involve some distor-
with the environment lead us to form certain tion of our epistemological position: can we say
beliefs; and if there is a worthwhile project here, it that identifying this distortion will let us see how
is presumably one for psychologists and neuro- knowledge is possible after all? Stroud suggests
physiologists. What is missing from both these not. We should not think that:
projects is the idea of an assessment. Each could
as well, in fact more properly, be offered as an if we did come to see how and why the epistemo-
account of the emergence of our beliefs. But only logical enterprise is not fully valid, or perhaps
a legitimating account of the basis or emergence not even fully coherent, we would then possess a
of our beliefs will give an account of our knowl- satisfactory explanation of how human knowl-
edge. The sort of theory Stroud has in mind is edge in general is possible. We would have seen,
therefore one that traces our knowledge of the at best, that we cannot have any such thing. And
world to something that is ours, and that is knowl- that too, I believe, would leave us dissatisfied!
edge, but not knowledge of the world. What could
this be except experiential knowledge? Even This is a powerful objection to any theoretical
Quine is forced to something like this position diagnosis of scepticism. Attempts to answer the

sceptic directly run into the epistemologist's direct refutation of scepticism, they call for a
dilemma. But, if Stroud is right, attempts at diagnosis of the sceptic's questions which will
diagnostic responses meet a similar fate. Suppose reveal them, first impressions to the contrary, as
we find that we cannot hope to ground our less than fully coherent. Even Stroud, who thinks
knowledge of the world in the way that tradi- our most pressing need as epistemologists is to
tional epistemology has invited us to, because of understand how traditional epistemological
some defect in the ideas about justification inquiry misrepresents our epistemic position, if it
involved in the notion of even trying: we would does, seems not to doubt the existence of its
still not have explained to ourselves how it is that objects. For the idea that there is something called
we ever come to know anything about the world. "our epistemic position" is just another aspect of
Unless we show that the sceptic's question is the idea that there is such a thing as "human
actually unintelligible, it will remain dissatisfy- knowledge" or "our view of reality." But is there?
ingly unanswered. So this is our new dilemma: if Or are there fewer things in heaven and earth
the traditional epistemological project is coher- than are dreamt of in our epistemology?
ent, it is doomed to fail; and if it isn't, we are still Now, it is tempting to use "human knowledge"
left in a position hard to distinguish from scepti- and "our knowledge of the external world" as
cism. It may be scepticism at second order, but it though it were obvious that such phrases pick out
is scepticism for all that. We may have knowledge reasonably definite objects of study. But it isn't
of the world, but we will never be able to explain obvious, or shouldn't be. We can talk of "our
to ourselves how we do. We may know things knowledge of the world," but do we have any
about the world, but we will never know that we reason to suppose that there is a genuine totality
know them. here and not just a loose aggregate of more or less
unrelated cases? My sense is that the totality con-
dition is far more problematic than it first seems.
Knowledge as an Object of Theory Consider, for example, Nagel's characteriza-
tion of the aim of epistemology as "to form a con-
In asking whether there is such a thing as knowl- ception of reality which includes ourselves and
edge of the world, I am not asking the very same our view of reality among its objects."5 This off-
question the sceptic asks but one that I think cuts hand allusion to "our view of reality" takes a lot
deeper. I am asking how we have to think about for granted. To suppose that there is such a thing
"knowledge of the world" for that phrase to pick as "our view of reality;' which might then be the
out a proper object of theory. So if it sounds too "object" of a single theoretical enterprise, is to
strange even to hint that there might not be any- assume that human knowledge constitutes some
thing for the theory of knowledge to be a theory kind of surveyable whole, an idea that is not, on
of, my question can be rephrased. What matters is the surface, very promising. There are no clear
whether "our knowledge of the world" picks out criteria for individuating beliefs and, even if
the kind of thing that might be expected to be there were, it is far from clear that there would be
susceptible of uniform theoretical analysis, so any systematic way of enumerating all the things
that failure to yield to such analysis would reveal we believe. Phrases like "our system of beliefs"
a serious gap in our understanding. and "our view of reality" are so vague that we
To raise these questions is to begin to examine cannot be confident they refer to anything.
a move that gets made before epistemological Nothing changes if we pull back to narrower
arguments, and particularly sceptical arguments, categories such as knowledge of the external
even get started. This is the introduction of the world. When it comes to such "specified domains;'
objects of epistemological inquiry. We shall be whether there is anything to understand will
trying to isolate views that, for the most part, depend on how the domains are specified. To try
even the most determined anti-sceptics share to understand all knowledge in the standard epis-
with their adversaries. Philosophers who respond temic domains is to suppose that the beliefs in
to scepticism do not doubt that there is some- those domains hang together in some important
thing to defend against the sceptic's attacks. If way. But how? "Knowledge of the external world"
they are dubious about our prospects for giving a covers not only all the natural sciences and all <2f

history, it covers all everyday, unsystematic factual The very nature of the traditional project
claims belonging to no particular investigative demands that the principles in question be all-
discipline. Since, even within a single subject, pervasive. For example, if we are to assess the
theories, problems and methods tend to prolifer- totality of our beliefs about the world, there must
ate with the progress of inquiry, so that even the be principles that inform all putative knowledge
most systematic disciplines tend to become less of the world as such. But what could they be?
rather than more unified, it is doubtful whether I take it to be obvious that, in one way, our beliefs
we can take a synoptic view of physics, never do not show any kind of theoretical integrity.
mind everything we believe about the external They do not, that is, add up to an ideally unified
world. It is not obvious that it makes sense even theory of everything. There is no way now, and
to try. none in prospect, of integrating all the sciences,
Recall Stroud's claim that in the philosophical much less all of anyone's everyday factual beliefs,
study of human knowledge we want "to under- into a single coherent system: for example, a
stand how any knowledge at all is possible - how finitely axiomatized theory with specified rules of
anything we currently accept amounts to knowl- inference. In this way, Nagel's phrase "our view of
edge." He finds that engaging in this project "feels reality" borders on the absurd. We have not got a
like the pursuit of a perfectly comprehensible "view of reality" but indefinitely many. The idea,
intellectual goal." Perhaps it does once we have taken for granted by coherence theorists of justi-
grown familiar with theoretical ideas that we shall fication, that we have a "system" of beliefs ought
be examining shortly. But we must try to recover to be suspect.
some naivete here. Then I think we see that, when "Our beliefs," then, do not amount to a single,
we first encounter the challenge to show how any integrated "view of reality." They are not topi-
knowledge of the world is possible, we cannot tell cally integrated. But this need not be fatal to the
whether we have been given a perfectly compre- project of understanding human knowledge in
hensible goal or not. In fact, the obvious diffi- general. For even if our beliefs are not topically
culty in commanding a synoptic view of our integrated, they might be epistemologically inte-
worldly beliefs suggests that we haven't. We grated. This is to say: they might be subject, in so
cannot, therefore, just see whether the epistemo- far as they are meant to be justified or to amount
logical challenge make sense. What we can do, to knowledge, to the same fundamental, episte-
however, is to ask how we might make sense of it. mological constraints. This is what is usually
I think that we can find a somewhat oblique suggested, or rather assumed. Thus Descartes
recognition of this problem even in Descartes. ties his pre-critical beliefs together, thereby
Descartes admits that getting to a general doubt constituting their totality as an object of theo-
by questioning his beliefs one at a time would not retical inquiry, by tracing them all to "the senses."
be easy: perhaps the examination would never be No matter how topically heterogeneous, and no
completed. Hume too dismisses a piecemeal matter how unsystematic, his beliefs have this
approach as a "tedious lingering method."7 But much in common: all owe their place to the
these grudging concessions are misleading: for authority of the senses. If this authority can be
they imply that the main obstacles to going over called in question, each loses its title to the rank
our beliefs seriatim are time and energy, whereas of knowledge.
the question is certainly not one of convenience. We have seen that this talk of "the senses" is
If we are to make sense of the project of explain- poised between a causal truism and a contentious
ing how anything we believe about the world epistemological doctrine. Now we see more clearly
amounts to knowledge, we need a way of reduc- why the epistemological doctrine is and must be
ing our beliefs to order. We have to bring them what is intended. Only by tracing our beliefs about
under principles or show them as resting on the world to a common "source;' which is to say a
commitments that we can survey. We must reveal common evidential ground, can we make "beliefs
some kind of theoretical integrity in the class of about the world" the name of a coherent kind.
beliefs we want to assess. H If we can do this, human In the absence of topical integration, we must
knowledge is a possible object of theoretical look to epistemological considerations for the
investigation. But not otherwise. theoretical integrity we require.

Hume may have seen, though perhaps dimly, of externalism in the theory of knowledge. By
that an epistemologically based form of theo- suggesting that our capacity for knowledge
retical integrity is a precondition for a properly depends on our situation in the world, and not
general, hence "philosophical;' understanding of just on our own "internal" capacities, externalism
human knowledge. He compares assessing par- challenges the idea of our "epistemic position" as
ticular beliefs and particular sciences one at a an autonomous object of theory. If our epistemic
time to a strategy of "taking now and then a position is not something that can be investigated
castle or village on the frontier"; and he contrasts without knowing something about how we are
this "tedious" method with marching up to "the placed in the world, there can be no question of
capital or center of these sciences, to human our assessing the totality of our knowledge of the
nature itself." In explaining the principles of world on the basis of insights into our epistemic
human nature, he tells us, "we in effect propose a position. Perhaps we do not even have a fixed
compleat system of the sciences." But the com- epistemic position. And if we find that we do not,
pleteness envisaged does not involve topical it is doubtful whether we will be able to retain a
integration. It derives rather from the fact that clear conception of "our knowledge of the world"
all sciences, whatever their subject matter, "lie as an appropriate object of theory.
under the cognizance of men, and are judged of Unlike Hume, Descartes aspires to topical as
by their powers and faculties." Their subjection well as epistemological integration: hence his
to the same underlying epistemological con- metaphor of the tree of knowledge whose roots
straints, rooted in our "powers and faculties" is are metaphysics, trunk physics, and branches
thus what makes possible a sweeping evaluation medicine, mechanics, and morals, a figure that
of "all the sciences."9 contrasts interestingly with Hume's citadel of
Hume sees the fact that all sciences lie "under reason. But even for Descartes, topical integration
the cognizance of men" as showing that all are "in is something to be achieved rather than assumed.
some measure dependent on the science of MAN:' His initial survey of his beliefs takes for granted
But it seems clear that the science of man is not, only their epistemological integrity. As is familiar,
or ought not to be, dependent on the other he makes the point in terms of the metaph~r of
sciences. (Hume is apologetic about his occa- foundations: undermine the foundations and the
sional excursions into natural philosophy.) This whole edifice crumbles. The metaphor is a very
asymmetry belongs to the logic ofHume's project, natural one for, as we have seen, there is a clear
indeed to the logic of the traditional epistemo- sense in which epistemology, understood as the
logical enterprise. Since he is attempting, with a attempt to comprehend how any knowledge is
view to its reform, a wholesale assessment of our possible, is intrinsically foundational. To see
knowledge of both the physical and the moral human knowledge as an object of theory, we must
world, he cannot take any of that knowledge for attribute to it some kind of systematic basis. This
granted. This means that it must be possible to may involve inference from some class of funda-
investigate our "powers and faculties," the episte- mental evidence-conferring beliefs, as traditional
mological aspect of the human condition, with- foundationalists maintain; or it may involve gov-
out relying on any worldly knowledge. Our ernance by certain "global" criteria of explanatory
epistemological self-knowledge must be both integration, as coherence theorists think. But
autonomous and fundamental. Thus the project something must regulate our knowledge of the
of assessing the totality of our knowledge of the world: something that we can identify and exam-
world does more than presuppose that experien- ine independently of any such knowledge. We
tial knowledge is in some very deep way prior to should therefore not be too eager to oppose the
knowledge of the world. It also assigns a definite account of scepticism that traces it to the general-
privilege to knowledge of such epistemological ity of the epistemological enterprise to that which
facts. These features of the traditional project traces it to foundational ism. (Nor, for that matter,
point to very extensive theoretical commitments. should we be too eager to oppose foundationalism
The fact that the traditional epistemological to the coherence theory.) If we give up the idea of
enterprise is committed to the autonomy of epis- pervasive, underlying epistemological constraints;
temology sheds further light on the significance if we start to see the plurality of constraints that

inform the various special disciplines, never mind theory that it doesn't cover the tremendous "heat"
ordinary, unsystematic factual discourse, as genu- produced in my mouth by a chicken vindaloo,
inely irreducible; if we become suspicious of the never mind the heat often generated by philo-
idea that "our powers and faculties" can be evalu- sophical arguments. We don't complain that,
ated independently of everything having to do since the theory doesn't apply to hot curries or
with the world and our place in it: then we lose heated arguments, it fails to explain heat in a sat-
our grip on the idea of "human knowledge" as an isfactorily general way.
object of theory. The clear contrast between cas- Given that we want to know whether there is
tles on the frontier and the fortress at the centre any such thing as, say, "our knowledge of the
disSDlves. Perhaps there is no capital, each prov- world;' this kind of failure may seem too weak to
ince, as Wittgenstein said of mathematics, having be of interest. Failure to take in hot curries and
to take care of itself. The quest for an understand- heated arguments does not tempt us to say that
ing of human knowledge as such, no longer feels there is no such thing as heat. But we could say
like "the pursuit of a perfectly comprehensible that there is no such thing as nominal heat, the
intellectual goal." nominal kind being merely nominal. We can tie
The same is true of more modest aims, such as together some of the examples of heat and,
understanding how our beliefs about the external having done so, treat them as the only genuine
world amount to knowledge. As a way of classify- examples, discarding the others as resembling the
ing beliefs, "beliefs about the external world" is genuine examples only superficially, hence as not
only quasi-topical, bringing together beliefs really, but only metaphorically, hot. This is,
belonging to any and every subject, or no well- indeed, what Bacon himself goes on to do when
defined subject at all. They are united only by he argues that heat is a form of motion. Anyway,
their supposed common epistemological status. it is clear that there need be no theory of all the
The essential contrast to "beliefs about the exter- things commonly called "hot": a hot curry is hot
nal world" is "experiential beliefs" and the basis even when it has gone cold. Nor need the lack of
for the contrast is the general epistemic priority such a theory because for intellectual dissatisfac-
of beliefs falling under the latter heading over tion. It is just another example of an ordinary
those falling under the former. "External" means principle of classification failing to cut nature at
"without the mind"; and it is taken for granted the joints. By the same token there does not have
that we have a firmer grasp of what is "in" the to be a theory of all the things normally called
mind than of what is outside it. "examples of knowledge." And if there isn't, it has
There is no doubt that this epistemological to be shown that this reveals a lack. It may be that
distinction is readily mastered: readily enough for there is no such thing as knowledge (or knowl-
arguments based on it to strike us as "immedi- edge of the external world, etc.) in just the way
ately gripping:' But a teachable distinction does that there is no such thing as Bacon's nominal
not guarantee theoretical integrity in the kinds of heat.
things distinguished. There are various ways of All this notwithstanding, I agree that the
failing. I discuss two examples in this section and example of heat doesn't get me very far. All that
one in the next. happens in this case is that a nominal kind fails to
My first example illustrates a relatively mild coincide exactly with a theoretically coherent
form of failure. In his natural history of heat, kind. So I move to my second example: the sup-
Bacon gives a long list of examples of heating. It posed division of sentences into analytic, or true
includes examples of heating by radiation, fric- by virtue of meaning, and synthetic, or true by
tion, exothermic reactions, and by "hot" spices virtue of fact. Quine is famously sceptical about
that "burn" the tongue. to Everything he mentions this distinction because he is dubious about the
is ordinarily said to involve "heat;' so we cannot atomistic conception of meaning that he takes to lie
deny that his list reflects ordinary usage. But what behind it.l! Quine's view of meaning is holistic -
we have here is a clear case in which a nominal the meaning of a given sentence depends on its
kind, comprising all the things commonly called role in a wider theory - and this holistic concep-
"hot;' has no automatic right to be considered a tion of meaning suggests that there is no privi-
natural kind. It is no objection to the kinetic leged way of distinguishing a theory's "meaning

postulates" from its empirical assumptions, any hot things. Explaining theoretically significant
more than there is a way of determining which out kinds this way is typical of scientific realism. For
of alternative complete axiom sets is the right one. the scientific realist, deep structural features of the
Against Quine, Grice and Strawson argue that elementary components of things determine the
the analytic/synthetic distinction must be genu- boundaries of natural, as opposed to merely nom-
ine and significant because it is teachable in such inal or conventional, kinds. This suggests an ana-
a way as to enable the student to apply it to new logy. Since, if human knowledge is to constitute
cases. 12 The reply, well known by now, is that all a genuine kind of thing - and the same goes for
kinds of dubious distinctions have proved to be knowledge of the external world, knowledge of
teachable in this way, for even terms belonging to other minds, and so on - there must be underlying
a false theory can admit of consensual application epistemological structures or principles, the tradi-
on the part of those who accept it. If the fact that, tional epistemologist is committed to epistemologi-
at one time, everyone could agree on who was the cal realism. This is not realism as a position within
village witch does not mean that there really were epistemology - the thesis that we have knowledge
witches, the fact that appropriately trained stu- of an objective, mind-independent reality - but
dents can pick out examples of analytic sentences something quite different: realism about the
does nothing to show that any sentences are gen- objects of epistemological inquiry.
uinely analytic. But the point I want to make does The epistemological realist thinks of knowledge
not require agreement on this particular example. in very much the way the scientific realist thinks of
Whether or not we agree with Quine on the ques- heat: beneath the surface diversity there is struc-
tion of analyticity, the fact remains that distinc- tural unity. Not everything we call knowledge need
tions can be teachable and projectible while be knowledge properly so called. But there is a way
failing to correspond to any theoretically coher- of bringing together the genuine cases into a coher-
ent division of objects. When a classification rests ent theoretical kind. By so doing - and only by so
on an implied background theory, there is no doing - we make such things as "knowledge of the
immediate inference from the existence of an easily external world" the objects of a distinctive form of
mastered kind-term to the theoretical integrity of theoretical investigation. We make it possible to
its associated kind. investigate knowledge, or knowledge of the world,
The application to our current problem is as such.
obvious. In accordance with my project of theo- I expect that at first it seemed bizarre to ques-
retical diagnosis, I have been arguing that the tion the existence of the objects of epistemologi-
kinds of knowledge investigated by the tradi- cal inquiry. Who can deny that we evaluate claims
tional epistemologist are theoretical kinds. So, and beliefs epistemologically, sometimes decid-
just as the ability of believers in the analytic/syn- ing that they express or amount to knowledge,
thetic distinction to agree on what to count as sometimes not? And who. can deny that these
paradigm instances of analytic sentences does claims or beliefs concern such things as objects in
n9t mean that there are analytic sentences, the our surroundings, other people's thoughts and
fact that we can agree on what to count as exam- experiences, events in the past, and so on? No
ples of knowledge of the external world does not one. So it is easy to assume that, if our claims ever
mean that there is knowledge of the external warrant positive assessment, there must be
world. The underlying principle of classification, knowledge of the external world, knowledge of
whatever it is, might be bogus. As a result, we other minds, knowledge of the past, and so on.
cannot simply help ourselves to classifications of Even more obviously, there must be knowledge.
this kind on the grounds that nothing else prom- But I hope the examples just considered make
ises the right kind of generality. That such princi- plausible the thought that there doesn't have to
ples of classification pick out coherent objects of be. All we know for sure is that we have various
theoretical investigation needs to be shown. practices of assessment, perhaps sharing certain
In the case of heat, to sort out the genuine from formal features. It doesn't follow from this that
the spurious examples we rely on a physical theory the various items given a positive rating add up
which identifies some underlying property, or struc- to anything like a natural kind. So it does not
ture of more elementary components, common to follow that they add up to a surveyable whole, to

a genuine totality rather than a more or less loose they are right, the analytic/synthetic distinction is
aggregate. Accordingly, it does not follow that a not essentially theoretical. But where a classifica-
failure to understand knowledge of the world tion is essentially theoretical, we are happy to say
with proper generality points automatically to an that there are no things of that kind, if we once
intellectual lack. To sum up, though I readily become convinced that the background theory is
admit that we have teachable distinctions here, all false. Thus there are no witches (or, if Quine is
this ensures is that there will be things that we can right, analytic sentences).
agree on as examples of, say, knowledge of the Though I do not claim that the concept of an
external world. It does not guarantee any theo- essentially theoretical classification is knife-edged,
retical integrity of the kind to which the examples I do want to say that "knowledge of the external
are assigned. This is the sense in which there need world" is quite clearly essentially theoretical.
be no such thing as knowledge of the world. There is no commonsense, pre-theoretical prac-
At this point, someone is likely to object that tice that this way of classifying beliefs rationalizes:
there is no immediate inference from the lack of a its sole function is to make possible a certain form
certain type of theoretical integrity in a given of theoretical inquiry, the assessment of knowl-
kind to its spuriousness. Still less is there an infer- edge of the world as such. As we have seen, this
ence to the non-existence of things of that kind. classification cuts across all familiar subject-
Take the sort of loose, functional classification of matter divisions and, in addition, presuppos~s
things that is common in everyday life, such as the autonomy of epistemology. Even the sense of
the division of dining room furniture into table "external" is unfamiliar from a commonsense
and chairs. We do not expect to be able to formu- standpoint. "External" does not mean "in one's
late a physical theory of what makes an object surroundings," for even one's own body, with its
a chair. But we are not tempted to conclude that "internal organs:' is an "external" object. It was a
chairs do not exist. 13 radical innovation on Descartes's part to exter-
This objection assumes that "knowledge of the nalize his own body. 14 As I have already remarked,
external world" is like "chair" rather than like "external" in "external world" means "without the
"witch" (or "analytic"). But is it? The distinctive mind." And since being within the mind depends
feature of terms like "witch" is that they are essen- on being given to consciousness, the essential
tially theoretical. Essentially theoretical distinc- contrast to "knowledge of the external world" is
tions are distinctions that we see no point in "experiential knowledge": the classification is
continuing to make, or even no way of drawing, epistemological through and through.
once the theory behind them has been rejected. But what if the proper analogy for "knowledge
If Quine is right, "analytic/synthetic" is like this, of the external world" were not "witch" but "heat?"
for he holds that giving up a certain conception I do not believe that it is because I do not see that
of meaning. involves losing all sense of how to there is any pre-theoretical utility to the concept, or
make a fixed, objective division between a theo- any theory-independent way of drawing even
ry's meaning postulates and its empirical assump- approximately the right boundaries round it. But
tions. Essentially theoretical classifications must this is not all. In bringing to centre-stage the issue
therefore be distinguished from classifications of epistemological realism, I am not questioning
that have been theoretically rationalized but particular theories of the structure of empirical
which retain independent utility. Distinctions knowledge, as we might question particular theo-
like this are apt to survive the rejection of theories ries of heat, but the very idea that knowledge has
with which they have become associated. Our any fixed, context-independent structure. The
first example, heat, is a case in point. Rejecting the analogy is therefore not with cases where one struc-
caloric theory of heat, or the phlogiston theory of tural theory replaces another but with those where
combustion, did not tempt us to conclude that we abandon any idea of coming up with a theory of
there are no hot things or that nothing burns. that kind. If there are no witches, we may debate
Some philosophers would take this view of "ana- witch -crazes and witchcraft beliefs, but not whether
lytic," for they think that there is a robust and sympathetic magic is superior to contagious.
useful pre-theoretical notion of synonymy that Suppose, however, that I am wrong about all
survives Quinean scepticism about meanings. If this. Suppose, that is, that "knowledge of the

external world" is like "chair": then what? So far deflationist will hold that his remarks on the
as I can see, nothing to the purpose. In connec- behaviour and utility of the truth-predicate say
tion with such loose, functional classifications, we just about everything there is to say about truth.
do not expect theoretical understanding, which is To approach truth in a deflationary spirit is
why such classifications survive the recognition emphatically not to think of "true" as denoting a
that no such understanding will be forthcoming. theoretically significant property, explicating which
We do not feel that there is an irremediable intel- will illuminate what is involved in any sentence's
lectuallack because there will never be a science being true. What is involved in a given sentence's
of chairs. But that is exactly what we are supposed being true is exhaustively captured by the sen-
to feel in the absence of a suitably anti-sceptical tence itself. On a deflationary view, then, true
theory of knowledge of the external world. This sentences constitute a merely nominal kind. We
shows that, even by the traditional epistemolo- could even say that, for a deflationist, though
gist's own standards, "knowledge of the external there are endlessly many truths, there is no such
world" cannot be like "chair." It must pick out thing as truth.
something in which theoretical integrity is to be The traditional theorist sees things quite dif-
expected, and this means that the existence of the ferently. In his eyes, "truth" is the name of an
objects of traditional epistemological inquiry is important property shared by all true sentences, a
far less assured than that of furniture. property that can be expected to repay theoretical
analysis. This property may be correspondence to
fact, incorporability in some ideally coherent
Explanation or Deflation? system of judgments, or goodness in the way of
belief, depending on whether he favours a corre-
Let me suggest one further case for comparison. spondence, coherence, or pragmatic theory. But
It has to do with deflationary views of truth. whatever his theoretical preference, he will hold
Philosophers who take a deflationary approach that, since true sentences constitute not just a
want no more from a theory of truth than a nominal but a theoretical kind, no theory of truth
description of the logical behaviour of "true" and is satisfactory which does not explain what makes
some account of why it is useful to have such a true sentences true. We set our sights too low if
device in our language. Quine is a good example we aim only to capture the use of a word or
of such a philosopher. According to Quine, if we explain the point of a concept:. there is more to
consider a sentence like" 'Snow is white' is true if understanding truth than appreciating the utility
and only if snow is white" we see that: "To ascribe of the truth-predicate.!?
truth to the sentence is to ascribe whiteness to We see, then, that traditional and deflationary
snow.... Ascription of truth just cancels the quo- theories are not theories of exactly the same kind.
tation marks. Truth is disquotation:'!5 As Stephen Leeds puts it, the traditional theories
Applied to a given sentence, the truth-predicate are genuinely theories of truth whereas deflation-
is dispensable. It comes into its own, however, ary theories are theories of the concept of truth
with respect to sentences that are not given, as (or, we could say, accounts of the use of "true") Y
when we say that all the consequences of a given Leeds's illuminating distinction is readily applied
theory are true. But even here, to say that certain to epistemological theories. We can distinguish
sentences are true is just to say that the world is theories of knowledge from theories of the con-
as they say it is. As Quine remarks, "one who cept of knowledge. I think that the debate sparked
puzzles over the adjective 'true' should puzzle by Gettier's demonstration that the standard
rather over the sentences to which he ascribes it. "justified true belief" analysis fails to state a suf-
'True' is transparent."!6 ficient condition for knowledge is best seen as
Tllough I am very sympathetic to this view, concerning the concept of knowledge. The kind
my interest here is less in its correctness than its of extra constraint on justification that seems to
character. This view of truth is striking on accouht be required - for example that an inference can-
of what it does not say. Compared with traditional not yield knowledge if it involves a false lemma
theories of truth, it says nothing about what essentially - is rather formal, nothing being said
makes all true sentences true. On the contrary, a about what beliefs can serve as justifying evidence

for what. This is why it is ptlssible to discuss issues This is in part what Austin is driving at in insist-
raised by the Gettier problem without ever get- ing that demands for justification are raised and
ting entangled in sceptical problems. Theories responded to against a background of specifically
that say nothing about whether examples of justi- relevant error possibilities. What is relevant will
fied beliefs about objective states of affairs reveal depend on both the content of the claim in ques-
any essential similarities, beyond highly formal tion and the context in which the claim is entered.
ones of the "no false lemmas" variety, are neutral If all evidence is relevant evidence, then, abstract-
with respect to whether we should think of our ing from such contextual details, there will be no
knowledge of the world as an appropriate object fact of the matter as to what sort of evidence
of theory. By contrast, traditional foundational could or should be brought to bear on a given
and coherence theories, which are much more proposition.
closely involved with scepticism, put forward If context-sensitivity goes all the way down,
general, substantive constraints on justification there is no reason to think that the mere fact that
and so make room for a project of assessing our a proposition is "about the external world" estab-
knowledge of the world as a whole. They are the- lishes that it needs, or is even susceptible of, any
ories of knowledge and not just theories of the particular kind of evidential support. No propo-
concept of knowledge. 19 sition, considered in abstraction, will have an
Of course, there is no obstacle in principle to epistemic status it can call its own. To suppose
supplementing one's views about the concept of that it must is precisely to fall in with what I call
knowledge with views about knowledge itself. 20 "epistemological realism." To treat "our knowl-
But one could also advance such views in a defla- edge of the world" as designating a genuine
tionary spirit. One philosopher who has done so, totality, thus as a possible object of wholesale
I believe, is Austin. Wittgenstein may be another. assessment, is to suppose that there are invariant
The availability of deflationary accounts of a epistemological constraints underlying the shift-
notion like truth changes the whole problem- ing standards of everyday justification, which it is
situation. Naively, we might be inclined to sup- the function of philosophical reflection to bring
pose that just as in physics we study the nature of to light. Exposing this epistemological deep struc-
heat, so in philosophy we study the nature of ture will be what allows us to determine, in some
truth. But once plausible deflationary views are general way, whether we are entitled to claim
on the table, the analogy between truth and things knowledge of the world. But if this is so, founda-
like heat can no longer be treated as unproblem- tionalist pre-suppositions are buried very deeply
atic, for the question raised by such views is pre- in the Cartesian project. They do not just fall out
cisely whether there is any need to think of truth of the totality condition's exclusion of any appeal
as having a "nature." We can conclude, mutatis to knowledge of the world in the course of our
mutandis, that if we have a plausible account of attempt to gain a reflective understanding of that
the concept of knowledge, it is a further step to knowledge. They turn out to be involved in the
insist on an account of knowledge as well. A defla- very idea of there being something to assess.
tionary account of "know" may show how the These are my suspicions in outline. Now we
word is embedded in a teachable and useful lin- must look at some details.
guistic practice, without supposing that "being
known to be true" denotes a property that groups
propositions into a theoretically significant kind. Foundationalism
We can have an account of the use and utility of
"know" without supposing that there is such a My main concern is the relation between scepti-
thing as human knowledge. cism and foundationalism. So having distin-
What makes this suggestion particularly guished between theories of knowledge and
pointed is that appearances certainly do not theories of the concept of knowledge, I must say
favour the view that a phrase like "knowledge of what kind of a theory I take foundationalism to be.
the world" picks out a theoretically coherent One way to understand foundationalism is to
kind. For one thing, justification, like explanation, see it as a doctrine about the formal character of
seems interest-relative, hence context-sensitive. justifying inferences. Formal foundationalism, as

we may call it, is the view that justification justification, while it might offer a way into the
depends on the availability of terminating beliefs fully general problem of the regress of justifica-
or judgments, beliefs or judgments which amount tion, gives no basis for supposing that there is a
to knowledge, or which are at least in some way particular sceptical problem about our knowl-
reasonably held to, without needing support from edge of objective reality. The transition to that
further empirical beliefs. Formal foundational- problem depends on the tacit assumption that
ism is sometimes thought to contrast with "coher- the fixed points recognized by commonsense jus-
entist" theories of knowledge or justification. tifications fall into some fairly obvious kind, so
According to theories of this type, a given belief that once they have been questioned there must
becomes justified through incorporation in some be some other, more primitive kind of judgment
suitably "coherent" system of beliefs or "total that we are forced to look to for their support.
view." Empirical inference is thus a matter of The thought that the functional role recognized
moving from one total view to another. The ter- by formal foundationalism corresponds to some
minating judgments, which the foundationalist kind of broad topical division of our beliefs is
sees as fixed points constraining the possibilities what I take to be the essential characteristic of
of inferential justification, are unnecessary. Some substantive, as opposed to merely formal, foun-
philosophers see the commitment to beliefs that dation ali sm.
function as fixed points as the essential feature of This is the way, then, in which there is more to
foundationalism, hence the complaint, promi- what I am calling (and what has generally been
nent in a recent systematic defence of the coher- called) "foundationalism" than the purely struc-
ence theory, that the key error in foundationalism tural doctrine of formal foundationalism. What
is its "linear" conception of inferenceY is missing from formal foundationalism is any
I have my doubts about the contrast between hint as to the kinds of beliefs that function as
foundationalism and the coherence theory, but fixed points or as to what qualifies a belief to play
they can wait. The point I want to make here is that role. But we have not yet got quite to the
that anyone who traces scepticism about our heart of why formal foundationalism is too weak
knowledge of the external world to the founda- a doctrine to capture all that is essential to a foun-
tionalist doctrine of epistemic priority must have dationalist conception of knowledge and justifi-
more than formal foundationalism in mind. We cation. The key point is this: that not only does
can call this stronger doctrine "substantive" foun- formal foundationalism give no account of what
dationalism. The distinction between formal and sorts of beliefs are epistemologically prior to
substantive foundationalism turns on the account what, and why, it does not even imply that any
given of terminating beliefs or judgments. Sub- such account needs to be given. If foundational-
stantive foundationalism involves more than the ism is a purely formal or structural doctrine, we
formal doctrine that inference depends on letting have no reason to think that a given belief has any
certain beliefs function as fixed points: it adds a particular or permanent epistemological status.
distinctive account of the kind of beliefs capable Perhaps the same belief can be a fixed point at
of performing that function. Since I think that a one time, or in one particular context of inquiry
genuinely foundationalist view of knowledge and or justification, but a candidate for justification at
justification must be substantive, whenever I refer another time or in another context. Nothing in
to foundationalism simpliciter I shall have sub- formal foundationalism excludes this.
stantive foundational ism in mind. 22 By contrast, substantive foundationalism pre-
Substantive foundationalism is a theory of supposes epistemological realism. I first intro-
knowledge, whereas formal foundationalism is duced the idea of epistemological realism by way
only (a contribution to) a theory of the concept of analogy with scientific realism. We can now
of knowledge. One way to see this is to recall that get a clearer sense of the appLOpriateness of the
Wittgenstein's view of knowledge, which con- analogy. A micro-structural theory of a physical
cedes that all justification takes place against a phenomenon is not purely structural. It will iden-
background of judgments affirmed without spe- tify both certain structures and the types of enti-
cial testing, can be seen as formally foundational- ties fitted to occupy appropriate places in them.
ist. But this point about our ordinary practices of (Think of models of the atom.) Similarly with the

foundationalist: he both attributes to justifying thought that any belief whatever about "external
inferences a certain structural character and iden- objects" must in the end derive its credibility from
tifies the types of beliefs fitted to play the various the evidence of "the senses;' knowledge of how
structurally defined roles: basic, inferential, etc. things appear.
Thus for the (substantive) foundationalist beliefs I call the foundationalist's supposed relations
have an intrinsic epistemological status that of epistemological priority "natural" to empha-
accounts for their ability to play one or other of size the fact that they are supposed to exist in
the formal roles the theory allows. Beliefs of one virtue of the nature of certain kinds of beliefs and
kind can be treated as epistemologically prior to not to depend on the changing and contingent
beliefs of some other kind because they are epis- contexts in which beliefs become embedded. For
temologically prior; some beliefs play the role of the foundationalist, in virtue of his epistemologi-
basic beliefs because they are basic; others receive cal realism, there is a level of analysis at which
inferential justification because they require it; epistemic status is not, as Quine once said of one
and all because of the kinds of beliefs they are. important epistemic feature, conventionality,
According to foundationalists, our beliefs arrange "a passing trait." Beliefs are more like the mem-
themselves into broad, theoretically coherent bers of a highly class-conscious society in which a
classes according to certain natural relations of person, no matter what he does, always carries the
epistemological priority. Beliefs to which no stigma or cachet of his origins. The quest for epis-
beliefs are epistemologically prior are epistemo- temic respectability is thus never entirely une
logically basic. Their credibility is naturally carriere ouverte aux talents. A given belief, though
intrinsic, as that of all other beliefs is naturally useful in all sorts of ways, generally and quite
inferential. This is a much more peculiar doctrine properly (in appropriate contexts) taken for
than is generally recognizedY granted, and beyond any specific reproach, can
On the foundationalist view, a belief's intrinsic never be allowed quite to forget that it presupposes
epistemological status derives from the content of the existence of the external world and is therefore,
the proposition believed. The foundationalist's by that fact alone, subject to some kind of residual
maxim is "Content determines status." Not, how- doubt, unless it can trace its lineage to more
ever, the details of content: what matter are cer- respectable data.
tain rather abstract features, for example that a The foundationalist conception of funda-
belief is about "external objects" or "experience." mental epistemological relations, cutting across
Thus it comes naturally to foundationalists to ordinary subject divisions and operating inde-
talk of basic propositions or basic statements, as pendently of all contextual constraints, receives
well as of basic beliefs. Propositions recording the an early articulation in Descartes's notion of "the
data of experience are held to be, by their very order of reasons." Descartes writes, "I do not
nature, epistemologically prior to propositions follow the order of topics but the order of argu-
about external objects, which is why they are apt ments. '" [In] orderly reasoning from easier mat-
for the expression of basic beliefs. In light of this, ters to more difficult matters I make what
we can characterize foundationalism as the view deductions I can, first on one topic, then on
that our beliefs, simply in virtue of certain ele- another:'24 However, it is far from obvious that
ments in their contents, stand in natural episte- there is such an order of reasons, operating inde-
mological relations and thus fall into natural pendently of the division of topics. It is not at all
epistemological kinds. The broad, fundamental clear that some matters are intrinsically - that is
epistemological classes into which all proposi- to say independently of all circumstances and all
tions, hence derivatively all beliefs, naturally fall collateral knowledge - "easier" than others. The
constitute an epistemic hierarchy which deter- way that justification and inquiry proceed in
mines what, in the last analysis, can be called on common life, or for the matter theoretical science,
to justify what. This means that, for a foundation- is far from evidently favourable to the foundation-
alist, every belief has an inalienable epistemic alist conception of epistemological relations. In
character which it carries with it wherever it goes both science and ordinary life, constraints on jus-
and which determines where its justification must tification are many and various. Not merely that,
finally be sought. The obvious illustration is the they shift with context in ways that are probably

impossible to reduce to rule. In part, they will were to have any doubt of it, then I don't know
have to do with the specific content of whatever why I should trust my eyes. For why shouldn't I
claim is at issue. But they will also be decisively test my eyes by looking to find out whether I see
influenced by the subject of inquiry to which the my two hands? What is to be tested by what?
claim in question belongs (history, physics, orni- (Who decides what stands fast?)26
thology, etc.). We can call these topical or, where
some definite subject or distinctive form of The point is that, in the absence of a detailed
inquiry is involved, disciplinary constraints. Not specification of a particular context of inquiry,
entertaining radical doubts about the age of the the sort of specification that would fix the rele-
Earth or the reliability of documentary evidence vant contextual constraints on justification, the
is a precondition of doing history at all. There are question "What is to be tested by what" has no
many things that, as historians, we might be answer. Questions about justification are essen-
dubious about, but not these. tially context-bound. This is something a founda-
Disciplinary constraints fix ranges of admis- tionalist will deny. He must of course make
sible questions. But what is and is not appropriate allowances for the way that what tests what can
in the way of justification may also be strongly shift with context. But - and this is the crucial
influenced by what specific objection has been point, he cannot allow that such contextual deter-
entered to a given claim or belief. So to discipli- mination goes all the way down. At the funda-
nary we must add dialectical constraints: con- mental level, what is to be tested by what is
straints reflecting the current state of a particular objectively fixed, which is why there is no ques-
argument or problem-situation. In this respect tion of anybody's deciding the matter. The answer
justification is closely akin to explanation, which is determined by the epistemological facts them-
is also context-sensitive because question-relative. selves: by fundamental, objective relations of
I shall have more to say about disciplinary con- epistemological priority. This is not exactly an
straints and about the relation between justifica- "intuitive" view.
tion and explanation. But for now let me note Continuing with the example of my knowing
that, in ordinary examples of requiring and pro- (in normal circumstances) that I have two hands,
ducing justifications, the epistemological status recall also that there is no obvious way to general-
of a given claim can also depend on the particular ize from an example like this. In normal circum-
situation in which the claim is entered, so that jus- stances, the proposition that I have two hands is
tification is also subject to a variety of situational as certain as anything we could cite as evidence
constraints. Here I have in mind the wordly and for it. But there is no obvious, non-trivial way of
not just the dialectical situation. Consider yet again saying what other propositions are, in normal cir-
Wittgenstein's remark that "My having two hands cumstances, as certain as anything we could cite
is, in normal circumstances, as certain as anything as evidence for them. Normally, I am as certain as
I could produce in evidence for it:'25 Entered in I could be of anything that my name is Michael
the right setting, a claim to have two hands might Williams: but beyond this, what does the proposi-
function like a foundationalist's basic statement, tion that my name is Michael Williams have in
providing a stopping place for requests for evidence common with the proposition that I have two
or justification: hence the element of formal foun- hands? What feature of their content explains
dationalism in Wittgenstein's view. But in other their belonging to the same epistemic kind? As far
circumstances the very same claim might be con- as I can see, there isn't one. So even if someone
testable and so might stand in need of evidential said that the claim to have two hands did have a
support. The content of what is claimed does not kind of intrinsic status - that of being certain in
guarantee a claim some particular epistemic stand- normal circumstances - we would still not be able
ing. Not merely is status often dependent on the to treat the example as paradigmatic of proposi-
details of content, it is never determined by content tions belonging to a definite epistemic kind, for
alone. As Wittgenstein notes: which we could articulate some alternative, non-
trivial criterion of membership.27 Again, the
If a blind man were to ask me "Have you got two foundationalist sees things quite differently. For
hands?" I should not make sure by looking. If I him, highly abstract divisions of propositions

according to content (propositions about external theoretical investigation, we have to suppose that
objects versus experiential propositions, proposi- there are pervasive epistemological constraints or
tions about the past versus propositions about the relations. That is to say, at least some constraints
present, etc.) have to coincide with fixed differ- on what propositions demand evidential support
ences in epistemological status. But what we and on what propositions can provide it must be
should learn from the example under discussion context-invariant. If we do not always insist on
is that no such coincidence can be simply assumed. respecting these constraints in a fully rigorous way,
To cite again another ofWittgenstein's reminders, this need not mean that they do not apply. To admit
"a proposition saying that here is a physical object that certain constraints are often waived is different
may have the same logical status as one saying that from, indeed incompatible with, claiming that they
here is a red patch."28 Without natural epistemo- are inapplicable.
logical kinds, the foundationalist's permanent This is a very substantial commitment and it
underlying structure of epistemological relations is not clear why we should accept it. An exami-
goes by the board. nation of ordinary practices of justification
We see from this that the antidote to founda- strongly suggests that constraints, governing
tionalism, indeed to epistemological realism gen- what sorts of evidence can properly be brought
erally, is a contextualist view of justification. 29 To to bear on a disputed claim, what needs to be
adopt contextualism, however, is not just to hold defended, and what can safely be taken for
that the epistemic status of a given proposition is granted, though subject to other kinds of con-
liable to shift with situational, disciplinary, and textual determination as well, are at least topic-
other contextually variable factors: it is to hold relative, which is to say determined in part by the
that, independently of all such influences, a propo- subject under discussion.
sition has no epistemic status whatsoever. There is We might criticize Hume's offhand suggestion
no fact of the matter as to what kind of justification that only carelessness and inattention save us
it eitl;1er admits of or requires. Thus stated, contex- from a permanent, debilitating awareness of the
tualism implies a kind of externalism, for though truth of scepticism, hence from lapsing into a
appropriate contextual constraints will have to be state of chronic, paralysing doubt. In particular
met, if a given claim is to express knowledge, they contexts, disciplines etc., exempting certain prop-
will not always have to be known, or even believed, ositions from doubt is what determines the direc-
to be met. 30 But when we realize that the point of tion of inquiry. As Wittgenstein remarks: "It may
contextualism is to oppose the sceptic's or tradi- be ... that all enquiry on our part is set so as to
tional epistemologist's epistemological realism, exempt certain propositions from doubt, if they
the externalist element in contextualism ought to are ever formulated. They lie apart from the route
be more palatable. The problem with externalism travelled by enquiry."31
was that it seemed to deprive us of the possibility If some of these propositions cease to lie apart
of answering a perfectly intelligible question: how from the route travelled by inquiry, then inquiry
do we come to know anything whatsoever about travels by a different route. Or perhaps no clear
the external world? What we now see is that this route remains for it to travel by. This is obviously
question is not at all intuitive but reflects theoreti- the case with investigations in particular scientific
cal presuppositions that are not easy to defend. or scholarly disciplines. Disciplinary constraints
Contextualism, with its implied externalism, is not have a great deal to do with the kinds of questions
offered as a question-begging direct answer to an that can and cannot legitimately be raised with-
undeniably compelling request for understanding, out radically affecting the direction of inquiry.
but as a challenge to justify the presumption that Thus, introducing sceptical doubts about whether
there is something to understand. the Earth really existed a hundred years (or five
minutes) ago does not lead to a more careful way
of doing history: it changes the subject, from his-
Methodological Necessity tory to epistemology. So when Wittgenstein asks:
"am I to say that the experiment which perhaps I
We have already seen that, to flesh out the idea make to test the truth of a proposition presupposes
of "human knowledge" as a possible object of the truth of the proposition that the apparatus

I believe I see is really there?"32 he is clearly invit- and propositions within a method."35 For a sub-
ing the answer "No." And the reason for answering ject like history, there is more to method than
"No" is that the possibility mentioned, while rele- abstract procedural rules. This is because the
vant to certain general, epistemological problems, exclusion of certain questions (about the exist-
is completely beside the point in the context of a ence of the Earth, the complete and total unreli-
specific experiment in chemistry or physics. To ability of documentary evidence, etc.) amounts
bring it up is not to introduce greater rigour into to the acceptance of substantial factual commit-
the investigation in hand but to shift attention ments. These commitments, which must be
to another kind of investigation entirely. accepted, if what we understand by historical
"[Tlhat something stands fast for me;' inquiry is to be conducted at all, have the status,
Wittgenstein remarks, "is not grounded in my relative to that form of inquiry, of methodological
stupidity or credulity."33 We now see that this is necessities.
so, at least in part, because it is grounded in my I have introduced the idea of a proposition's
interests. It is not that I think that no proposition being exempted from doubt as a matter of meth-
that stands fast could ever be questioned, though odological necessity in connection with the disci-
in certain cases I should be likely to feel, as plinary constraints that determine the general
Wittgenstein says, "intellectually very distant" directions of highly organized forms of inquiry.
from someone inclined to raise questions. It is But it is evident that something similar goes on in
just that some doubts are logically excluded by more informal, everyday settings. Asking some
forms of investigation that I find significant, questions logically precludes asking others: all
important, or perhaps just interesting. This has sorts of everyday certainties have to stand fast if
nothing to do with dogmatism, credulity or care- we are to get on with life. Again, however, I want
lessness. Wittgenstein sums up the key points in to emphasize that our situation is misread both
the following well-known passages: by the Human naturalist and by the sceptic. The
naturalist sees our everyday inability to entertain
The questions that we raise and our doubts radical doubts as showing that nature has simply
depend on the fact that some propositions are determined us to believe certain things, however
exempt from doubt, are as it were like hinges on groundless they seem to us in our more reflective
which those turn. moments. By contrast, I want to claim that
That is to say, it belongs to the logic of our exemption from doubt - epistemic privilege - is a
scientific investigations that certain things are matter of methodology, not psychology. In a spe-
indeed not doubted. cific context, certain exemptions will be logically
But it isn't that the situation is like this: We just required by the direction of inquiry. We are there-
can't investigate everything, and for that reason fore determined by Nature to hold certain things
we are forced to rest content with assumption. If I fast only in so far as we are naturally inclined to
want the door to turn, the hinges must stay put. 34 interest ourselves in matters requiring us to
exempt them from doubt.
Of course, if I do not want the door to turn I can This is far from the only point that we must
nail it shut; or I might want it to open the other emphasize. It is also crucial to note that, if epis-
way, in which case I will move the hinges. But if temic status is determined by the direction of
I want the door to turn this way, it is not just inquiry, the reason why, in a given inquiry, cer-
more convenient, if a little slapdash, to place the tain propositions have to stand fast has to be
hinges where they are: there is nowhere else to separated from the reason why that inquiry
put them. results in knowledge, if it does. Here we recur,
By fixing a range of admissible questions, we from a slightly different angle, to the externalist
determine a form of inquiry. But this means that element in contextualism. In particular contexts
a form of inquiry is determined by more than of inquiry, certain propositions stand fast as a
purely formal constraints. As Wittgenstein puts it: matter of methodological necessity. But inquiries
'''The question doesn't arise at all.' Its answer informed by them will yield knowledge only if
would characterise a method. But there is no sharp those propositions are true, which they need not
boundary between methodological propositions always be.

The general moral here is that questions about as somehow derivative from experience. No other
a proposition's epistemic status must always be way of seeing it permits an assessment, hence a
separated from questions about its truth. If epis- legitimating explanation, at the proper level of
temic status is fixed by the direction of inquiry, generality.
epistemic status is context-sensitive. Truth how- We are now in a position to see why this argu-
ever is not. A proposition is either true or not. ment does not prove what it needs to prove. All it
But, according to the contextualist view I favour, shows is that the doctrine of the priority of expe-
we cannot say, in a similarly unqualified way, that riential knowledge over knowledge of the world is
a proposition is either open to doubt or not. a methodological necessity of the traditional episte-
Sometimes it will be and sometimes it won't. mological project. But since the sceptic himself is
Generally speaking, a proposition is neither true irrevocably committed to distinguishing between
because it stands fast nor stands fast because it methodological necessity and truth, it does not
is true. show, nor by his own standards can the sceptic
We can also see why it was so important at the take it to show, that that doctrine is true.
outset to distinguish between formal and sub- The result is that the inference from the essen-
stantive foundationalism. If foundationalism is tial generality of the traditional epistemological
equated with a certain view of the formal struc- project fails to establish the kind of relations of
ture of justification - i.e. with the view that infer- epistemological priority needed to threaten us
ential justification always requires beliefs that with scepticism. To yield sceptical results, these
function as "fixed points - a contextualist view of relations must reflect more than mere methodo-
justification can be seen as (formally) founda- logical necessities: they must correspond to fully
tionalist. But it certainly need not be substantively objective epistemological asymmetries. It is not
foundationalist. There are no limits as to what enough to point out that if we are to attempt an
might or might not, in an appropriate context, assessment of our knowledge of the world as a
be fixed. whole we must take experiential knowledge to be
In an earlier chapter, I tried to show that argu- epistemologically prior to the knowledge we want
ments for radical scepticism presuppose the pri- to assess. Success or failure in the enterprise will
ority of experiential knowledge over knowledge have the significance the sceptic and the tradi-
of the world. This enabled me to conclude that tional epistemologist mean it to have only if expe-
attempts to establish the intrinsic epistemological riential knowledge really is, as a matter of objective
priority of experiential knowledge on the basis epistemological fact, more basic than knowledge
of the greater intrinsic dubitability of objective of the world. If it isn't, or more generally if no
knowledge are question-begging. The only reason epistemological relations are in the sense I have
for thinking that such knowledge is intrinsically indicated fully objective, no attempt to ground
more dubitable is provided by the existence of knowledge of some allegedly problematic kind on
sceptical arguments which, when unpacked, turn some appropriately prior kind of knowledge will
out to take the doctrine of the priority of experi- amount to an attempt at assessment. Should the
ential knowledge for granted. attempt fail, or even inevitably fail, the sceptic
This result did not allow us to conclude will be left with a harmless logico-conceptual
straight away that scepticism rests on a gratuitous point but with no way of advancing to his pessi-
epistemological assumption. What it did suggest, mistic epistemological conclusion.
however, is that the source of the doctrine of the I remarked that the argument from the total-
priority of experiential knowledge is not evidence ity condition to the absolute priority of experi-
from our ordinary justificational practices but ential knowledge over knowledge of the world
rather the distinctively philosophical project of rests on two assumptions: that there is some-
trying to understand how it is possible for us to thing to assess, and that charting its relation to
know anything whatsoever about the external experience amounts to assessing it. I have con-
world. The totality condition that the sceptic centrated on the first, but by so doing have shown
(or the traditional philosopher) imposes on a what to say about the second. As a pure method-
philosophical understanding of our knowledge of ological proposal, there is nothing wrong with
the world is what forces us to see that knowledge setting propositions about the world against

experiential propositIOns, for the purposes of defended in discussions of philosophical scepti-

exploring possible relations between them. Like cism. But in plain contexts, nobody doubts that
Goodman, we could think of phenomenalism as they are true, even though plain common sense
an interesting constructive project. We could ask, recognizes the very phenomena - dreaming, hal-
"To what extent can a phenomenalist reconstruc- lucinating, and so on - that the sceptic appeals to
tion of the world be carried through?" without in his attempt to show that we can never know
thinking that we were even addressing any ques- that we are in touch with "a real public world."
tions of epistemic legitimacy.36 Think of the way Whether there is a clash between philosophy and
we can model arithmetic in set theory: though common sense will depend, therefore, on the rela-
this is an interesting piece of mathematics, we tion between philosophical and plain knowing.
need ancillary epistemological assumptions to Here Clarke is more subtle than Carnap, for
think of it as relevant to an "assessment" of arith- he recognizes that the sceptic has an account of
metic. But this is not the spirit in which the scep- the relation between philosophy and common
tic thinks of the relation between experiential sense which both preserves the relevance of phil-
knowledge and knowledge of the world. He needs osophical discoveries to ordinary plain knowing
a fully objective epistemological asymmetry, and and makes it hard to think that sceptiCal claims
this is what no argument from methodological are less than fully meaningfuJ.38 Ordinary, plain
necessity will ever yield. knowing is hemmed by practical considerations.
Some philosophers, Carnap for example, hold By contrasts, to philosophize is "to step outside
that the sceptic fails to undermine ordinary knowl- the nonsemantical practice" and, meaning simply
edge of the world because his statements, as he what one's words mean, ask whether we really
intends them to be taken, mean nothing at all. As know what we (plainly) take ourselves to know.
a statement "internal" to our everyday linguistic Compared with our philosophizing, ordinary
framework, "There are material objects" is a trivial thinking is "restricted." All the sceptic has to do is
consequence of any statement about the world. to get us to look beyond the restrictions. This is
But as an "external" statement about that frame- easy enough since there is a standing invitation to
work, an attempted statement, though made in look "beyond the plain" in our conception of
the very same words, will lack "cognitive signifi- knowledge as knowledge of an objective world.
cance." However, the sceptic might be equally We want to know what there is: not just relative to
unsuccessful if his statements, as they must be this and that particular restriction, imposed by
understood in the unusual context of philosophi- this or that practical purpose or limitation, but
cal reflection, mean something different from absolutely.
what they ordinarily mean. Thus Thompson Still, the final distance between Clarke and
Clarke suggests that the very general common- Carnap is not as great as their initial divergence
sense propositions with which Moore confronts might suggest. Clarke too holds that, in the end,
the sceptic can be taken two ways, the "plain" way both "philosophical common sense" and its
and the "philosophical" way. For example: sceptical denial "are a spurious fiction if our
conceptual-human constitution is not standard."
Suppose a physiologist lecturing on mental Amongst other things, a conceptual-human con-
abnormalities observes: Each of us who is normal stitution of the standard type requires that "Each
knows that he is now awake, not dreaming or hallu- concept or the conceptual scheme must be
cinating, that there is a real public world outside his divorceable intact from our practices, from what-
mind which he is now perceiving, that in this world ever constitutes the essential character of the plain"
there are three-dimensional animate and inanimate and that we, as concept users, are "purely ascer-
bodies of many shapes and sizes. ... In contrast, taining observers who, usually by means of our
individuals suffering from certain mental abnor- senses, ascertain, when possible, whether items
malities each believes that what we know to be the fulfill the conditions legislated by concepts."39 But
real public world is his imaginative creation. 37 the sceptic himself shows that our conceptual-
human constitution cannot be of the standard
The italicized, plain propositions are "verbal type. Our plain knowledge that we are not dream-
twins" of propositions typically attacked and ing right now - the sort of knowledge expressed

by the physiologist - cannot be undermined by driven by that same doctrine. If a statement is

the plain possibility that we might, in fact, be certain in one context but not in another, the
asleep. But it would be if our conceptual-human argument assumes, this can only be because a
constitution were of the standard type. For on change in context induces a change in meaning.
this point the sceptic is right: there are no marks So if, plainly speaking, we do know that we are
or features that conclusively distinguish waking awake at the moment, whereas, philosophically
experience from dreaming. So the fact of plain speaking, we don't, our plain and philosophical
knowing, combined with the sceptic's point about propositions can only be "verbal twins:' But if, as
dreaming or hallucinating, shows that our con- I have argued, epistemological status is never
ceptual-human constitution is not of the stand- determined by content alone, there is no such
ard type. This insight is part of the legacy of easy inference from a difference in status to a dif-
scepticism. ference in content. We can explain the context-
In representing the sceptic as helping bring boundedness of sceptical doubts without getting
about his own undoing, Clarke prefigures the entangled in this baroque apparatus of plain and
strategy followed by Wright. Wright, we may philosophical meanings. As we shall see in a
recall, argues that the sceptic does indeed show moment, this is all to the good.
that his target-propositions - for example, that Once again, I must emphasize that my argu-
there is a real, public world - are beyond justifica- ment on these matters will not be complete until
tion. They are beyond justification because the I have examined the sceptic's own favoured
sole evidence we can bring to bear on them only account of the nature of philosophical reflection.
functions as evidence if they are already known to Even so, however, I think it is fair to conclude that
be true. Thus sensory experience only counts in we are well on the way to accomplishing the pri-
favour of any proposition about the public world mary goal of theoretical diagnosis, which is to get
on assumptions that already commit us to that the sceptic to share the burden of theory. But
world's existence. But the lesson to learn from this there is a nagging question that is likely to surface
is that the propositions the sceptic represents as again at this point. If we are left with one theory
groundless, factual assumptions, are not really of knowledge confronting another, and we will
factual at all. If a proposition's factuality requires never be able to determine conclusively which is
some account of the cognitive powers that would correct, doesn't the sceptic win ties and so tri-
be required for knowing that proposition to be umph at second order?
true, and if the sceptic shows that, in the case of If we abandon epistemological realism, there
some propositions, no such account can be given, is a clear sense in which we no longer see such
scepticism is self-undermining. This argument things as "knowledge of the world" as appropriate
shares with Clarke's more than just structural objects of theory. At most, we will have a theory
similarities. of the concept of knowledge. We will not have a
None of these arguments appeals to me. I do theory of knowledge as well. A fortiori, we will not
not want to distinguish between internal and be left confronting the sceptic's theory with a
external questions or between plain and philo- theory of our own.
sophical meanings of statements. Nor do I wish to Perhaps this will look like a purely verbal
claim that, for deep philosophical reasons, appar- manoeuvre, for we shall certainly be left with
ently factual statements are really not factual at epistemological views, whether or not we want to
all. The reason is that I think that all these reac- think of them as a theory of knowledge. But the
tions to scepticism reveal the deep and pervasive point isn't just verbal. For what we have seen is
influence of epistemological realism. I suggested that the sceptic's theoretical commitments are in
earlier that one of the epistemological realist's fact far more extensive than those of his contextu-
central commitments is to the doctrine that con- alist opponent. Contextualism simply takes seri-
tent determines status. Now I claim that the ously and at face-value what seem to be evident
attempt to insulate common sense from sceptical facts of ordinary epistemic practices: that relevant
undermining by finding a different meaning, or evidence varies with context, that content alone
no factual meaning at all, in the apparently com- never determines epistemological status, and so
monsensical propositions the sceptic examines is on. The theoretical resources required to explain

these appearances away belong entirely to the have seen repeatedly, attempts to argue for it
sceptic. So it might be reasonable to object that directly beg the question. So the doctrine has to
the sceptic wins ties, if the outcome of my theo- be true but unarguable.
retical diagnosis were a tie. And if I had followed I think that the sceptic's difficulties are com-
philosophers like Carnap, Clarke, or Wright and pounded when we turn from this relatively par-
rested my diagnosis on difficult and controversial ticular doctrine to epistemological realism in
views about meaning, perhaps it would have been. general. It is not easy to imagine what a convinc-
But as things stand it isn't. ing argument for epistemological realism would
This is not all. It seems to me entirely reason- even look like, or what evidence it could appeal to.
able to hold that extra theoretical commitments This is where the clash between scepticism and
demand extra arguments. But where will the our ordinary attitudes really does work to the
sceptic find them? Not in evidence from everyday sceptic's disadvantage. It does so because our
practice, which fits in as well or better with con- ordinary practices of justification not only toler-
textualism. Presumably, then, in some kind of ate but invite a contexualist construction: and
general, theoretical considerations. Here, how- contextualism is the antidote to epistemological
ever, we run into the fallaciousness of the argu- realism.
ment from methodological necessity: by the True, a contextual view of knowledge and jus-
sceptic's own standards, there is no inference tification will seem unsatisfactory to a philosopher
from the fact that we must take experiential who continues to feel the lack of an understanding
knowledge to be generally prior to knowledge of of human knowledge in general. But if my argu-
the world, if we are to make room for a project of ment to this point is correct, he will feel this lack
assessing our knowledge of the world as a whole, only if he is already predisposed to epistemological
to its really being so. But if the argument from realism. Once more, we are starting to run round a
methodological necessity does not show that the very small circle of ideas. The sceptic's foundation-
sceptic's principles are true, what would? It is alism, together with the epistemological realism it
hard to say: for although the argument from embodies, is a brute metaphysical commitment.
methodological necessity is fallacious, it is not as The theoretical diagnostician could hardly ask
if there are other ways of arguing for the priority for more.
of experiential knowledge. On the contrary, as we


Barry Stroud, "Understanding Human 7 Rene Descartes, "Meditations on First

Knowledge in General," in Marjorie Clay and Philosophy: First Meditation;' in The
Keith Lehrer (eds), Knowledge and Skepticism Philosophical Works of Descartes, vol. 1, trans.
(Boulder, CO: Westview, 1989), p. 32. Elizabeth Haldane and G. R. T. Ross
2 W. V. Quine, "Epistemology Naturalized;' this (Cambridge: Cambridge University Press,
vol., ch. 39. On Quine's problematic attitude 1972), p. 145; David Hume, A Treatise of
towards traditional epistemology, see Barry Human Nature, ed. L. A. Selby-Bigge, 2nd
Stroud, The Significance of Philosophical edn rev. P. H. Nidditch (Oxford: Oxford
Scepticism (Oxford: Oxford University Press, University Press, 1978), p. xx (hereafter,
1984), ch. VI. THN).
3 Barry Stroud, "Skepticism and the Possibility 8 I am grateful to Simon Blackburn for this
of Knowledge;' Journal of Philosophy 81, pp. useful phrase.
545-51. p. 551. 9 These quotations and the next, THN,
4 "Understanding Human Knowledge;' p. 49. pp. xv-xvi.
5 Thomas Nagel, The View from Nowhere 10 Novum Organum, bk II in J. Spedding,
(Oxford: Oxford University Press, 1986), R. Ellis, and D. Heath (eds), The Works of
p.68. Francis Bacon (London: Longman, 1857-8),
6 "Understanding Human Knowledge;' p. 32. vol. IV. My example is slightly unfair to

Bacon in that he did not intend to give 22 The terminology of formal versus substan-
instances of everything that would, III tive foundationalism is also employed by
common parlance, be said to involve "heat- Ernest Sosa: see "The Raft and the Pyramid,"
ing." His aim was to collect instances which this vol., ch. l3. However, I am uncertain
"agree in the same nature, though in sub- whether my usage is the same as Sosa's.
stance the most unlike": Novum Organum, According to Sosa, "A type of formal founda-
II, aphorism xi. tionalism with respect to a normative or
11 W. V. Quine, "Two Dogmas of Empiricism," evaluative property I/J is the view that the
in W. V. Quine (ed.), From a Logical Point of conditions (actual and possible) within
View (New York: Harper, 1963). which I/J would apply can be specified in gen-
12 H. P. Grice and P. F. Strawson, "In Defence of eral, perhaps recursively. Substantive founda-
a Dogma," Philosophical Review (1956). tionalism is only a particular way of doing
13 I must thank Alvin Goldman for pressing me so" (p. 278, italics in original). From my point
on this point. of view, everything depends on what is
14 Myles Burnyeat ("Idealism and Greek allowed to count as a "general" specification.
Philosophy: What Descartes Saw and Berkeley 23 Though I have grave doubts about the notion
Missed;' Philosophical Review 90, pp. 3-40) of intrinsic credibility, having written about
thinks that Descartes's externalization of his all this elsewhere (Groundless Belief (Oxford:
own body is the key move in his invention of Blackwell), chs 2, 3, and 5) I will not repeat
the problem of our knowledge of the exter- myself. My interest here is in the foundation-
nal world. I see it as a consequence of his alist's conception of epistemological prior-
epistemological realism. ity, which I see as his deepest theoretical
15 W. V. Quine, Pursuit of Truth (Cambridge commitment.
MA: Harvard University Press, 1990), p. 80. 24 Letter to Mersenne (24 December, 1640),
16 Ibid., p. 82. quoted from Anthony Kenny (ed.), Descartes:
17 However, it seems to me that the more a PhilosophicalLetters (Minneapolis: University
purely disquotational account of "true" can of Minnesota Press, 1981), p. 87.
be shown to capture whatever we want out 25 Ludwig Wittgenstein, On Certainty (Oxford:
of the truth-predicate, the less reason there is Blackwell, 1969), p. 250; hereafter OC.
for thinking that there must be some "truth- 26 Ibid., p. 125.
making" property that all true sentences 27 Failure to appreciate this point is what viti-
share: the invitation to apply Occam's Razor ates Marie McGinn's intuitive reconstruc-
ought to be, irresistible. tion of the case for scepticism.
18 Stephen Leeds, "Theories of Reference and 28 ~C, p. 53.
Truth;' Erkenntnis (1978). 29 For a succinct defence of contextualism, see
19 Attempts to discuss the Gettier problem and David B. Annis, ''A Contextualist Theory of
traditional sceptical questions in the same Epistemic Justification;' American Philo-
breath often seem rather contrived. The sophical Quarterly (1978), reprinted III
epistemological analogue of Leed's distinc- Moser, Empirical Knowledge (Totowa, NJ:
tion explains why. Rowman and Littlefield, 1986). Annis sees
20 Keith Lehrer notes that the "definitional or Pierce, Dewey, and Popper as having been,
formal" in his theory of knowledge, which historically, the key contextualists. He may
"constitutes an analysis or explication of the be right, though I have doubts about how
concept of knowledge;' "leaves open sub- far these philosophers saw into the implica-
stantive issues:' See Lehrer, "Knowledge tions of contextualism. For example, I doubt
Reconsidered" in Clay and Lehrer (eds), whether Popper would be as suspicious as
Knowledge and Skepticism, quotation p. l32. he is about justification if he were really a
21 Laurence BonJour, The Structure of Empirical thoroughgoing contextualist.
Knowledge (Cambridge, MA: Harvard Univer- 30 Obviously, this kind of externalism is not the
sity Press, 1985), pp. 89ff. Subsequent cita- same as pure reliabilism, except in so far as
tions given by Structure and page numbers. the apparent theoretical simplicity of some

forms of reliabilism is a sham. Consider, 33 Ibid., p. 235.

for example, Colin McGinn's suggestion 34 Ibid., pp. 341-3.
(in "The Concept of Knowledge;' in Midwest 35 Ibid., p. 318, emphasis in original. Cf. the
Studies in Philosophy IX (1984)) that know- metaphor of the river bed at pp. 95-8.
ing that p depends on the availability of a 36 Nelson Goodman, Ways of Worldmaking
"way of telling" that p. This analysis does not (Indianapolis: Hackett, 1978), ch. 1.
guarantee any theoretical integrity in "ways 37 Thompson Clarke, "The Legacy of
of telling" and is therefore compatible with Skepticism;' Journal of Philosophy 69, pp.
a contextual, hence anti-epistemological- 754-69; p. 756.
realist, conception of knowledge. 38 Ibid., pp. 758 ff.
31 OC,p.88. 39 Ibid., pp. 762, 761.
32 Ibid., p. 163.
The Structure of Knowledge
and Justification

The Pyrrhonian problematic can be formulated as follows. One can be justified in

believing that p only if one has a reason to believe that p. But if a proposition that q is
one's reason to believe that p, it can provide justification only if it is a good reason - that
is, only if it, too, is something one is justified in believing. This leaves three possibilities
for any tree of justification: (1) all its branches terminate; (2) at least one of its branches
contains a loop; (3) at least one of its branches is infinite. Thus, we have the three tra-
ditional theories of justification: foundationalism, coherentism, and the rather less
popular infinitism. To be complete, there is a fourth option not mentioned, namely
that skepticism is true, and there are no trees of justification, for no one is ever justified
in believing anything.
This description of the Pyrrhonian problematic corresponds closely to the way
Roderick Chisholm sees the epistemological terrain. Faced with these options, he
chooses foundationalism. This means admitting that there are some propositions that
we are justified in believing but for which we lack reasons in the form of further prop-
ositions we are justified in believing. Chisholm fully embraces this consequence. Having
made the statement "There lies a key:' if one is asked "What is your justification for
thinking that?," one must provide an answer, but eventually, in the chain of questions,
a claim about one's present experience will be challenged: "What is your justification
for thinking you have such-and-such experience?" To this, Chisholm thinks one can do
no better than to answer: "My justification is that I have such-and-such experience."
Similarly, faced with a challenge to a claim regarding one's present belief that p, one
must repeat oneself, saying "My justification for thinking I believe that p is that I believe
that p."
Wilfrid Sellars attacks the doctrine of the given precisely on the issue of the episte-
mological status of these foundational beliefs. If there is knowledge that is unsupported
by further knowledge, as Chisholm would have to acknowledge, then reports of this
knowledge, like reports of any piece of knowledge, must have authority (Chisholm and
Sellars seem to assume that the structure of justification is the same as the structure of

knowledge). But a report can have authority only if the person making it recognizes its
authority. Thus, even in the case of my knowledge that what I see before me is green,
my report "This is green" must have authority that I recognize. Moreover, in this case,
the authority can only lie in the reliable connection between the production of tokens
of "This is green" and the presence of green objects. So if I am to know through obser-
vation that what I see is green, I must recognize the truth of this generalization. How,
then, do I know the truth of the generalization? My present knowledge is based on
memory knowledge of instances of it. What of my knowledge of these instances? Are
we headed for a regress? No, answers Sellars, for although I have such memory knowl-
edge, the experiential beliefs from which these memories are derived need not have
been instances of knowledge. (Presumably, these instances of memory knowledge, too,
have authority that is also recognized by the subject in the form of a belief that reports
of "This was green, and I experienced it to be so" co-vary with actual past encounters
with green objects.)
Sellars' account rules out Chisholm's given. Formulated in the language of beliefs
rather than of reports: if all knowledge of particular matters of fact, including observa-
tional knowledge, depends on general knowledge, and if in turn this general knowledge
itself depends on knowledge of particular matters of fact, then empirical knowledge
has no foundation.
Laurence BonJour joins Sellars in arguing that foundationalism fails to solve the
problem that it was designed to solve, viz. the Pyrrhonian problematic. He bases his
argument on the very nature of epistemic justification as essentially connected to the
cognitive goal of truth. What this connection amounts to, he claims, is that a belief is
justified only if one has good reason to think it is true. Not only this: one must have
good reason to think, regarding whatever feature that in fact makes the belief justified,
that beliefs possessing that feature are likely to be true. The problem for foundational-
ists is that there is only one conceivable way foundational beliefs could count as justi-
fied: one would have to know, regarding whatever feature makes foundational beliefs
foundational, that beliefs with that feature were likely to be true. Yet this is not know-
able a priori, at least about any of the sources of empirical knowledge. Nor can it be
known a posteriori, for it could only be established using circular reasoning, which is
eschewed by foundationalists.
BonJour anticipates objections to his interpretation of the conditions required for
having good reason to think a belief true. In favor of givenists, he acknowledges that the
search for the given is a search for something that justifies foundational beliefs.
Nonetheless, they seek the impossible. There cannot be a state of mind that is able to
impart justification but needs no justification itself. To impart justification, a state must
have assertive content, and having assertive content suffices for standing in need of
If Sellars and BonJour are right that foundationalism is inadequate, then the door is
open for alternatives. Donald Davidson proposes a coherentist theory based on conclu-
sions about meaning and content. Meaning, coherence, and truth, he argues, are inter-
nally connected. The meaning of one's words and one's thoughts depends on one's
being interpretable as a coherent (indeed a rational) believer, most of whose beliefs are
true. Since a fully informed or omniscient interpreter would also interpret any believer
as having mostly true beliefs, it follows that all believers, ourselves included, have mostly
true beliefs. This would establish a further connection to justification, according to

Davidson, for in seeing that most of our beliefs are true, we gain a presumptive reason
in favor of retaining any arbitrary one of them.
Susan Haack sees merit in both the coherentist and foundationalist approaches. Her
aim is to connect epistemic justification essentially with truth-conduciveness. After
examining varieties of foundationalist and coherentist theories, she claims that we
remain in need of an account of how there could be both logical and causal relations
between experience and beliefs. Only a logical relation can ensure the rational or justi-
ficatory connection between experience and belief. And only a causal connection can
ensure the linking of empirical justification with truth. For an empirical worldly fact
can enter our cognitive economy only through experience.
Haack proposes to provide both the logical and causal connection by employing a
distinction between belief states (S-beliefs) and the contents of those states (C-beliefs).
She begins by giving an evidentialist account of justification: agent A is more/less justi-
fied in believing that p depending upon how good 1\s evidence is for p. The distinction
between S- and C-beliefs is then employed in characterizing 1\s evidence.1\s evidence
consists of three sorts of items: 1\s S-reasons, 1\s C-reasons, and 1\s experiential C-evi-
dence for believing that p. The S-reasons are themselves S-beliefs sustained ultimately
by 1\s experiential S-states. The role of experience in sustaining S-beliefs, Haack claims,
identifies what was right about experientialist foundationalism. 1\s C-reasons for
believing that p are the C-beliefs that serve as the contents of 1\s S-reasons for believing
that p. Coherentists were right to emphasize the non-linear character of C-reasons in
justification. No class of C-beliefs is basic in the nexus of C-reasons. Finally, 1\s experi-
ential C-evidence consists of true propositions to the effect that A is in a certain state,
viz. the state that constitutes 1\s experiential S-evidence for believing that p. It is the last
element of 1\s evidence, Haack believes, that supplies the necessary connection between
justification and truth. One might be tempted to doubt this: surely, a proposition to the
effect that it seems visually as if there is something green before me provides no guar-
anteed link to truth. It could very well be that my experience is unveridical. How could
the mere fact that I have an experience as of something green before me be evidence in
favor of there being something green before me? Haack's answer is that the appropriate
description of the experience characterizes it in a world-involving way. Thus, a visual
experience as if there being something green before me is to be described as the kind
of experience a normal subject would be in, in normal circumstances, when looking at
a green thing. This would seem to provide the link between justification and truth.
That I am in the kind of experiential state that is normally or typically caused by a
green thing does seem to make it objectively more probable that there is a green thing
before me.
Thus, we have foundherentism. Foundationalist elements survive in the claim that
experiential S-reasons form the causal bedrock, coherentist elements in the claim that
the structure of C-reasons do not have a linear structure. The connection with truth,
missed by coherentism and by many forms of foundationalism, is secured through the
claim that part of the C-evidence for a belief includes truths describing experiences in
terms of their typical external causes.
Like Haack, Ernest So sa attempts to reconcile coherentism and foundationalism -
the raft and the pyramid. Traditional foundationalism and coherentism alike are com-
mitted to a kind of "formal" foundationalism, which holds that epistemic conditions
supervene on non-epistemic conditions in a way that can be specified in general,

perhaps recursively. Formal foundationalism, according to Sosa, derives its plausibility

from the claims that epistemic conditions are normative and that all normative condi-
tions are supervenient. If a state of affairs is good, it must be good because it is a state
of pleasure or because it is a state of desire satisfaction, etc. It cannot be barely good or
good ultimately owing merely to the goodness of some other state( s). So, too, if a belief
is justified, some non-epistemic condition must account for its justification. Sosa goes
on to argue that the thesis of formal foundationalism conflicts with internalist theories
of justification (perhaps such as Sellars'). If one's justification for believing that p is
fixed ultimately by non -epistemic facts, then such justification cannot in every case also
require the possession of further justified beliefs.
In the final selection of this section, Peter Klein argues in favor of the often dis-
missed position of infinitism. Infinitism provides an account of justification accord-
ing to which the structure of justificatory reasons is infinite and non-repeating. Klein
argues that infinitism provides an acceptable account of rational beliefs, while other
epistemic theories, such as foundationalism and coherentism, cannot. This is because
infinitism is the only epistemic theory that can satisfy two plausible constraints upon
reasoning - that reasoning neither be arbitrary nor beg the question. Infinitism is
similar to foundationalism in holding that not every belief counts as a reason but dif-
fers to the extent that the infinitist also holds PAA (the principle of avoiding arbi-
trariness), which states that there are no foundational reasons and so every reason
stands in need of another reason. Infinitism is similar to coherentism in holding that
only reasons can justify a belief but differs to the extent that the infinitist also holds
PAC (the principle of avoiding circularity), which states that justifying reasons cannot
beg the question. For much of the paper Klein deals with the main objections to infin-
itism, including (1) the finite mind objection, (2) the objection that if some knowledge
is inferential then some knowledge must not be, (3) a reductio argument against the
possibility of an infinite regress providing a justification for beliefs, and (4) skeptical

Further Reading

Alston, William, "Two Types of Foundationalism," Chisholm, Roderick, The Foundations of Knowing
in Epistemic Justification (Ithaca, NY: Cornell (Minneapolis: University of Minnesota Press,
University Press, 1989), pp. 19-38. 1982).
Audi, Robert, The Structure of Justification (New - - , Theory of Knowledge (Englewood Cliffs:
York: Cambridge University Press, 1993). Prentice-Hall, 1966, 2nd edn 1977, 3rd edn
Bender, J. (ed.), The Current State of the Coherence 1989).
Theory (Dordrecht: Kluwer Academic Publishers, DePaul, Michael (ed.), Resurrecting Old-Fashioned
1989). Foundationalism (Langham, MD: Rowman
BonJour, Laurence, The Structure of Empirical and Littlefield, 2000).
Knowledge (Cambridge, MA: Harvard Fant!, Jeremy, "Modest Infinitism," Canadian
University Press, 1985). Journal of Philosophy 33 (2003), pp. 537-62.
BonJour, Laurence and Ernest Sosa, Epistemic Gillett, Carl, "Infinitism Redux? A Response to
Justification: Internalism vs. Externalism, Foun- Klein," Philosophy and Phenomenological Research
dations vs. Virtues (Oxford: Blackwell, 2003). 66 (2003),pp. 709-17.

Haack, Susan, Evidence and Inquiry: Towards Recon- Sellars, Wilfrid, "Empiricism and the Philosophy
struction in Epistemology (Oxford: Blackwell, of Mind," reprinted in Sellars, Science,
1993). Perception and Reality (London: Routledge
Lehrer, Keith, Theory of Knowledge (Boulder, CO: and Kegan Paul, 1963).
Westview Press, 1990). - - , "Givenness and Explanatory Coherence;'
Lewis, C. I., An Analysis of Knowledge and Journal of Philosophy 70 (1973), pp. 612-24.
Valuation (LaSalle, IL: Open Court Publishing Sosa, Ernest, Knowledge in Perspective: Selected
Company, 1946). Essays in Epistemology (Cambridge: Cambridge
Plantinga, Alvin, Warrant: The Current Debate University Press, 1991).
(Oxford: Oxford University Press, 1993).
Rescher, Nicholas, Methodological Pragmatism
(New York: New York University Press, 1977).
The Myth of the Given

Roderick M. Chisholm

1. The doctrine of "the given" involved two theses apprehension of "appearances" (etc.) -
about our knowledge. We may introduce them by our apprehension of the given.
means of a traditional metaphor:
Theses (A) and (B) constitute the "doctrine of the
(A) The knowledge which a person has at any given"; thesis (C), if a label were necessary,
time is a structure or edifice, many parts might be called "the phenomenalistic version"
and stages of which help to support each of the doctrine. The first two theses are essential
other, but which as a whole is supported to the empirical tradition in Western philosophy.
by its own foundation. The third is problematic for traditional empiri-
cism and depends in part, but only in part, upon
The second thesis is a specification of the first: the way in which the metaphor of the edifice and
its foundation is spelled out.
(B) The foundation of one's knowledge con- I believe it is accurate to say that, at the time at
sists (at least in part) of the apprehension which our study begins, most American episte-
of what have been called, variously, "sensa- mologists accepted the first two theses and thus
tions;, "sense-impressions," "appearances;' accepted the doctrine of the given. The expres-
"sensa," "sense-qualia;' and "phenomena." sion "the given" became a term of contemporary
philosophical vocabulary partly because of its
These phenomenal entities, said to be at the base use by c.1. Lewis in his Mind and the World-Order
of the structure of knowledge, are what was called (Scribner, 1929). Many of the philosophers who
"the given." A third thesis is sometimes associated accepted the doctrine avoided the expression
with the doctrine of the given, but the first two because of its association with other more con-
theses do not imply it. We may formulate it in the troversial parts of Lewis's book - a book which
terms of the same metaphor: might be taken (though mistakenly, I think) also
to endorse thesis (C), the "phenomenalistic ver-
(C) The only apprehension which is thus basic sion" of the doctrine. The doctrine itself - theses
to the structure of knowledge is our (A) and (B) - became a matter of general contro-
versy during the period of our survey.
Thesis (A) was criticized as being "absolute" and
thesis (B) as being overly "subjective." Both criti-
Originally published in R. Chisholm, Philosophy cisms may be found in some of the "instrumental-
(Englewood Cliffs, NJ: Prentice-Hall, 1964), pp. 261-86. istic" writings of John Dewey and philosophers

associated with him. They may also be found in one; the essential thing, if there is such an n, is
the writings of those philosophers of science that it provides a stopping place in the process, or
("logical empiricists") writing in the tradition of dialectic, of justification.
the Vienna Circle. (At an early stage of this tradi- We may now re-express, somewhat less meta-
tion, however, some of these same philosophers phorically, the two theses which I have called the
seem to have accepted all three theses.) Discussion "doctrine of the given:' The first thesis, that our
became entangled in verbal confusions - espe- knowledge is an edifice or structure having its
cially in connection with the uses of such terms own foundation, becomes (A) "every statement,
as "doubt;' "certainty," "appearance;' and "imme- which we are justified in thinking that we know, is
diate experience." Philosophers, influenced by justified in part by some statement which justifies
the work that Ludwig Wittgenstein had been itself." The second thesis, that there are appear-
doing in the 1930s, noted such confusions in ances ("the given") at the foundation of our
detail, and some of them seem to have taken the knowledge, becomes (B) "there are statements
existence of such confusions to indicate that (A) about appearances which thus justify themselves."
and (B) are false. l Many have rejected both theses (The third thesis - the "phenomenalistic version"
as being inconsistent with a certain theory of of the doctrine of the given - becomes (C) "there
thought and reference; among them, in addition are no self-justifying statements which are not
to some of the critics just referred to, we find phi- statements about appearances:')
losophers in the tradition of nineteenth-century Let us now turn to the first of the two theses
"idealism." constituting the doctrine of the given.
Philosophers of widely diverging schools now
believe that "the myth of the given" has finally 3. "Every justified statement is justified in part by
been dispelled. 2 I suggest, however, that, although some statement which justifies itself." Could it be
thesis (C), "the phenomenalistic version;' is false, that the question which this thesis is supposed to
the two theses, (A) and (B), which constitute the answer is a question which arises only because of
doctrine of the given are true. some mistaken assumption? If not, what are the
The doctrine is not merely the consequence of alternative ways of answering it? And did any of
a metaphor. We are led to it when we attempt to the philosophers with whom we are concerned
answer certain questions about justification - our actually accept any of these alternatives? The first
justification for supposing, in connection with two questions are less difficult to answer than the
anyone of the things that we know to be true, third.
that it is something that we know to be true. There are the following points of view to be
considered, each of which seems to have been
2. To the question "What justification do I have taken by some of the philosophers in the period
for thinking I know that a is true?" one may reply: of our survey.
"I know that b is true, and if I know that b is true
then I also know that a is true." And to the ques- (1) One may believe that the questions about
tion "What justification do I have for thinking I justification which give rise to our problem
know that b is true?" one may reply: "I know that are based upon false assumptions and hence
c is true, and if I know that c is true then I also that they should not be asked at all.
know that b is true." Are we thus led, sooner or (2) One may believe that no statement or claim
later, to something n of which one may say: "What is justified unless it is justified, at least in
justifies me in thinking I know that n is true is part, by some other justified statement or
simply the fact that n is true." If there is such an n, claim which it does not justify; this belief
then the belief or statement that n is true may be may suggest that one should continue the
thought of either as a belief or statement which process of justifying ad indefinitum, justify-
"justifies itself" or as a belief or statement which is ing each claim by reference to some addi-
itself "neither justified nor unjustified." The dis- tional claim.
tinction - unlike that between a Prime Mover (3) One may believe that no statement or claim
which moves itself and a Prime Mover which is a is justified unless it is justified by some
neither in motion nor at rest - is largely a verbal other justified statement or claim b, and that

b is not justified unless it in turn is justified in an extraordinarily misleading way. The ques-
by a; this would suggest that the process of tion "Does this doubtful statement rest upon
justifying is, or should be, circular. statements which are certain and incorrigible?" -
(4) One may believe that there are some par- if taken as one would ordinarily take it - does rest
ticular claims n at which the process of jus- upon a false assumption, for (we may assume) the
tifying should stop, and one may then hold statement that there is a clock on the mantelpiece
of any such claim n either: (a) n is justified is one which is not doubtful at all.
by something - viz., experience or observa- John Dewey, and some of the philosophers
tion - which is not itself a claim and which whose views were very similar to his, tended to
therefore cannot be said itself either to be suppose, mistakenly, that the philosophers who
justified or unjustified; (b) n is itself unjusti- asked themselves "What justification do I have
fied; (c) n justifies itself; or (d) n is neither for thinking I know this?" were asking the quite
justified nor unjustified. different question "What more can I do to verify
or confirm that this is so?" and they rejected
These possibilities, I think, exhaust the sig- answers to the first question on the ground that
nificant points of view; let us now consider them they were unsatisfactory answers to the second. 3
in turn. Philosophers influenced by Wittgenstein tended
to suppose, also mistakenly, but quite under-
4. "The question about justification which give standably, that the question "What justification
rise to the problem are based upon false assump- do I have for thinking I know this?" contains an
tions and therefore should not be asked at all." implicit challenge and presupposes that one does
The questions are not based upon false not have the knowledge concerned. They then
assumptions; but most of the philosophers who pointed out, correctly, that in most of the cases
discussed the questions put them in such a mis- where the question was raised (e.g., "What justi-
leading way that one is very easily misled into fies me in thinking I know that this is a table?")
supposing that they are based upon false there is no ground for challenging the claim to
assumptions. knowledge and that questions presupposing that
Many philosophers, following Descartes, the claim is false should not arise. But the question
Russell, and Husserl, formulated the questions "What justifies me in thinking I know that this is
about justification by means of such terms as a table?" does not challenge the claim to know
"doubt;"'certainty," and "incorrigibility;' and they that this is a table, much less presuppose that the
used, or misused, these terms in such a way that, claim is false.
when their questions were taken in the way in The "critique of cogency," as Lewis described
which one would ordinarily take them, they could this concern of epistemology, presupposes that we
be shown to be based upon false assumptions. are justified in thinking we know most of the things
One may note, for example, that the statement that we do think we know, and what it seeks to elicit
"There is a clock on the mantelpiece" is not self- is the nature of this justification. The enterprise is
justifying - for to the question "What is your jus- like that of ethics, logic, and aesthetics:
tification for thinking you know that there is a
clock on the mantelpiece?" the proper reply would
The nature of the good can be learned from
be to make some other statement (e.g., "I saw it
experience only if the content of experience be
there this morning and no one would have taken first classified into good and bad, or grades of
it away") - and one may then go on to ask "But better and worse. Such classification or grading
are there any statements which can be said to jus- already involves the legislative application of the
tify themselves?" If we express these facts, as many same principle which is sought. In logic, princi-
philosophers did, by saying that the statement ples can be elicited by generalization from exam-
"There is a clock on the mantelpiece" is one which ples only if cases of valid reasoning have first
is not "certain;' or one which may be "doubted," been segregated by some criterion. In esthetics,
and if we then go on to ask "Does this doubtful the laws of the beautiful may be derived from
statement rest upon other statements which are experience only if the criteria of beauty have first
certain and incorrigible?" then we are using terms been correctly applied. 4

When Aristotle considered an invalid mood of the form "It is probable that any particular a
the syllogism and asked himself "What is wrong is a b" may be explicated as saying "Most a's
with this?" he was not suggesting to himself that are b's." Or, still more accurately, to say "The
perhaps nothing was wrong; he presupposed probability is n that a particular a is a b" is
that the mood was invalid, just as he presup- to say "The limit of the relative frequency
posed that others were not, and he attempted, with which the property of being a b occurs
successfully, to formulate criteria which would in the class of things having the property a
enable us to distinguish the two types of mood. is n."
When we have answered the question "What (3) Hence, by (2), to show that a proposition
justification do I have for thinking I know this?" is probable it is necessary to show that a
what we learn, as Socrates taught, is something certain statistical frequency obtains; and, by
about ourselves. We learn, of course, what the jus- 0), to show that a certain statistical fre-
tification happens to be for the particular claim quency obtains it is necessary to show that it
with which the question is concerned. But we also is probable that the statistical frequency
learn, more generally, what the criteria are, if any, obtains; and therefore, by (2), to show that it
in terms of which we believe ourselves justified in is probable that a certain statistical frequency
counting one thing as an instance of knowing and obtains, it is necessary to show that a certain
another thing not. The truth which the philoso- frequency of frequencies obtains ... ,
pher seeks, when he asks about justification, is (4) And therefore "there is no Archimedean
"already implicit in the mind which seeks it, and point of absolute certainty left to which to
needs only to be elicited and brought to clear attach our knowledge of the world; all we
expression."5 have is an elastic net of probability connec-
Let us turn, then, to the other approaches to tions floating in open space" (p. 192).
the problem of "the given."
This reasoning suggests that an infinite
5. "No statement or claim would be justified number of steps must be taken in order to justify
unless it were justified, at least in part, by some acceptance of any statement. For, according to the
other justified claim or statement which it does reasoning, we cannot determine the probability
not justify." of one statement until we have determined that of
This regressive principle might be suggested a second, and we cannot determine that of the
by the figure of the building and its supports: no second until we have determined that of a third,
stage supports another unless it is itself supported and so on. Reichenbach does not leave the matter
by some other stage beneath it - a truth which here, however. He suggests that there is a way of
holds not only of the upper portions of the build- "descending" from this "open space" of probabil-
ing but also of what we call its foundation. And ity connections, but, if I am not mistaken, we can
the principle follows if, as some of the philoso- make the descent only by letting go of the concept
phers in the tradition of logical empiricism of justification.
seemed to believe, we should combine a frequency He says that, if we are to avoid the regress of
theory of probability with a probability theory of probabilities of probabilities of probabilities ...
justification. we must be willing at some point merely to make
In Experience and Prediction (u. of Chicago, a guess; "there will always be some blind posits on
1938) and in other writings, Hans Reichenbach which the whole concatenation is based" (p. 367).
defended a "probability theory of knowledge" The view that knowledge is to be identified with
which seemed to involve the following conten- certainty and that probable knowledge must be
tions: "imbedded in a framework of certainty" is "a
remnant of rationalism. An empiricist theory of
(1) To justify accepting a statement, it is neces- probability can be constructed only if we are will-
sary to show that the statement is probable. ing to regard knowledge as a system of posits."6
(2) To say of a statement that it is probable is to But if we begin by assuming, as we do, that
say something about statistical frequencies. there is a distinction between knowledge, on the
Somewhat more accurately, a statement of one hand, and a lucky guess, on the other, then we

must reject at least one of the premises of any Brand Blanshard, who defended the coherence
argument purporting to demonstrate that knowl- theory in The Nature of Thought, said that a pro-
edge is a system of "blind posits." The unaccept- position is true provided it is a member of an
able conclusion of Reichenbach's argument internally consistent system of propositions and
may be so construed as to follow from premises provided further this system is "the system in
(1) and (2); and premise (2) may be accepted as a which everything real and possible is coherently
kind of definition (though there are many who included."7 In one phase of the development of
believe that this definition is not adequate to all "logical empiricism" its proponents seem to have
of the uses of the term "probable" in science and held a similar view: a proposition - or, in this
everyday life). Premise (1), therefore, is the one case, a statement - is true provided it is a member
we should reject, and there are good reasons, of an internally consistent system of statements
I think, for rejecting (1), the thesis that "to justify and provided further this system is "the system
accepting a proposition it is necessary to show which is actually adopted by mankind, and espe-
that the proposition is probable." In fairness to cially by the scientists in our culture circle."8
Reichenbach, it should be added that he never A theory of truth is not, as such, a theory of
explicitly affirms premise (1); but some such justification. To say that a proposition is true is
premise is essential to his argument. not to say that we are justified in accepting it as
true, and to say that we are justified in accepting
6. "No statement or claim a would be justified it as true is not to say that it is true. Whatever
unless it were justified by some other justified merits the coherence theory may have as an
statement or claim b which would not be justified answer to certain questions about truth, it throws
unless it were justified in turn by a." no light upon our present epistemological ques-
The "coherence theory of truth;' to which some tion. If we accept the coherence theory, we may
philosophers committed themselves, is sometimes still ask, concerning any proposition a which we
taken to imply that justification may thus be cir- think we know to be true, "What is my justifica-
cular; I believe, however, that the theory does not tion for thinking I know that a is a member of the
have this implication. It does define "truth" as a system of propositions in which everything real
kind of systematic consistency of beliefs or propo- and possible is coherently included, or that a is a
sitions. The truth of a proposition is said to consist, member of the system of propositions which is
not in the fact that the proposition "corresponds" actually adopted by mankind and by the scientists
with something which is not itself a proposition, of our culture circle?" And when we ask such a
but in the fact that it fits consistently into a certain question, we are confronted, once again, with our
more general system of propositions. This view original alternatives.
may even be suggested by the figure of the build-
ing and its foundations. There is no difference in 7. If our questions about justification do have a
principle between the way in which the upper sto- proper stopping place, then, as I have said, there
ries are supported by the lower, and that in which are still four significant possibilities to consider.
the cellar is supported by the earth just below it, or We may stop with some particular claim and say
the way in which that stratum of earth is sup- of it that either.
ported by various substrata farther below; a good
building appears to be a part of the terrain on (a) it is justified by something - by experience,
which it stands and a good system of propositions or by observation - which is not itself a
is a part of the wider system which gives it its claim and which, therefore, cannot be said
truth. But these metaphors do not solve philo- either to be justified or to be unjustified;
sophical problems. (b) it is justified by some claim which refers to
The coherence theory did in fact appeal to our experience or observation, and the
something other than logical consistency; its pro- claim referring to our experience or obser-
ponents conceded that a system of false proposi- vation has no justification;
tions may be internally consistent and hence that (c) it justifies itself; or
logical consistency alone is no guarantee of truth. (d) it is itself neither justified nor unjustified.

The first of these alternatives leads readily to the It is relevant to note, moreover, that there
second, and the second to the third or to the may be conditions under which seeing a key
fourth. The third and the fourth - which differ does not justify one in accepting the statement
only verbally, I think - involve the doctrine of "There is a key" or in believing that one sees a
"the given." key. If the key were so disguised or concealed
Carnap wrote, in 1936, that the procedure of that the man who saw it did not recognize it to
scientific testing involves two operations: the be a key, then he might not be justified in accept-
"confrontation of a statement with observation" ing the statement "There is a key." Just as, if
and the "confrontation of a statement with previ- Mr. Jones unknown to anyone but himself is a
ously accepted statements." He suggested that thief, then the people who see him may be said
those logical empiricists who were attracted to the to see a thief - but none of those who thus sees
coherence theory of truth tended to lose sight of a thief is justified in accepting the statement
the first of these operations - the confrontation of "There is a thief."10
a statement with observation. He proposed a way Some of the writings of logical empiricists
of formulating simple "acceptance rules" for such suggest that, although some statements may be
confrontation and he seemed to believe that, justified by reference to other statements, those
merely by applying such rules, we could avoid the statements which involve "confrontation with
epistemological questions with which the adher- observation" are not justified at all. C. G. Hempel,
ents of "the given" had become involved. for example, wrote that "the acknowledgement of
Carnap said this about his acceptance rules: "If an experiential statements as true is psychologi-
no foreign language or introduction of new terms cally motivated by certain experiences; but within
is involved, the rules are trivial. For example: 'If the system of statements which express scientific
one is hungry, the statement "I am hungry" may knowledge or one's beliefs at a given time, they
be accepted'; or: 'If one sees a key one may accept function in the manner of postulates for which
the statement "there lies a key." "'9 As we shall note no grounds are offered."!! Hempel conceded,
later, the first of these rules differs in an impor- however, that this use of the term "postulate" is
tant way from the second. Confining ourselves for misleading and he added the following note of
the moment to rules of the second sort - "If one clarification: "When an experiential sentence is
sees a key one may accept the statement 'there lies accepted 'on the basis of direct experiential evi-
a key'" -let us ask ourselves whether the appeal to dence: it is indeed not asserted arbitrarily; but to
such rules enables us to solve our problem of the describe the evidence in question would simply
stopping place. mean to repeat the experiential statement itself.
When we have made the statement "There lies Hence, in the context of cognitive justification,
a key:' we can, of course, raise the question "What the statement functions in the manner of a prim-
is my justification for thinking I know, or for itive sentence:' 12
believing, that there lies a key?" The answer would When we reach a statement having the pro-
be "I see the key." We cannot ask "What is my jus- perty just referred to - an experiential statement
tification for seeing a key?" But we can ask "What such that to describe its evidence "would simply
is my justification for thinking that it is a key that mean to repeat the experiential statement itself" -
I see?" and, if we do see that the thing is a key, the we have reached a proper stopping place in the
question will have an answer. The answer might process of justification.
be "I see that it's shaped like a key and that it's in
the lock, and I remember that a key is usually 8. We are thus led to the concept of a belief, state-
here." The possibility of this question, and its ment, claim, proposition, or hypothesis, which
answer, indicates that we cannot stop our ques- justifies itself. To be clear about the concept, let us
tions about justification merely by appealing to note the way in which we would justify the state-
observation or experience. For, of the statement ment that we have a certain belief. It is essential,
"I observe that that is an A," we can ask, and of course, that we distinguish justifying the state-
answer, the question "What is my justification for ment that we have a certain belief from justifying
thinking that I observe that there is an A?" the belief itself.

Suppose, then, a man is led to say "I believe good president" might properly be taken to indi-
that Socrates is mortal" and we ask him "What is cate that, if the speaker does believe that Johnson
your justification for thinking that you believe, or is a good president, he is not yet firm in that belief.
for thinking that you know that you believe, that Hence there is a temptation to infer that, if we say
Socrates is mortal?" To this strange question, the of a man who is firm in his belief that Socrates is
only appropriate reply would be "My justification mortal, that he is "justified in believing that he
for thinking I believe, or for thinking that I know believes that Socrates is mortal," our statement
that I believe, that Socrates is mortal is simply the "makes no sense." And there is also a temptation
fact that I do believe that Socrates is mortal." One to go on and say that it "makes no sense" even to
justifies the statement simply by reiterating it; the say of such a man, that his statement "I believe
statement's justification is what the statement that Socrates is mortal" is one which is "justified"
says. Here, then, we have a case which satisfies for him. 14 After all, what would it mean to say of a
Hempel's remark quoted above; we describe the man's statement about his own belief, that he is
evidence for a statement merely by repeating the not justified in accepting it?15
statement. We could say, as C. J. Ducasse did, that The questions about what does or does not
"the occurrence of belief is its own evidence:'13 "make any sense" need not, however, be argued.
Normally, as I have suggested, one cannot jus- We may say, if we prefer, that the statements about
tify a statement merely by reiterating it. To the the beliefs in question are "neither justified nor
question "What justification do you have for unjustified:' Whatever mode of description we
thinking you know that there can be no life on the use, the essential points are two. First, we may
moon?" it would be inappropriate, and imperti- appeal to such statements in the process of justi-
nent, to reply by saying simply "There can be no fying some other statement or belief. If they have
life on the moon," thus reiterating the fact at issue. no justification they may yet be a justification -
An appropriate answer would be one referring to for something other than themselves. ("What jus-
certain other facts - for example, the fact that we tifies me in thinking that he and I are not likely to
know there is insufficient oxygen on the moon to agree? The fact that I believe that Socrates is
support any kind of life. But to the question mortal and he does not.") Second, the making of
"What is your justification for thinking you know such a statement does provide what I have been
that you believe so and so?" there is nothing to say calling a "stopping place" in the dialectic of justi-
other than "I do believe so and so." fication; but now, instead of signalizing the stop-
We may say, then, that there are some state- ping place by reiterating the questioned statement,
ments which are self-justifying, or which justify we do it by saying that the question of its justifi-
themselves. And we may say, analogously, that cation is one which "should not arise."
there are certain beliefs, claims, propositions, or It does not matter, then, whether we speak of
hypotheses which are self-justifying, or which certain statements which "justify themselves" or
justify themselves. A statement, belief, claim, of certain statements which are "neither justified
proposition, or hypothesis may be said to be self- nor unjustified;' for in either case we will be refer-
justifying for a person, if the person's justification ring to the same set of statements. I shall continue
for thinking he knows it to be true is simply the to use the former phrase.
fact that it is true. There are, then, statements about one's own
Paradoxically, these things I have described by beliefs ("I believe that Socrates is mortal") - and
saying that they "justify themselves" may also be for statements about many other psychological
described by saying they are "neither justified nor attitudes - which are self-justifying. "What justi-
unjustified." The two modes of description are fies me in believing, or in thinking I know, that I
two different ways of saying the same thing. hope to come tomorrow? Simply the fact that I do
If we are sensitive to ordinary usage, we may hope to come tomorrow." Thinking, desiring,
note that the expression "I believe that I believe" wondering, loving, hating, and other such attitudes
is ordinarily used, not to refer to a second-order are similar. Some, but by no means all, of the state-
belief about the speaker's own beliefs, but to indi- ments we can make about such attitudes, when the
cate that the speaker has not yet made up his attitudes are our own, are self-justifying - as are
mind. "I believe that I believe that Johnson is a statements containing such phrases as "I think

I remember" or "I seem to remember" (as distin- affect our sense organs or the conditions of obser-
guished from "I remember"), and "I think that I vation. One of the important epistemological
see" and "I think that I perceive" (as distinguished questions about appearances is "Are there self-
from "I see" and "I perceive"). Thus, of the two justifying statements about the ways in which
examples which Carnap introduced in connec- things appear?"
tion with his "acceptance rules" discussed above, Augustine, refuting the skeptics of the late
viz., "I am hungry" and "I see a key," we may say Platonic Academy, wrote:
that the first is self-justifying and the second not.
The "doctrine of the given," it will be recalled, I do not see how the Academician can refute him
tells us (A) that every justified statement, about who says: I know that this appears white to me, I
what we think we know, is justified in part by know that my hearing is delighted with this,
some statement which justifies itself and (B) that I know this has an agreeable odor, I know this
there are statements about appearances which tastes sweet to me, I know that this feels cold to
thus justify themselves. The "phenomenalistic me .... When a person tastes something, he can
version" of the theory adds (C) that statements honestly swear that he knows it is sweet to his
about appearances are the only statements which palate or the contrary, and that no trickery of the
justify themselves. What we have been saying is Greeks can dispossess him of that knowledge. l7
that the first thesis, (A), of the doctrine of the
given is true and that the "phenomenalistic ver- Suppose, now, one were to ask "What justification
sion," (C), is false; let us turn now to thesis (B). do you have for believing, or thinking you know,
that this appears white to you, or that tastes bitter
9. In addition to the self-justifying statements to you?" Here, too, we can only reiterate the state-
about psychological attitudes, are there self-justi- ment: "What justifies me in believing, or in think-
fying statements about "appearances"? Now we ing I know, that this appears white to me and that
encounter difficulties involving the word "appear- the tastes bitter to me is the fact that this does
ance" and its cognates. appear white to me and that does taste bitter."
Sometimes such words as "appears," "looks;' An advantage of the misleading substantive
and "seems" are used to convey what one might "appearance;' as distinguished from the verb
also convey by such terms as "believe." For exam- "appears," is that the former may be applied to
ple, if I say "It appears to me that General de those sensuous experiences which, though capa-
Gaulle was successful," or "General de Gaulle ble of being appearances of things, are actually
seems to have been successful," I am likely to not appearances of anything. Feelings, imagery,
mean only that I believe, or incline to believe, that and the sensuous content of dreams and halluci-
he has been successful; the words "appears" and nation are very much like the appearances of
"seems" serve as useful hedges, giving me an out, things and they are such that, under some cir-
should I find out later that de Gaulle was not suc- cumstances, they could be appearances of things.
cessful. When "appear" -words are used in this But if we do not wish to say that they are experi-
way, the statements in which they occur add noth- ences wherein some external physical things
ing significant to the class of "self-justifying" appears to us, we must use some expression other
statements we have just provided. Philosophers than "appear." For "appear;' in its active voice,
have traditionally assumed, however, that such requires a grammatical subject and thus requires
terms as "appear" may also be used in a quite dif- a term which refers, not merely to a way of appear-
ferent way. If this assumption is correct, as I ing, but also to something which appears.
believe it is, then this additional use does lead us But we may avoid both the objective" Something
to another type of self-justifying statement. appears blue to me," and the substantival "I sense
The philosophers who exposed the confusions a blue appearance." We may use another verb, say
to which the substantival expression "appearance" "sense," in a technical way, as many philosophers
gave rise were sometimes inclined to forget, I did, and equate it in meaning with the passive
think, that things do appear to us in various voice of "appear;' thus saying simply "I sense blue,"
ways.16 We can alter the appearance of anything or the like. Or better still, it seems to me, and at
we like merely by doing something which will the expense only of a little awkwardness, we can

use "appear" in its passive voice and say "1 am the "appear" -statements be interpreted as state-
appeared to blue." ments asserting a comparison between a present
Summing up, in our new vocabulary, we may object and any other object or set of objects.
say that the philosophers who talked of the The question now becomes "Can we formu-
"empirically given" were referring, not to "self- late any significant 'appear' -statements without
justifying" statements and beliefs generally, but thus comparing the way in which some object
only to those pertaining to certain "ways of being appears with the way in which some other object
appeared to." And the philosophers who objected appears, or with the way in which the object in
to the doctrine of the given, or some of them, question has appeared at some other time? Can
argued that no statement about "a way of being we interpret 'This appears white' in such a way
appeared to" can be "self-justifying." that it may be understood to refer to a present
way of appearing without relating that way of
10. Why would one suppose that "This appears appearing to any other object?" In Experience and
white" (or, more exactly, "1 am now appeared Prediction, Reichenbach defended his own view
white to") is not self-justifying? The most con- (and that of a good many others) in this way:
vincing argument was this: If I say "This appears
white;' then, as Reichenbach put it, 1 am making a The objection may be raised that a comparison
"comparison between a present object and a for- with formerly seen physical objects should be
merly seen object." 18 What I am saying could have avoided, and that a basic statement is to concern
been expressed by "The present way of appearing the present fact only, as it is. But such a reduction
is the way in which white objects, or objects which would make the basic statement empty. Its con-
I believe to be white, ordinarily appear." And this tent is just that there is a similarity between the
present object and one formerly seen; it is by
new statement, clearly, is not self-justifying; to
means of this relation that the present object is
justify it, as Reichenbach intimated, 1 must go on
described. Otherwise the basic statement would
and say something further - something about the
consist in attaching an individual symbol, say a
way in which 1 remember white objects to have number, to the present object; but the introduc-
appeared. tion of such a symbol would help us in no way,
"Appears white" may thus be used to abbrevi- since we could not make use of it to construct a
ate "appears the way in which white things nor- comparison with other things. Only in attaching
mally appear." Or "white thing:' on the other the same symbols to different objects, do we
hand, may be used to abbreviate "thing having the arrive at the possibility of constructing relations
color of things which ordinarily appear white." between the objects. (pp. 176-7)
The phrase "appear white" as it is used in the
second quoted expression cannot be spelled out It is true that, if an "appear" -statement is to be
in the manner of the first; for the point of the used successfully in communication, it must
second can hardly be put by saying that "white assert some comparison of objects. Clearly, if
thing" may be used to abbreviate "thing having I wish you to know the way things are now
the color of things which ordinarily appear the appearing to me, I must relate these ways of
way in which white things normally appear." In appearing to something that is familiar to you.
the second expression, the point of "appears But our present question is not "Can you under-
white" is not to compare a way of appearing with stand me if I predicate something of the way in
something else; the point is to say something which something now appears to me without
about the way of appearing itself. It is in terms of relating that way of appearing to something that
this second sense of "appears white" - that in is familiar to you?" The question is, more simply,
which one may say significantly and without "Can I predicate anything of the way in which
redundancy "Things that are white may normally something now appears to me without thereby
be expected to appear white" - that we are to comparing that way of appearing with something
interpret the quotation from Augustine above. else?" From the fact that the first of these two
And, more generally, when it was said that questions must be answered in the negative it
"appear" -statements constitute the foundation of does not follow that the second must also be
the edifice of knowledge, it was not intended that answered in the negative.'"

The issue is not one about communication, fore, as we might expect, this conception of
nor is it, strictly speaking, an issue about lan- thought and reference has been associated with
guage; it concerns, rather, the nature of thought skepticism.
itself. Common to both "pragmatism" and "ideal- Blanshard conceded that his theory of thought
ism," as traditions in American philosophy, is the "does involve a degree of scepticism regarding
view that to think about a thing, or to interpret or our present knowledge and probably all future
conceptualize it, and hence to have a belief about knowledge. In all likelihood there will never be a
it, is essentially to relate the thing to other things, proposition of which we can say, 'This that I am
actual or possible, and therefore to "refer beyond asserting, with precisely the meaning I now attach
it:' It is this view - and not any view about lan- to it, is absolutely true:"21 On Dewey's theory, or
guage or communication - that we must oppose on one common interpretation of Dewey's theory,
if we are to say of some statements about appear- it is problematic whether anyone can now be said
ing, or of any other statements, that they "justify to know that Mr Jones is working in his garden. A.
themselves." O. Lovejoy is reported to have said that, for Dewey,
To think about the way in which something is "I am about to have known" is as close as we ever
now appearing, according to the view in question, get to "I know."22 C. I. Lewis, in his An Analysis of
is to relate that way of appearing to something Knowledge and Valuation (Open Court, 1946)
else, possibly to certain future experiences, possi- conceded in effect that the conception of thought
bly to the way in which things of a certain sort suggested by his earlier Mind and the World-Order
may be commonly expected to appear. According does lead to a kind of skepticism; according to the
to the "conceptualistic pragmatism" of c.1. Lewis's later work there are "apprehensions of the given"
Mind and the World-Order (1929), we grasp the (cf. An Analysis, pp. 182-3) - and thus beliefs
present experience, any present way of appearing, which justify themselves.
only to the extent to which we relate it to some What is the plausibility of a theory of thought
future experience. 2o According to one interpreta- and reference which seems to imply that no one
tion of John Dewey's "instrumentalistic" version knows anything?
of pragmatism, the present experience may be Perhaps it is correct to say that when we think
used to present or disclose something else but it about a thing we think about it as having certain
does not present or disclose itself. And according properties. But why should one go on to say that
to the idealistic view defended in Brand to think about a thing must always involve think-
Blanshard's The Nature of Thought, we grasp our ing about some other thing as well? Does thinking
present experience only to the extent that we are about the other thing then involve thinking about
able to include it in the one "intelligible system of some third thing? Or can we think about one
universals" (vol. I, p. 632). thing in relation to a second thing without thereby
This theory of reference, it should be noted, thinking of a third thing? And if we can, then why
applies not only to statements and beliefs about can we not think of one thing - of one thing as
"ways of being appeared to" but also to those having certain properties - without thereby relat-
other statements and beliefs which I have called ing it to another thing?
"self-justifying." If "This appears white," or "I am The linguistic analogue of this view of thought
appeared white to," compares the present experi- is similar. Why should one suppose - as
ence with something else, and thus depends for Reichenbach supposed in the passage cited above
its justification upon what we are justified in and as many others have also supposed - that to
believing about the something else, then so, too, refer to a thing, in this instance to refer to a way of
does "I believe that Socrates is mortal" and "I appearing, is necessarily to relate the thing to
hope that the peace will continue." This general some other thing?
conception of thought, therefore, would seem to Some philosophers seem to have been led to
imply that no belief or statement can be said to such a view of reference as a result of such consid-
justify itself. But according to what we have been erations as the following: We have imagined a
saying, if there is no belief or statement which man saying, in agreement with Augustine, "It just
justifies itself, then it is problematic whether any does appear white - and that is the end of the
belief or statement is justified at all. And there- matter." Let us consider now the possible reply

"That it is not the end of the matter. You are edifice: our knowledge of the world is a structure
making certain assumptions about the language supported entirely by a foundation of such self-
you are using; you are assuming, for example, that justifying statements or beliefs. We should recall,
you are using the word 'white; or the phrase however, that the answers to our original Socratic
'appears white; in the way in which you have for- questions had two parts. When asked "What is
merly used it, or in the way in which it is ordinar- your justification for thinking that you know a?"
ily used, or in the way in which it would ordinarily one may reply "I am justified in thinking I know
be understood. And if you state your justification a, because (1) I know band (2) ifI know b then
for this assumption, you will refer to certain other I know a." We considered our justification for the
things - to yourself and to other people, to the first part of this answer, saying "I am justified in
word 'white; or to the phrase 'appears white: and thinking I know b, because (1) I know c and (2) if
to what the word or phrase has referred to or might I know c then I know b:' And then we considered
refer to on other occasions. And therefore, when our justification for the first part of the second
you say 'This appears white' you are saying some- answer, and continued in this fashion until we
thing, not only about your present experience, but reached the point of self-justification. In thus
also about all of these other things as well." moving toward "the given:' we accumulated, step
The conclusion of this argument - the part by step, a backlog of claims that we did not
that follows the "therefore" - does not follow attempt to justify - those claims constituting the
from the premises. In supposing that the argu- second part of each of our answers. Hence our
ment is valid, one fails to distinguish between (1) original claim - "I know that a is true" - does not
what it is that a man means to say when he uses rest upon "the given" alone; it also rests upon all
certain words and (2) his assumptions concern- of those other claims that we made en route. And
ing the adequacy of these words for expressing it is not justified unless these other claims are
what it is that he means to say; one supposes, mis- justified.
takenly, that what justifies (2) must be included A consideration of these other claims will lead
in what justifies (1). A Frenchman, not yet sure of us, I think, to at least three additional types of
his English, may utter the words "There are apples "stopping place," which are concerned, respec-
in the basket," intending thereby to express his tively, with memory, perception, and what Kant
belief that there are potatoes in the basket. If we called the a priori. Here I shall comment briefly
show him that he has used the word "apples" on the first two.
incorrectly, and hence that he is mistaken in his It is difficult to think of any claim to empiri-
assumptions about the ways in which English cal knowledge, other than the self-justifying
speaking people use and understand the word statements we have just considered, which does
"apples," we have not shown him anything rele- not to some extent rest upon an appeal to
vant to his belief that there are apples in the memory. But the appeal to memory- "I remem-
basket. ber that A occured" - is not self-justifying. One
Logicians now take care to distinguish between may ask "And what is your justification for
the use and mention oflanguage (e.g., the English thinking that you remember that A occured?"
word "Socrates" is mentioned in the sentence and the question will have an answer - even if
'''Socrates' has eight letters" and is used but not the answer is only the self-justifying "I think
mentioned, in "Socrates is a Greek.")23 As we shall that I remember that A occurred." The statement
have occasion to note further, the distinction has "I remember that A occured" does, of course,
not always been observed in writings on episte- imply "A occurred"; but "I think that I remem-
mology. ber that A occurred" does not imply "A occurred"
and hence does not imply "I remember that A
11. If we decide, then, that there is a class of occured." For we can remember occasions - at
beliefs or statements which are "self-justifying," least we think we can remember them - when
and that this class is limited to certain beliefs or we learned, concerning some event we had
statements about our own psychological states thought we remembered, that the event had not
and about the ways in which we are "appeared to," occurred at all, and consequently that we had
we may be tempted to return to the figure of the not really remembered it. When we thus find

that one memory conflicts with another, or, formulation, and validity, of the rules of logic.
more accurately, when we thus find that one Nor do they differ from the problems posed by
thing that we think we remember conflicts with the moral and religious "cognitivists" (the "non-
another thing that we think we remember, we intuitionistic cognitivists") that I have referred to
may correct one or the other by making further elsewhere. The status of ostensible memories and
inquiry; but the results of any such inquiry will perceptions, with respect to that experience
always be justified in part by other memories, which is their "source;' is essentially like that
or by other things that we think that we remem- which such "cognitivists" claim for judgments
ber. How then are we to choose between what having an ethical or theological subject matter.
seem to be conflicting memories? Under what Unfortunately, it is also like that which other
conditions does "I think that I remember that "enthusiasts" claim for still other types of subject
A occurred" serve to justify "I remember that A matter.
The problem is one of formulating a rule of 12. What, then, is the status of the doctrine of
evidence - a rule specifying the conditions "the given" - of the "myth of the given"? In my
under which statements about what we think opinion, the doctrine is correct in saying that
we remember can justify statements about what there are some beliefs or statements which are
we do remember. A possible solution, in very "self-justifying" and that among such beliefs and
general terms, is "When we think that we statements are some which concern appearances
remember, then we are justified in believing or "ways of being appeared to;" but the "phenom-
that we do remember, provided that what we enalistic version" of the doctrine is mistaken in
think we remember does not conflict with any- implying that our knowledge may be thought of
thing else that we think we remember; when as an edifice which is supported by appearances
what we think we remember does conflict with alone. 26 The cognitive significance of "the empiri-
anything else we think we remember, then, of cally given" was correctly described - in a vocab-
the two conflicting memories (more accurately, ulary rather different from that which I have been
ostensible memories) the one that is justified is using - by John Dewey:
the one that fits in better with the other things
that we think we remember." Ledger Wood The alleged primacy of sensory meanings is
made the latter point by saying that the justified mythical. They are primary only in logical
memory is the one which "coheres with the status; they are primary as tests and confirma-
system of related memories"; C. I. Lewis used tion of inferences concerning matters of fact,
"congruence" instead of "coherence."24 But we not as historic originals. For, while it is not usu-
cannot say precisely what is meant by "fitting ally needful to carry the check or test of theo-
retical calculations to the point of irreducible
in," "coherence," or "congruence" until certain
sensa, colors, sounds, etc., these sensa form a
controversial questions of confirmation theory
limit approached in careful analytic certifica-
and the logic of probability have been answered.
tions, and upon critical occasions it is necessary
And it may be that the rule of evidence is
to touch the limit .... Sensa are the class of irre-
too liberal; perhaps we should say, for example, ducible meanings which are employed in verify-
that when two ostensible memories conflict ing and correcting other meanings. We actually
neither one of them is justified. But these are set out with much coarser and more inclusive
questions which have not yet been satisfactorily meanings and not till we have met with failure
answered. from their use do we even set out to discover
If we substitute "perceive" for "remember" in those ultimate and harder meanings which are
the foregoing, we can formulate a similar set of sensory in character. 27
problems about perception; these problems, too,
must await solution. 25 The Socratic questions leading to the concept of
The problems involved in formulating such "the given" also lead to the concept of "rules of
rules of evidence, and in determining the validity evidence:' Unfortunately some of the philoso-
of these rules, do not differ in any significant way phers who stressed the importance of the former
from those which arise in connection with the concept tended to overlook that of the latter.


Philosophers in other traditions also noted discussed in this collection of essays, the
these confusions. See, for example, John Wild, fault is with our conventional classification
"The Concept of the Given in Contemporary of philosophical disciplines. The book, which
Philosophy:' PhilosophyandPhenomenological is concerned with an area falling between
Research I (1940), pp. 70-82. logic and metaphysics, is one of the most
2 The expression "myth of the given" was used important philosophical works written by
by Wilfrid Sellars in "Empiricism and the an American during the period being sur-
Philosophy of Mind," in Herbert Feigl and veyed.
Michael Scriven (eds), Foundations of Science 11 C. G. Hempel, "Some Theses on Empirical
and the Concepts of Psychology and Psycho- Certainty," Review of Metaphysics V (1952),
analysis, Minnesota Studies in the Philosophy pp. 621-9; the quotation is from p. 621.
of Science, vol. I (u. of Minn., 1956), pp. 12 Ibid., p. 628. Hempel's remarks were made in
253-329. an "Exploration" in which he set forth sev-
3 Dewey also said that, instead of trying to eral theses about "empirical certainty" and
provide "Foundations for Knowledge:' the then replied to objections by Paul Weiss,
philosopher should apply "what is known Roderick Firth, Wilfrid Sellars, and myself.
to intelligent conduct of the affairs of 13 C. J. Ducasse, "Propositions, Truth, and the
human life" to "the problems of men." John Ultimate Criterion of Truth;' Philosophy and
Dewey, Problems of Men (Philosophical, Phenomenological Research IV (1939), pp.
1946), pp. 6-7. 317-40; the quotation is from p. 339.
4 C. I. Lewis, Mind and the World-Order 14 Cf. Norman Malcolm, "Knowledge of Other
(Scribner, 1929), p. 29. Minds," Journal of Philosophy LV (1958), pp.
5 Ibid., p. 19. Cf. Hans Reichenbach, Experience 969-78. Reprinted in Malcolm, Knowledge
and Prediction (University of Chicago, 1938), and Certainty: Essays and Lectures (Prentice-
p. 6; C. J. Ducasse, "Some Observations Hall, 1963).
Concerning the Nature of Probability;' 15 The principle behind this way of looking at
Journal of Philosophy XXXVIII (1941), esp. the matter is defended in detail by Max Black
pp.400-1. in Language and Philosophy (Cornell, 1949),
6 Hans Reichenbach, "Are Phenomenal pp. 16 ff.
Reports Absolutely Certain?" Philosophical 16 One of the best criticisms of the "appear-
Review LXI (1952), pp. 147-59; the quota- ance" (or "sense-datum") terminology was
tion is from p. 150. O. K. Bouwsma's "Moore's Theory of Sense-
7 Brand Blanshard, The Nature of Thought, Data;' in The Philosophy of G. E. Moore, ed.
vol. II (Macmillan, 1940), p. 276. Schilpp, pp. 201-21. In Perceiving: A
8 C. G. Hempel, "On the Logical Positivists' Philosophical Study (Cornell, 1957), I tried to
Theory of Truth," Analysis II (1935), pp. call attention to certain facts about appear-
49-59; the quotation is from p. 57. ing which, I believe, Bouwsma may have
9 Rudolf Carnap, "Truth and Confirmation," overlooked.
in Herbert Feigl and W. S. Sellars (eds), 17 Augustine, Contra academicos, xi, 26; trans-
Readings in Philosophical Analysis (Appleton, lated by Sister Mary Patricia Garvey as
1949), p. 125. The portions of the article Saint Augustine Against the Academicians
quoted above first appeared in "Wahrheit und (Marquette, 1942); the quotations are from
Bewahrung," Actes du congres international de pp.68-9.
philosophie scientifique, IV (Paris; 1936), pp. 18 Experience and Prediction, p. 176.
18-23. 19 It may follow, however, that "the vaunted
10 Cf. Nelson Goodman, The Structure of incorrigibility of the sense-datum language
Appearance (Harvard, 1951), p. 104. If can be achieved only at the cost of its perfect
Goodman's book, incidentally, is not utility as a means of communication" (Max

Black, Problems of Analysis (Cornell, 1954), of inquiry, however, was not intended to be
p. 66), and doubtless, as Black added, it an epistemology and he did not directly
would be "misleading, to say the least" to address himself to the questions with which
speak of a "language that cannot be commu- we are here concerned.
nicated" - cf. Wilfrid Sellars, "Empiricism 23 Cf. W. V. Quine, Mathematical Logic (Norton,
and the Philosophy of Mind" - but these 1940; rev. edn, Harvard, 1951), sec. 4.
points do affect the epistemological question 24 Ledger Wood, The Analysis of Knowledge
at issue. (Princeton, 1941), p. 81; C. I. Lewis, An
20 This doctrine was modified in Lewis's later Analysis of Knowledge and Valuation, p. 334.
An Analysis of Knowledge and Valuation 25 Important steps toward solving them were
(Open Court, 1946) in a way which enabled taken by Nelson Goodman in "Sense and
him to preserve the theory of the given. Certainty," Philosophical Review LXI (1952),
21 The Nature of Thought, vol. II, pp. 269-70. pp. 160-7, and by Israel Scheffler in "On
Blanshard added, however, that "for all the Justification and Commitment:' Journal of
ordinary purposes of life" we can justify Philosophy, LI (1954), pp. 180-90. The
some beliefs by showing that they cohere former paper is reprinted in Philosophy of
"with the system of present knowledge"; and Knowledge, ed. Roland Houde and J. P.
therefore, he said, his theory should not be Mullally (Lippincott, 1960), pp. 97-103.
described as being "simply sceptical" (vol. II, 26 Alternatives to the general metaphor of the
p. 271). Cf. W. H. Werkmeister, The Basis edifice are proposed by W. V. Quine in the
and Structure of Knowledge (Harper, 1948), introduction to Methods ofLogic (Holt, 1950;
part II. rev. edn, 1959), in From a Logical Point of
22 Quoted by A. E. Murphy in "Dewey's View (Harvard, 1953), and in Word and
Epistemology and Metaphysics:' in P. A. Object (Wiley, 1960).
Schilpp (ed.), The Philosophy of John Dewey, 27 John Dewey, Experience and Nature, 2nd edn
(Northwestern, 1939), p. 203. Dewey's theory (Norton, 1929), p. 327.

Does Empirical Knowledge

Have a Foundation?

Wilfrid Sellars

I have arrived at a stage in my argument which is, objects with the vocable "This is green:' Not only
at least prima facie, out of step with the basic pre- must the conditions be of a sort that is appropri-
suppositions of logical atomism. Thus, as long as ate for determining the color of an object by look-
looking green is taken to be the notion to which ing, the subject must know that conditions of this
being green is reducible, it could be claimed with sort are appropriate. And while this does not imply
considerable plausibility that fundamental con- that one must have concepts before one has them,
cepts pertaining to observable fact have that it does imply that one can have the concept of
logical independence of one another which is green only by having a whole battery of concepts
characteristic of the empiricist tradition. Indeed, of which it is one element. It implies that while the
at first sight the situation is quite disquieting, for process of acquiring the concept green may -
if the ability to recognize that x looks green pre- indeed does - involve a long history of acquiring
supposes the concept of being green, and if this in piecemeal habits of response to various objects in
turn involves knowing in what circumstances to various circumstances, there is an important sense
view an object to ascertain its color, then, since in which one has no concept pertaining to the
one can scarcely determine what the circum- observable properties of physical objects in Space
stances are without noticing that certain objects and Time unless one has them all- and, indeed, as
have certain perceptible characteristics - includ- we shall see, a great deal more besides.
ing colors - it would seem that one couldn't form [ J
the concept of being green, and, by parity of rea-
soning, of the other colors, unless he already had One of the forms taken by the Myth of the Given
them. is the idea that there is, indeed must be, a struc-
Now, it just won't do to reply that to have the ture of particular matter of fact such that (a) each
concept of green, to know what it is for something fact can not only be noninferentially known to be
to be green, it is sufficient to respond, when one is the case, but presupposes no other knowledge
in point of fact in standard conditions, to green either of particular matter of fact, or of general
truths; and (b) such that the non inferential
knowledge of facts belonging to this structure
Originally published in H. Feigl and M. Scriven (eds), constitutes the ultimate court of appeals for all
The Foundations ofScience and the Concepts ofPsychology factual claims - particular and general - about
and Psychoanalysis, Minnesota Studies in the Philosophy the world. It is important to note that I character-
of Science, vol. I (Minneapolis: University of Minnesota ized the knowledge of fact belonging to this
Press, 1956), pp. 293-300. stratum as not only noninferential, but as

presupposing no knowledge of other matter of thought I am developing in terms of the fact-

fact, whether particular or general. It might be stating and observation-reporting roles of certain
thought that this is a redundancy, that knowledge sentences). Roughly, two verbal performances
(not belief or conviction, but knowledge) which which are tokens of a non-token-reflexive sen-
logically presupposes knowledge of other facts tence can occur in widely different circum-
must be inferential. This, however, as I hope to stances and yet make the same statement;
show, is itself an episode in the Myth. whereas two tokens of a token-reflexive sentence
Now, the idea of such a privileged stratum of can make the same statement only if they are
fact is a familiar one, though not without its dif- uttered in the same circumstances (according to
ficulties. Knowledge pertaining to this level is a relevant criterion of sameness). And two
tokens of a sentence, whether it contains a
non-inferential, yet it is, after all, knowledge. It is
token-reflexive expression - over and above a
ultimate, yet it has authority. The attempt to make
tensed verb - or not, can make the same report
a consistent picture of these two requirements
only if, made in all candor, they express the pres-
has traditionally taken the following form:
ence - in some sense of "presence" - of the state
of affairs that is being reported; if, that is, they
Statements pertaining to this level, in order to stand in that relation to the state of affairs,
"express knowledge" must not only be made, but, whatever the relation may be, by virtue of which
so to speak, must be worthy of being made, cred- they can be said to formulate observations of it.
ible, that is, in the sense of worthy of credence. lt would appear, then, that there are two ways
Furthermore, and this is a crucial point, they in which a sentence token can have credibility: (1)
must be made in a way which involves this credi- The authority may accrue to it, so to speak, from
bility. For where there is no connection between above, that is, as being a token of a sentence type
the making of a statement and its authority, the all the tokens of which, in a certain use, have cred-
assertion may express conviction, but it can ibility, e.g. "2 + 2 = 4:' In this case, let us say that
scarcely be said to express knowledge. token credibility is inherited from type authority.
The authority - the credibility - of statements (2) The credibility may accrue to it from the fact
pertaining to this level cannot exhaustively consist that it came to exist in a certain way in a certain set
in the fact that they are supported by other state- of circumstances, e.g. "This is red." Here token
ments, for in that case all knowledge pertaining to credibility is not derived from type credibility.
this level would have to be inferential, which not Now, the credibility of some sentence types
only contradicts the hypothesis, but flies in the appears to be intrinsic - at least in the limited
face of good sense. The conclusion seems inevita- sense that it is not derived from other sentences,
ble that if some statements pertaining to this level type or token. This is, or seems to be, the case
are to express noninferential knowledge, they must with certain sentences used to make analytic
have a credibility which is not a matter of being statements. The credibility of some sentence
supported by other statements. Now there does types accrues to them by virtue of their logical
seem to be a class of statements which fill at least relations to other sentence types, thus by virtue
part of this bill, namely such statements as would of the fact that they are logical consequences of
be said to report observations, thus, "This is red." more basic sentences. It would seem obvious,
These statements, candidly made, have authority. however, that the credibility of empirical sen-
Yet they are not expressions of inference. How, tence types cannot be traced without remainder
then, is this authority to be understood? to the credibility of other sentence types. And
Clearly, the argument continues, it springs since no empirical sentence type appears to have
from the fact that they are made in just the cir- intrinsic credibility, this means that credibility
cumstances in which they are made, as is indi- must accrue to some empirical sentence types by
cated by the fact that they characteristically, virtue of their logical relations to certain sen-
though not necessarily or without exception, tence tokens, and, indeed, to sentence tokens the
involve those so-called token-reflexive expres- authority of which is not derived, in its turn,
sions which, in addition to the tenses of verbs, from the authority of sentence types.
serve to connect the circumstances in which a The picture we get is that of there being two
statement is made with its sense. (At this point ultimate modes of credibility: (1) The intrinsic
it will be helpful to begin putting the line of credibility of analytic sentences, which accrues to

tokens as being tokens of such a type; (2) the weakened beyond all recognition into the bare
credibility of such tokens as "express observa- notion of exhibiting a uniformity - in which case
tions;' a credibility which flows from tokens to the lightning, thunder sequence would "follow a
types. rule" - then it is the knowledge or belief that the
circumstances are of a certain kind, and not the
Let us explore this picture, which is common to mere fact that they are of this kind, which con-
all traditional empiricisms, a bit further. How is tributes to bringing about the action.
the authority of such sentence tokens as "express In the light of these remarks it is clear that if
observational knowledge" to be understood? It has observation reports are construed as actions, if
been tempting to suppose that in spite of the obvi- their correctness is interpreted as the correctness
ous differences which exist between "observation of an action, and if the authority of an observa-
reports" and "analytic statements;' there is an tion report is construed as the fact that making it
essential similarity between the ways in which they is "following a rule" in the proper sense of this
come by their authority. Thus, it has been claimed, phrase, then we are face to face with giveness in
not without plausibility, that whereas ordinary its most straightforward form. For these stipula-
empirical statements can be correctly made with- tions commit one to the idea that the authority of
out being true, observation reports resemble ana- Konstatierungen rests on nonverbal episodes of
lytic statements in that being correctly made is a awareness - awareness that something is the case,
sufficient as well as necessary condition of their e.g. that this is green - which nonverbal episodes
truth. And it has been inferred from this - some- have an intrinsic authority (they are, so to speak,
what hastily, I believe - that "correctly making" the "self-authenticating") which the verbal perform-
report "This is green" is a matter of "following the ances (the Konstatierungen) properly performed
rules for the use of 'this; 'is' and 'green:" "express:' One is committed to a stratum of
Three comments are immediately necessary: authoritative nonverbal episodes ("awareness")
(1) First a brief remark about the term the authority of which accrues to a superstructure
"report." In ordinary usage a report is a report of verbal actions, provided that the expressions
made by someone to someone. To make a report occurring in these actions are properly used.
is to do something. In the literature of epistemol- These self-authenticating episodes would consti-
ogy,however, the word "report" or"Konstatierung" tute the tortoise on which stands the elephant on
has acquired a technical use in which a sentence which rests the edifice of empirical knowledge.
token can playa reporting role (a) without being The essence of the view is the same whether these
an overt verbal performance, and (b) without intrinsically authoritative episodes are such items
having the character of being "by someone to as the awareness that a certain sense content is
someone" - even oneself. There is, of course, green or such items as the awareness that a certain
such a thing as "talking to oneself" - in foro physical object looks to someone to be green.
interno - but, as I shall be emphasizing in the But what is the alternative? We might begin by
closing stages of my argument, it is important trying something like the following: An overt or
not to suppose that all "covert" verbal episodes covert token of "This is green" in the presence of
are of this kind. a green item is a Konstatierung and express obser-
(2) My second comment is that while we shall vational knowledge if and only if it is a manifesta-
not assume that because "reports" in the ordinary tion of a tendency to produce overt or covert
sense are actions, "reports" in the sense of tokens of "This is green" - given a certain set - if
Konstatierungen are also actions, the line of and only if a green object is being looked at in
thought we are considering treats them as such. standard conditions. Clearly on this interpreta-
In other words, it interprets the correctness of tion the occurrence of such tokens of "This is
Konstatierungen as analogous to the rightness of green" would be "following a rule" only in the
actions. Let me emphasize, however, that not all sense that they are instances of a uniformity, a
ought is ought to do, nor all correctness the cor- uniformity differing from the lightning-thunder
rectness of actions. case in that it is an acquired causal characteristic
(3) My third comment is that if the expression of the language user. Clearly the above sugges-
"following a rule" is taken seriously, and is not tion, which corresponds to the "thermometer

view" criticized by Professor Price, and which we very like it is true. The point I wish to make now,
have already rejected elsewhere, won't do as it however, is that if it is true, then it follows, as a
stands. Let us see, however, if it can't be revised to matter of simple logic, that one couldn't have
fit the criteria I have been using for "expressing observational knowledge of any fact unless one
observational knowledge." knew many other things as well. And let me
The first hurdle to be jumped concerns the emphasize that the point is not taken care of by
authority which, as I have emphasized, a sentence distinguishing between knowing how and know-
token must have in order that it may be said to ing that, and admitting that observational knowl-
express knowledge. Clearly, on this account the edge requires a lot of "know how." For the point is
only thing that can remotely be supposed to con- specifically that observational knowledge of any
stitute such authority is the fact that one can infer particular fact, e.g. that this is green, presupposes
the presence of a green object from the fact that that one knows general facts of the form X is a
someone makes this report. As we have already reliable symptom of Y. And to admit this requires
noticed, the correctness of a report does not have an abandonment of the traditional empiricist
to be construed as the rightness of an action. idea that observational knowledge "stands on its
A report can be correct as being an instance of a own feet." Indeed, the suggestion would be anath-
general mode of behavior which, in a given lin- ema to traditional empiricists for the obvious
guistic community, it is reasonable to sanction reason that by making observational knowledge
and support. presuppose knowledge of general facts of the form
The second hurdle is, however, the decisive X is a reliable symptom ofY, it runs counter to the
one. For we have seen that to be the expression of idea that we come to know general facts of this
knowledge, a report must not only have authority, form only after we have come to know by obser-
this authority must in some sense be recognized vation a number of particular facts which support
by the person whose report it is. And this is a the hypothesis that X is a symptom ofY.
steep hurdle indeed. For if the authority of the And it might be thought that there is an obvi-
report "This is green" lies in the fact that the exist- ous regress in the view we are examining. Does it
ence of green items appropriately related to the not tell us that observational knowledge at time t
perceiver can be inferred from the occurrence of presupposes knowledge of the form X is a reliable
such reports, it follows that only a person who is symptom of Y, which presupposes prior observa-
able to draw this inference, and therefore who tional knowledge, which presupposes other knowl-
has not only the concept green, but also the con- edge of the form X is a reliable symptom of Y,
cept of uttering "This is green" - indeed, the which presupposes still other, and prior observa-
concept of certain conditions of perception, those tional knowledge, and so on? This charge, how-
which would correctly be called "standard condi- ever, rests on too simple, indeed a radically
tions" - could be in a position to token "This is mistaken, conception of what one is saying of
green" in recognition of its authority. In other Jones when one says that he knows that p. It is not
words, for a Konstatierung "This is green" to just that the objection supposes that knowing is
"express observational knowledge;' not only must an episode; for clearly there are episodes which we
it be a symptom or sign of the presence of a green can correctly characterize as knowings, in partic-
object in standard conditions, but the perceiver ular, observings. The essential point is that in
must know that tokens of "This is green" are symp- characterizing an episode or a state as that of
toms of the presence of green objects in conditions knowing, we are not giving an empirical descrip-
which are standard for visual perception. tion of that episode or state; we are placing it in
Now it might be thought that there is some- the logical space of reasons, of justifying and
thing obviously absurd in the idea that before a being able to justify what one says.
token uttered by, say, Jones could be the expres- Thus, all that the view I am defending requires
sion of observational knowledge, Jones would is that no tokening by S now of "This is green" is
have to know that overt verbal episodes of this to count as "expressing observational knowl-
kind are reliable indicators of the existence, suit- edge" unless it is also correct to say of S that he
ably related to the speaker, of green objects. I do now knows the appropriate fact of the form X is
not think that it is. Indeed, I think that something a reliable symptom ofY, namely that {and again I

oversimplify) utterances of "This is green" are "knowings in presence" which are presupposed
reliable indicators of the presence of green objects by all other knowledge, both the knowledge of
in standard conditions of perception. And while general truths and the knowledge "in absence" of
the correctness of this statement about Jones other particular matters of fact. Such is the frame-
requires that Jones could now cite prior particular work in which traditional empiricism makes its
facts as evidence for the idea that these utterances characteristic claim that the perceptually given is
are reliable indicators, it requires only that it is the foundation of empirical knowledge.
correct to say that Jones now knows, thus remem- If I reject the framework of traditional empiri-
bers, that these particular facts did obtain. It does cism, it is not because I want to say that empirical
not require that it be correct to say that at the knowledge has no foundation. For to put it this
time these facts did obtain he then knew them to way is to suggest that it is really "empirical knowl-
obtain. And the regress disappears. edge so-called;' and to put it in a box with rumors
Thus, while Jones' ability to give inductive and hoaxes. There is clearly some point to the pic-
reasons today is built on a long history of acquir- ture of human knowledge as resting on a level of
ing and manifesting verbal habits in perceptual propositions - observation reports - which do
situations, and, in particular, the occurrence of not rest on other propositions in the same way as
verbal episodes, e.g. "This is green," which is other propositions rest on them. On the other
superficially like those which are later properly hand, I do wish to insist that the metaphor of
said to express observational knowledge, it does "foundation" is misleading in that it keeps us
not require that any episode in this prior time be from seeing that if there is a logical dimension in
characterizeable as expressing knowledge. (At which other empirical propositions rest on obser-
this point, the reader should reread the opening vation reports, there is another logical dimension
section of this chapter.) in which the latter rest on the former.
The idea that observation "strictly and prop- Above all, the picture is misleading because of
erly so-called" is constituted by certain self- its static character. One seems forced to choose
authenticating nonverbal episodes, the authority between the picture of an elephant which rests on
of which is transmitted to verbal and quasi-verbal a tortoise (What supports the tortoise?) and the
performances when these performances are made picture of a great Hegelian serpent of knowledge
"in conformity with the seman tical rules of the with its tail in its mouth (Where does it begin?).
language;' is, of course, the heart of the Myth of Neither will do. For empirical knowledge, like its
the Given. For the given, in epistemological tradi- sophisticated extension, science, is rational, not
tion, is what is taken by these self-authenticating because it has a foundation but because it is a self-
episodes. These "takings" are, so to speak, the correcting enterprise which can put any claim in
unmoved movers of empirical knowledge, the jeopardy, though not all at once.

Epistemic Principles

Wilfrid Sellars

I as contrasted with its contradictory. But how

good? Adequate? Conclusive? If adequate, ade-
The explication of knowledge as "justified true quate for what? If conclusive, the conclusion of
belief", though it involves many pitfalls to which what is at stake?
attention has been called in recent years, remains We are all familiar with Austin's point con-
the orthodox or classical account and is, I believe, cerning the performative character of "I know".
essentially sound. Thus, in the present lecture I We are also familiar with the fact that, whereas
shall assume that it can be formulated in such a to say "I promise to do A" is, other things being
way as to be immune from the type of counter- equal, to promise to do A, to say "I know that-p"
examples with which it has been bombarded since is not, other things being equal, to know that-po
Gettier's pioneering paper in Analysis and turn Chisholm's distinction between the strict and
my attention to another problem which has the extended sense of "performative utterance"
dogged its footsteps since the very beginning. is helpful in this connection. According to
This problem can be put in the form of two ques- Chisholm,
tions: Ifknowledge is justified true belief, how can
there be such a thing as self-evident knowledge? An utterance beginning with "I want" is not
And if there is no such thing as self-evident performative in [the1strict sense, for it cannot be
knowledge, how can any true belief be, in the said to be an "act" of wanting. But "I want" is
relevant sense, justified? often used to accomplish what one might accom-
But first let us beat about in the neighboring plish by means of the strict performative "I
fields, perhaps to scare up some game, but, in any request". Let us say, then, that "I want" may be a
case, to refamiliarize ourselves with the terrain. "performative utterance" in an extended sense of
the latter expression. I
Thus, are there not occasions on which a person
can be said to be justified in believing something
which he would not appropriately be said to He asks in which, if either, of these senses an
know? Presumably, to be justified in believing utterance of "I know" may be performative. After
something is to have good reasons for believing it, reminding us that "I know" is not performative in
the strict sense of the term, he allows that "[it] is
often used to accomplish what one may accomplish
Originally published in H. Castaneda (ed.), Action, by the strict performative 'I guarantee' or 'I give you
Knowledge, and Reality (Indianapolis: Bobbs-Merrill, my word' " and "hence may be performative in an
1975), pp. 332-49. extended sense of the term':2

He argues, however, that "I know" is not always II

a substitute for "I guarantee", pointing out that:
Yet even after justice has been done, perhaps along
Just as an utterance of "I want" may serve both to the above lines, to the performative element in
say something about me and to get you to do the meaning of the verb "to know", it seems to
something, an utterance of "I know" may serve me that we must recognize a closely related use of
both to say something about me and to provide this expression which, though it may have impli-
you with guarantees. To suppose that the per- cations concerning action, is not in any of the
formance of the nondescriptive function is incon- above senses performative. For once the ethical
sistent with the simultaneous performance of the issue of how good one's reasons for a belief must
descriptive function might be called, therefore, be in order to justify giving a guarantee is solved,
an example of the performative fallacy. 3 there remains the problem of how good reasons
must be to justify believing that-p, where to
I think that Chisholm is quite right about this. believe that-p is obviously not an action, let alone
On the other hand, it seems to me that he over- a performatory action in either the strict or the
looks the possibility of a connection between extended sense.
"I know" and "I guarantee" other than the one he Confronted by this question, we are tempted
considers. "I know that-p" might be related to to set apart a class of cases in which the reasons
"I guarantee that-p" not just as an autobiograph- are not only good enough to justify believing
ical description which on occasion performs the that-p but good enough to make it absurd not to
same role as the latter but as one which contains believe that-p (or, perhaps, to believe its contra-
a reference to guaranteeing in its very meaning. dictory). It is perhaps, some such concept as this
Is it not possible to construe "I know that-p" as which is (in addition to the truth condition) the
essentially equivalent to "p, and I have reasons non-performative core of the meaning of the verb
good enough to support a guarantee" (i.e., to say "to know':
"I guarantee" or "You can rely on my state- I think the above discussion has served its
ment")? Such an account would enable us to primary purpose by highlighting the concept of
recognize a performative element in the very having good reasons for believing that-po For the
meaning of the verb "to know" without constru- solution of the problem which was posed in my
ing "I know" as a performative in the strict sense. opening remarks hinges ultimately on a distinc-
It would also preserve the symmetry between tion between two ways in which there can be,
first person and other person uses of the verb "to and one can have, good reasons for believing
know" which seems to be a pre-analytic datum. that-p.5
Thus, "He knows that-p" would entail "He has Now one pattern for justifying a belief in terms
reasons good enough to support a guarantee of good reasons can be called inferential. Consider
that-p".4 the schema:
Furthermore, this account would enable us to
appreciate the context dependence of the adequacy p;
involved. Reasons which might be adequately So, I have good reasons, all things considered,
good to justify a guarantee on one occasion might for believing q.
not be adequate to justify a guarantee on another.
Again, the presence of such a performative ele- On reflection, this schema tends to expand into:
ment in the very meaning of the verb "to know"
would account for the fact (if it is a fact) that we I have good reasons, all things considered, for
rarely think in terms of "I know" in purely self- believing p;
directed thinkings; that we rarely have thoughts So,p;
of the form "I know that-p" unless the question of So, I have good reasons, all things considered,
a possible guarantee to someone other than our- for believing q.
selves has arisen. Of course, we can "tell ourselves"
that we know something, but, then, so can we be Further reflection suggests that arguments
said to make promises to ourselves. conforming to this schema have a suppressed

premise. What might it be? Consider the follow- where "E" formulates the evidence, to a conclu-
ing expanded schema: sion of the form:

I have, all things considered, good reasons for It is reasonable, all things considered, to
believing p; believe H,
p logically implies q; where "H" formulates in the first case a law-like
So, I have, all things considered, good reasons statement and in the second case a body of theo-
for believing q. retical assumptions.

The line of thought thus schematically repre-

sented would seem to involve the principle, III
Logical implication transmits reasonableness. As has been pointed out since time immemorial
it is most implausible to suppose that all epis~
In cases of this type, we are tempted to say, we
temic justification is inferential, at least in the
have derivative good reasons, all things considered,
sense of conforming to the patterns described
for believing q. We say, in other words, that the
above. Surely, it has been argued, there must be
reasonableness of believing q is "inferential':
beliefs which we are justified in holding on
Notice that the above line of thought is obvi-
grounds other than that they can be correctly
ously an oversimplification, undoubtedly in
inferred, inductively or deductively, from other
several respects. In particular, it is important to
beliefs which we are justified in holding. In tradi-
note that if I have independent grounds for
tional terms, if there is to be inferential knowledge,
believing not-q, I may decide that I do not have
must there not be non-inferential knowledge -
good reasons, all things considered, for believing
beliefs, that is, the reasonableness of which does
that-po After all, if p implies q, not-q equally
not rest on the reasonableness of beliefs which
implies not-po Yet in spite of its oversimplifica-
logically or probabilistically imply them?
tions, the above train of thought takes us nearer
We are clearly in the neighborhood of what has
to the distinctions necessary to solve our problem.
been called the "self-evident", the "self-certifying",
I have been considering the case where one
in short, of "intuitive knowledge': It is in this
proposition, p, logically implies another, q, and
neighborhood that we find what has come to be
have claimed, with the above qualifications, that
called the foundational picture of human knowl-
logical implication transmits reasonableness.
edge. According to this picture, beliefs which have
Perhaps we can also take into account, with trepi-
inferential reasonableness ultimately rely for their
dation, "probabilistic" implication, which would
authority on a stratum of beliefs which are, in
give us the following schema:
some sense, self-certifying. The reasonableness of
It is reasonable, all things considered, to moves from the level of the self-evident to higher
believe p; levels would involve the principles oflogic (deduc-
So,p; tive and inductive) and, perhaps, certain addi-
p probabilistically implies q to a high degree; tional principles which are sui generis. They would
So, all things considered, it is reasonable to have in common the character of transmitting
believe q. authoritativeness from lower-level beliefs to
higher-level beliefs.
Probabilistic justification of beliefs in accordance
with this pattern would, presumably, be illus-
trated by inductive arguments and theoretical IV
explanations. In each case, we move from a
premise of the form: Let us reflect on the concept of such a founda-
tionallevel of knowledge. It involves the concept
It is reasonable, all things considered, to of beliefs which are reasonable, which have epis-
believe E, ternic authority or correctness, but which are not

reasonable or authoritative by virtue of the fact It is reasonable to believe it to be a fact that

that they are beliefs in propositions which are a isF;
implied by other propositions which it is reason- So, it is reasonable to believe that a is F,
able to believe. Let us label them, for the moment,
"non-inferentially reasonable beliefs': which, in virtue of the equivalence of
How can there be such beliefs? For the con-
cept of a reason seems so clearly tied to that of an believing a to be F
inference or argument that the concept of non-
inferential reasonableness seems to be a contra- with
dictio in adjecto. Surely, we are inclined to say, for
a belief (or believing) to be reasonable, there must believing it to be a fact that a is F,
be a reason for the belief (or believing). And must
not this reason be something other than the belief is obviously unilluminating.
or believing for which it is the reason? And surely,
we are inclined to say, to believe something because
it is reasonable (to believe it) involves not only that v
there be a reason but that, in a relevant sense, one
has or is in possession of the reason. Notice that I Now many philosophers who have endorsed a
have deliberately formulated these expostulations concept of intuitive knowledge are clearly com-
in such a way as to highlight the ambiguities mitted to the position that there is a level of cogni-
involved when one speaks of reasonable beliefs. tion more basic than believing. This more basic
In attempting to cope with these challenges, level would consist of a sub-conceptual' aware-
I shall leave aside problems pertaining to inferen- ness of certain facts. In terms of the framework
tial and non-inferential reasonableness in logic that I have sketched elsewhere, there would be a
and mathematics and concentrate on the appar- level of cognition more basic than thinkings or
ent need for "self evidence" in the sphere of tokenings of sentences in Mentalese - more basic,
empirical matters of fact. in fact, than symbolic activity, literal or analogi-
How might a self-justifying belief be con- cal. It would be a level of cognition unmediated
strued? One suggestion, modified from Chisholm's by concepts; indeed it would be the very source of
Theory of Knowledge,6 is to the effect that the concepts in some such way as described by tradi-
justification of such beliefs has the form, tional theories of abstraction. It would be "direct
apprehension" of facts; their "direct presence" to
What justifies me in claiming that my belief the mind. 8
that a is F is reasonable is simply the fact that Schematically we would have,
a isF.
It is a fact (which I directly apprehend) that a
But this seems to point to the existence of infer- is F;
ences of the form, So, it is reasonable to have the conceptual belief
that a is F.
It is a fact that a is F;
So, it is reasonable to believe that a is F, This multiplication of distinctions raises two seri-
ous problems: (1) What sort of entities are facts?
and one might begin to wonder what principle Do they belong to the real (extra-conceptual)
authorizes this inference. order? That "fact" is roughly a synonym for
Something, clearly, has gone wrong. In order "truth", and "true" is appropriately predicated of
for any such argument to do the job, its premise conceptual items (in overt speech or Mentalese)
would have to have authority; it would have to be should give pause for thought.
something which it is reasonable to believe. But if Then there is also the question: (2) How is
we modify the schema to take this into account, it "direct apprehension" to be understood? If
becomes: the apprehending is distinguishable from the

apprehended, is it not also "separable"? Might not The thought occurred to Jones that snow is
apprehending occur without any fact being appre- white
hended? If so, an "apprehending that-p" might
not be an apprehending of the fact that-po Hitting, is
in baseball, implies that something is hit.
"Swinging" does not. To hit is to swing success- Jones said "snow is white",
fully. Of course, "apprehend", like "see", is, in its
ordinary sense, an achievement word. But does where the verb "to say" was stripped of some of
this not mean that, as in the case of "see", there is its ordinary implications and roughly equated
a place for "ostensibly apprehending", i.e., seeming with "to utter words candidly as one who knows
to apprehend, a concept which does not imply the language". In particular, it was purged of the
achievement? illocutionary and perlocutionary forces which
Many who use the metaphor "to see" in intel- Austin and Grice find so central to their theory of
lectual contexts overlook the fact that in its literal meaning. "To say", in this sense, was also equated
sense "seeing" is a term for a successful conceptual with "thinking-out-loud".
activity which contrasts with "seeming to see". No According to the VB, as I describe him, we
piling on of additional metaphors (e.g., "grasp- must also introduce, in order to take account of
ing", which implies an object grasped) can blunt those cases where one thinks silently, a secondary
this fact. Now the distinction between seeing and sense of
merely seeming to see implies a criterion. To rely
on the metaphors of "apprehending" or "presence The thought occurred to Jones that snow IS
of the object" is to obscure the need of criteria for white,
distinguishing between "knowing" and "seeming
to know", which ultimately define what it means in which it refers to a short-term proximate propen-
to speak of knowledge as a correct or well-founded sityto think-out-loud that snow is white. When this
thinking that something is the case. propensity is "uninhibited'; one thinks-aut-loud,
If so, to know that we have apprehended a fact, i.e., thinks in the primary sense of this term (as con-
we would have to know that the criteria which strued by VB). There can be many reasons why, on a
distinguish apprehending from seeming to appre- particular occasion, this propensity is inhibited. But,
hend were satisfied. In short, I suspect that the for our purposes, the most important is the general
notion of a non-conceptual "direct apprehension" inhibition acquired in childhood when, after being
of a "fact" provides a merely verbal solution to taught to think-out-loud, one is trained not to be a
our problem. The regress is stopped by an ad hoc "babbler': One might use the model of an on-off
regress-stopper. Indeed, the very metaphors switch which gets into the wiring diagram when the
which promised the sought-for foundation child learns to keep his thoughts to himself.
contain within themselves a dialectical moment I have argued elsewhere that yet another con-
which takes us beyond them. cept of "having the thought occur to one that-p"
can be introduced which stands to the second as
the theoretical concept of electronic processes
VI stands to the acquisition (and loss) of the power
to attract iron filings (or a bell clapper) by a
What is the alternative? I suggest that the key to piece of soft iron in a coil of wire attached to an
our problem is provided by the Verbal Behaviorist electric circuit. I argued that the classical con-
model, developed elsewhere. It is, we have seen, a cept of thought-episodes can be construed as
simple, indeed radically over-simplified, model, part of a theoretical framework designed to
but it will provide us, I believe, with the outline of explain the acquisition and loss of verbal pro-
a strategy for getting out of the classical laby- pensities to think-out-loud. In approaching the
rinth. problem of the status of non-inferential knowl-
According to this model, it will be remem- edge, however, I shall return to the VB model
bered, the primary sense of and concentrate, indeed, on the primary sense of

having the thought occur to one that-p, i.e., I just thought-out-loud "Lo! Here is a red
think-out-loud that-po apple" (no countervailing conditions obtain);
I have argued elsewhere that perceptual So, there is good reason to believe that there is
experience involves a sensory element which is a red apple in front of me.
in no way a form of thinking, however inti-
mately it may be connected with thinking. This Of course, the conclusion of this reasoning is
element consists of what I have variously called not the thinking involved in his original percep-
"sense impressions", "sensations", or "sensa". tual experience. Like all justification arguments, it
I argued that these items, properly construed, is a higher-order thinking. He did not originally
belong in a theoretical framework designed to infer that there is a red apple in front of him. Now,
explain: however, he is inferring from the character and
context of his experience that it is veridical and
(a) the difference between merely thinking of that there is good reason to believe that there is
(believing in the existence of) a perceptible indeed a red apple in front of him.
state of affairs and seeing (or seeming to see) Notice that although the justification of the
that such a state of affairs exists; belief that there is a red apple in front of (Jones)
(b) how it can seem to a person that there is a is an inferential justification, it has the peculiar
pink ice cube in front of him when there character that its essential premise asserts the
isn't one - either because there is something occurrence of the very same belief in a specific
there which is either not pink or not cubical, context.!O It is this fact which gives the appearance
or because there is nothing there and he is that such beliefs are self-justifying and hence gives
having a realistic hallucination. the justification the appearance of being non-
I've explored problems pertaining to the It is, as I see it, precisely this feature of the
nature and status of this sensory element on many unique pattern of justification in question which,
occasions,9 but further exploration of this theme misinterpreted, leads Chisholm to formulate as
would leave no time for the problem at hand. his principle for the "directly evident",
What is important for our purposes is that
perceptual experience also involves a conceptual What justifies me in counting it as evident that
or propositional component - a "thinking" in a a is F is simply the fact that a is F.ll
suitably broad sense of this accordion term.
In perception, the thought is caused to occur to To be sure, Chisholm's examples of the "directly
one that, for example, there is a pink ice cube in evident" are not taken from the domain of percep-
front of one. It is misleading to call such a tual beliefs, but rather, in true Cartesian spirit,
thought a "perceptual judgment" - for this from one's knowledge about what is going on in
implies question-answering activity of estimat- one's mind at the present moment. Indeed, he
ing, for example, the size of an object. (I judge rejects the idea that particular perceptual beliefs
that the room is ten feet tall.) Perhaps the best of the kind which I illustrated by my example of
term is "taking something to be the case". Thus, the red apple are ever directly evident.
on the occasion of sensing a certain color con- On the other hand, though he does think
figuration, one takes there to be an object or sit- that particular perceptual beliefs of this type
uation of a certain description in one's physical can at best be indirectly evident, he does think
environment. that they can be reasonable. Should we say
Let us consider the case where "directly reasonable"? I, of course, would answer
in the affirmative. Yet it is not clear to me that
Jones sees there to be a red apple III front Chisholm would be happy with this suggestion.
of him. If (as he should) he has at the back of his mind
the reasoning;
Given that Jones has learned how to use the rele-
vant words in perceptual situations, he is justified There (visually) appears to me to be a red
in reasoning as follows: apple here;

So, it is reasonable for me (to believe) that The contrasting reasoning would be:
there is a red apple here,
There appears to me to be a pink object here;
then he should not object to speaking of the rea- So, it is evident to me that I perceive a pink
sonableness in question as "direct'; for the premise object to be here and evident to me that there
does not contain a predicate of epistemic evalua- is a pink object here.
tion. If, on the other hand (as he should not), he has
at the back of his mind the following reasoning, Now I suspect that what has misled Chisholm
is the fact that if I were to argue,
It is evident to me that there (visually) appears
to me to be a red apple here; There appears to me to be a pink cube here;
So, it is reasonable for me (to believe) that So, it is highly reasonable for me (to believe)
there is a red apple here, that there is a pink object here,

we could expect him to object to speaking of his a skeptic could be expected to challenge me by
reasonableness as "direct". asking "What right have you to accept your con-
This tension sets the stage for a corresponding clusion, unless you have a right to accept the
comment on Chisholm's third epistemic princi- premise? Are you not implying that you know that
ple, which concerns the case where what we visu- there appears to you to be a pink object here; and
ally take to be the case is the presence of something must not this claim be a tacit premise in your
having a "sensible characteristic F" (where "F' argument?" But, surely, the skeptic would just be
ranges over the familiar Aristotelian list of proper mistaken - not, indeed, in asserting that in some
and common sensibles). The principle reads as sense I imply that I know that there appears to me
follows: to be a pink object here, but in asserting that this
implication must be taken to be a premise in my
(C) If there is a certain sensible characteristic reasoning, if it is to be valid, and, hence, if the
F such that S believes that he perceives corresponding epistemic principle is to be true.
something to be F, then it is evident to But in that case, the latter principle would be not
S that he is perceiving something to have Chisholm's (C), but rather:
that characteristic F, and also evident that
there is something that is F. (C') If it is evident to S that there is a certain
sensible characteristic F ...
I shall not pause to quibble over such matters
as whether, in the light of Chisholm's definition The larger import of the above reply to the
of "evident", it can ever be evident to me that I am skeptic will be sketched in my concluding remarks.
perceiving something to be pink or that some- For the moment, let me say that from my point of
thing in front of me is pink - even if the claim is view something very like Chisholm's principle
limited to the facing side. A high degree of rea- (C) is sound but concerns the direct evidence (or,
sonableness will do. The point which I wish to better, direct high degree of reasonableness) of
stress is that once again the question arises, does certain perceptual beliefs. Let me formulate it as
Chisholm think of the evidence involved in the follows:
principles as "direct" or "indirect"? This time it is
clear that he thinks of it as indirect. As I see it, (5) If there is a certain sensible characteristic
then, he has at the back of his mind the following F such that S believes that he perceives
reasoning: something to be F, then it is evident to S
that there is something that is F and,
It is evident to me that there appears to me to hence, that he is perceiving something to
be a pink object here; beF.
So, it is evident to me that I perceive a pink
object to be here and evident to me that there Notice that I have reversed the relative posi-
is a pink object here. tion of the two clauses in the consequent as they

appear in Chisholm's principle. This is because, is at stake. Must we not conclude that any such
on my interpretation, the core of the principle is account as I give of the principle that perceptual
beliefs occurring in perceptual contexts are likely
(S1) If I ostensibly see there to be an F object to be true is circular? It must, indeed, be granted
here, then it is highly reasonable for me that principles pertaining to the epistemic
(to believe) that there is an F object here. authority of perceptual and memory beliefs are
not the sort of thing which could be arrived at by
And the move to inductive reasoning from perceptual belief. But
the best way to make this point is positive. We
(S2) If I ostensibly see there to be an F object have to be in this framework to be thinking and
here, then it is highly reasonable for me perceiving beings at all. I suspect that it is this
(to believe) that I see there to be an F plain truth which is the real underpinning of the
object here idea that the authority of epistemic principles
rests on the fact that unless they were true we
IS justified by the conceptual tie between could not see that a cat is on the roof.
"ostensibly see", "see", and truth. I pointed out a moment ago that we have to
be in the framework of these (and other) princi-
ples to be thinking, perceiving, and, I now add,
VII acting beings at all. But surely this makes it clear
that the exploration of these principles is but
Chisholm's principle (C) and his other epistemic part and parcel of the task of explicating the
principles pertaining to perception and memory concept of a rational animal or, in VB terms, of a
are themselves justified, as he sees it, by the fact language-using organism whose language is
that unless they, or something like them, are true, about the world in which it is used. It is only in
then there could be no such thing as perceptual the light of this larger task that the problem of
knowledge to the effect, to use his example, that the status of epistemic principles reveals its true
there is a cat on the roof. We have here a justifica- meaning.
tion of the "this or nothing" kind familiar to the From the perspective of this larger task, the
Kantian tradition. The principles also seem, on metaphor of "foundation and superstructure" is
occasion, to be treated as candidates for the status seen to be a false extrapolation, to use a Deweyan
of synthetic a priori (and even, one suspects, self- turn of phrase, from specific "problematic situa-
evident) truth. tions" with respect to which it is appropriate. And
As I see it, on the other hand, these epistemic when we concern ourselves, as Philosophy ulti-
principles can be placed in a naturalistic setting mately demands, with how it is with man and his
and their authority construed in terms of the world, as contrasted with the catch-as-catch-can
nature of concept formation and of the acquisi- procedures which generate man's awareness of
tion of relevant linguistic skills. The model which himself and his world, surely we can say, as I wrote
I have been using is, indeed, a very simple one, some fifteen years ago in an earlier essay on this
and I have largely limited my use of it to the epis- topic,
temic authority of perceptual beliefs. But if the
strategy which I have suggested is successful, it is
There is clearly some point to the picture of
a relatively simple matter to extend it to memory
human knowledge as resting on a level of propo-
beliefs. I have discussed the case of non-inferen- sitions - observation reports - which do not rest
tial knowledge of our own mental states in some on other propositions in the same way as other
detail, using this same general strategy, on a propositions rest on them. On the other hand,
number of occasions. 12 I do wish to insist that the metaphor of "founda-
But, surely, it will be urged, facts about learn- tion" is misleading in that it keeps us from seeing
ing languages and acquiring linguistic skills are that if there is a logical dimension in which other
themselves empirical facts; and to know these empirical propositions rest on observation
facts involves perception, memory, indeed, all reports, there is another logical dimension in
the epistemic activities the justification of which which the latter rest on the former.

Above all, the picture is misleading because of begin?). Neither will do. For empirical knowl-
its static character. One seems forced to choose edge, like its sophisticated extension, science, is
between the picture of an elephant which rests rational, not because it has a foundation but
on a tortoise (What supports the tortoise?) and because it is a self-correcting enterprise which
the picture of a great Hegelian serpent of know1- can put any claim in jeopardy, though not all at
edge with its tail in its mouth (Where did it once. 13


R. M. Chisholm, Theory of Knowledge That it is implicit in Chisholm's posltIon

(Englewood Cliffs, NJ: Prentice-Hall, 1966), becomes clear not only when we reflect (as
pp.16-17. above) on what his principle concerning the
2 Ibid. directly evident might mean, but when we
3 Ibid., p. 17. take into account his use of such phrases as
4 Notice that the above account of the relation "state of affairs" that '''presents itself to him'"
of "I know" to a performative is not quite the or that '''is apprehended through itself'"
same as Urmson's. According to the latter, as (Chisholm, Theory of Knowledge, p. 28) and
represented by Chisholm, to say that Mr Jones his general commitment to a fact ontology
knew some proposition to be true is to say (ibid., chap. 7, passim), a "fact", in the rele-
that Mr Jones was "in a position in which he vant sense, being a "state of affairs which
was entitled to say 'I know": This account, as exists" (ibid., p. 104). "Exists" in this context
Chisholm points out, brings us back to the should not be confused with the "existential
original problem of how the first person use quantifier" but should be considered as a
of the verb is to be construed. synonym for "obtains': It is obviously not
5 I have called attention elsewhere to the self-contradictory to say that some states of
importance of distinguishing between ques- affairs do not obtain.
tions concerning the reasonableness of 9 Most recently in my Science and Metaphysics
believing that-p from questions concerning (London: Routledge and Kegan Paul, 1967),
the reasonableness of "acting on the proposi- ch. 1, and in "Science, Sense Impressions,
tion that-p", including guaranteeing that-po and Sensa: A Reply to Cornman", Review of
The concept of acting on a proposition is Metaphysics 25 (1971), which is a reply to
clear only in simple cases, as when, for exam- Cornman's "Sellars, Scientific Realism, and
ple, the proposition occurs as a premise in the Sensa", Review of Metaphysics 24 (1970).
agent's practical reasoning. When the agent 10 I called attention to this feature of the
takes probabilities into account, a far more justification involved in "non-inferential"
complicated story is necessary to clarify the knowledge in Science, Perception and Reality
sense in which a person can be said to have (London: Routledge and Kegan Paul, and
acted on a given proposition. For a discussion New York: Humanities Press, 1963), chap. 3.
of these problems, see my "Induction as Thus, I wrote" ... one only knows what one
Vindication", Philosophy of Science 31 (1964), has a right to think to be the case. Thus, to
pp. 197-232. say that one directly knows that-p is to say
6 Chisholm, Theory of Knowledge, p. 28. that his right to the conviction that-p essen-
Chisholm's principle concerns "what justifies tially involves the fact that the idea that-p
us in counting it as evident that a is F': But the occurred to the knower in a specific way"
"evident" is defined on p. 22 as a special case (ibid., p. 88). I suggested that this "kind of
of the "reasonable". credibility" be called "trans-level credibility",
7 Where "sub-conceptual" is far from being and the pattern of inference involved in the
used as a pejorative term. reasoning which mobilizes this credibility,
8 It is clearly some such position which is envis- "trans-level inference': A similar point was
aged by many who explicitly reject the equa- less clearly made in Sections 32-9 of my
tion of knowledge with justified true belief. "Empiricism and the Philosophy of Mind",

in Herbert Feigl and Michael Scriven (eds), 12 Most recently in my Science and Metaphysics,
Minnesota Studies in the Philosophy of Science esp. pp. 71 ff., 151 ff.
vol. I (Minneapolis: University of Minnesota 13 "Empiricism and the Philosophy of Mind':
Press, 1956). Reprinted as chapter 5 of my sec. 38; quoted from Science, Perception and
Science, Perception and Reality. Reality, p. 170.
11 Chisholm, Theory of Knowledge, p. 28.

Can Empirical Knowledge

Have a Foundation?

Laurence BonJour

The idea that empirical knowledge has, and must the main dialectical variants of foundationism, by
have, a foundation has been a common tenet of viewing them as responses to one fundamental
most major epistemologists, both past and problem which is both the main motivation and the
present. There have been, as we shall see further primary obstacle for foundationism; and second, as
below, many importantly different variants of a result of this discussion to offer schematic reasons
this idea. But the common denominator among for doubting whether any version of foundationism
them, the central thesis of epistemological foun- is finally acceptable.
dationism as I shall understand it here, is the The main reason for the impressive durability
claim that certain empirical beliefs possess a of foundationism is not any overwhelming plau-
degree of epistemic justification or warrant which sibility attaching to the main foundationist thesis
does not depend, inferentially or otherwise, on in itself, but rather the existence of one appar-
the justification of other empirical beliefs, but is ently decisive argument which seems to rule out
instead somehow immediate or intrinsic. It is all non-skeptical alternatives to foundation ism,
these non-inferentially justified beliefs, the thereby showing that some version of founda-
unmoved (or self-moved) movers of the epis- tionism must be true (on the assumption that
temic realm as Chisholm has called them, 1 that skepticism is false). In a recent statement by
constitute the foundation upon which the rest of Quinton, this argument runs as follows:
empirical knowledge is alleged to rest.
If any beliefs are to be justified at all, ... there
In recent years, the most familiar foundationist
must be some terminal beliefs that do not owe
views have been subjected to severe and continuous
their ... credibility to others. For a belief to be
attack. But this attack has rarely been aimed directly
justified it is not enough for it to be accepted, let
at the central foundationist thesis itself, and new alone merely entertained: there must also be
versions of foundationism have been quick to good reason for accepting it. Furthermore, for an
emerge, often propounded by the erstwhile critics inferential belief to be justified the beliefs that
themselves. Thus foundationism has become a support it must be justified themselves. There
philosophical hydra, difficult to come to grips with must, therefore, be a kind of belief that does not
and seemingly impossible to kill. The purposes owe its justification to the support provided by
of this paper are, first, to distinguish and clarify others. Unless this were so no belief would be
justified at all, for to justify any belief would
require the antecedent justification of an infinite
Originally published in American Philosophical series of beliefs. The terminal ... beliefs that are
Quarterly 15, 1(1978), pp. 1-13. needed to bring the regress of justification to a

stop need not be strictly self-evident in the sense Two further points about inferential justifica-
that they somehow justify themselves. All that tion, as understood here, must be briefly noted.
is required is that they should not owe their First, the belief in question need not have been
justification to any other beliefs. 2 arrived at as the result of an inference in order to
be inferentially justified. This is obvious, since a
I shall call this argument the epistemic regress belief arrived at in some other way (e.g., as a result
argument, and the problem which generates it, the of wishful thinking) may later come to be main-
epistemic regress problem. Since it is this argument tained solely because it is now seen to be inferen-
which provides the primary rationale and argu- tially justifiable. Second, less obviously, a person
mentative support for foundationism, a careful for whom a belief is inferentially justified need
examination of it will also constitute an explora- not have explicitly rehearsed the justificatory
tion of the foundationist position itself. The main argument in question to others or even to him-
dialectical variants of foundation ism can best be self. It is enough that the inference be available to
understood as differing attempts to solve the him if the belief is called into question by others
regress problem, and the most basic objection to or by himself (where such availability may itself
the foundationist approach is that it is doubtful be less than fully explicit) and that the availability
that any of these attempts can succeed. (In this of the inference be, in the final analysis, his reason
paper, I shall be concerned with the epistemic for holding the belief.6 It seems clear that many
regress argument and the epistemic regress prob- beliefs which are quite sufficiently justified to
lem only as they apply to empirical knowledge. It satisfy the justification criterion for knowledge
is obvious that an analogous problem arises also depend for their justification on inferences which
for a priori knowledge, but there it seems likely have not been explicitly formulated and indeed
that the argument would take a different course. which could not be explicitly formulated without
In particular, a foundationist approach might be considerable reflective effort (e.g., my current
inescapable in an account of a priori knowledge.) belief that this is the same piece of paper upon
which I was typing yesterday).7
Suppose then that belief A is (putatively)
I justified via inference, thus raising the question
of how the justifying premise-belief B is justified.
The epistemic regress problem arises directly out Here again the answer may. be in inferential
of the traditional conception of knowledge as terms: B may be (putatively) justified in virtue of
adequately justified true belief3 - whether this be being inferable from some further belief C. But
taken as a fully adequate definition of knowledge then the same question arises about the justifica-
or, in light of the apparent counter-examples tion of C, and so on, threatening an infinite and
discovered by Gettier,4 as merely a necessary but apparently vicious regress of epistemic justifica-
not sufficient condition. (I shall assume through- tion. Each belief is justified only if an epistemi-
out that the elements of the traditional concep- cally prior belief is justified, and that epistemically
tion are at least necessary for knowledge.) Now prior belief is justified only if a still prior belief is
the most natural way to justify a belief is by justified, etc., with the apparent result that justi-
producing a justificatory argument: belief A is fication can never get started - and hence that
justified by citing some other (perhaps conjunc- there is no justification and no knowledge. The
tive) belief B, from which A is inferable in some foundationist claim is that only through the
acceptable way and which is thus offered as a adoption of some version of foundationism can
reason for acceptingA.5 Call this inferential justi- this skeptical consequence be avoided.
fication. It is clear, as Quinton points out in the Prima facie, there seem to be only four basic
passage quoted above, that for A to be genuinely possibilities with regard to the eventual outcome
justified by virtue of such a justificatory argu- of this potential regress of epistemic justification:
ment, B must itself be justified in some fashion; (i) the regress might terminate with beliefs for
merely being inferable from an unsupported which no justification of any kind is available,
guess or hunch, e.g., would confer no genuine even though they were earlier offered as justifying
justification upon A. premises; (ii) the regress might proceed infinitely

backwards with ever more new premise beliefs in the circle seems now to presuppose its own
being introduted and then themselves requiring epistemically prior justification: such a belief
justification; (iii) the regress might circle back must, paradoxically, be justified before it can be
upon itself, so that at some point beliefs which justified. Advocates of views resembling alterna-
appeared earlier in the sequence of justifying tive (iii) have generally tended to respond to this
arguments are appealed to again as premises; sort of objection by adopting a holistic conception
(iv) the regress might terminate because beliefs of justification in which the justification of indi-
are reached which are justified - unlike those in vidual beliefs is subordinated to that of the closed
alternative (i) - but whose justification does not systems of beliefs which such a view implies; the
depend inferentially on other empirical beliefs property of such systems usually appealed to as a
and thus does not raise any further issue of justi- basis for justification is internal coherence. Such
fication with respect to such beliefs.s The founda- coherence theories attempt to evade the regress
tionist opts for the last alternative. His argument problem by abandoning the view of justification as
is that the other three lead inexorably to the skep- essentially involving a linear order of dependence
tical result, and that the second and third have (though a non-linear view of justification has
additional fatal defects as well, so that some ver- never been worked out in detail). 10 Moreover, such
sion of the fourth, foundationist alternative must a coherence theory of empirical knowledge is
be correct (assuming that skepticism is false). subject to a number of other familiar and seem-
With respect to alternative (i), it seems apparent ingly decisive objectionsY Thus alternative (iii)
that the foundationist is correct. If this alternative seems unacceptable, leaving only alternative (iv),
were correct, empirical knowledge would rest ulti- the foundationist alternative, as apparently viable.
mately on beliefs which were, from an epistemic As thus formulated, the epistemic regress
standpoint at least, entirely arbitrary and hence argument makes an undeniably persuasive case
incapable of conferring any genuine justification. for foundationism. Like any argument by elimi-
What about the other two alternatives? nation, however, it cannot be conclusive until the
The argument that alternative (ii) leads to a surviving alternative has itself been carefully
skeptical outcome has in effect already been examined. The foundationist position may turn
sketched in the original formulation of the prob- out to be subject to equally serious objections, thus
lem. One who opted for this alternative could forcing a re-examination of the other alternatives,
hope to avoid skepticism only by claiming that a search for a further non-skeptical alternative, or
the regress, though infinite, is not vicious; but conceivably the reluctant acceptance of the skep-
there seems to be no plausible way to defend such tical conclusion. '2 In particular, it is not clear on
a claim. Moreover, a defense of an infinite regress the basis of the argument thus far whether and
view as an account of how empirical knowledge is how foundationism can itself solve the regress
actually justified - as opposed to how it might in problem; and thus the possibility exists that the
principle be justified - would have to involve the epistemic regress argument will prove to be a
seemingly dubious thesis that an ordinary knower two-edged sword, as lethal to the foundationist as
holds a literally infinite number of distinct beliefs. it is to his opponents.
Thus it is not surprising that no important phi-
losopher, with the rather uncertain exception of
Peirce,. seems to have advocated such a position. II
Alternative (iii), the view that justification
ultimately moves in a closed curve, has been The most straightforward interpretation of
historically more prominent, albeit often only as alternative (iv) leads directly to a view which I
a dialectical foil for foundationism. At first glance, will here call strong foundationism. According to
this alternative might seem even less attractive strong foundation ism, the foundational beliefs
than the second. Although the problem of the which terminate the regress of justification pos-
knower having to have an infinite number of sess sufficient epistemic warrant, independently
beliefs is no longer present, the regress itself, still of any appeal to inference from (or coherence
infinite, now seems undeniably vicious. For the with) other empirical beliefs, to satisfy the justifi-
justification of each of the beliefs which figure cation condition of knowledge and qualify as

acceptable justifying premises for further beliefs. beliefs. Thus this independent warrant must
Since the justification of these basic beliefs, as they somehow be argumented if knowledge is to be
have come to be called, is thus allegedly not achieved, and the usual appeal here is to coher-
dependent on that of any other empirical belief, ence with other such minimally warranted beliefs.
they are uniquely able to provide secure starting- By combining such beliefs into larger and larger
points for the justification of empirical knowl- coherent systems, it is held, their initial, minimal
edge and stopping-points for the regress of degree of warrant can gradually be enhanced
justification. until knowledge is finally achieved. Thus weak
The position just outlined is in fact a fairly foundationism, like the pure coherence theories
modest version of strong foundation ism. Strong mentioned above, abandons the linear concep-
foundationists have typically made considerably tion of justification. 14
stronger claims on behalf of basic beliefs. Basic Weak foundation ism thus represents a kind of
beliefs have been claimed not only to have suffi- hybrid between strong foundationism and the
cient non-inferential justification to qualify as coherence views discussed earlier, and it is often
knowledge, but also to be certain, infallible, indu- thought to embody the virtues of both and the
bitable, or incorrigible (terms which are usually vices of neither. Whether or not this is so in other
not very carefully distinguished)Y And most of respects, however, relative to the regress problem
the major attacks on foundationism have focused weak foundation ism is finally open to the very
on these stronger claims. Thus it is important to same basic objection as strong foundationism,
point out that nothing about the basic strong with essentially the same options available for
foundationist response to the regress problem meeting it. As we shall see, the key problem for
demands that basic beliefs be more than ade- any version of foundation ism is whether it can
quately justified. There might of course be other itself solve the regress problem which motivates
reasons for requiring that basic beliefs have some its very existence, without resorting to essentially
more exalted epistemic status or for thinking that ad hoc stipulation. The distinction between the
in fact they do. There might even be some sort of two main ways of meeting this challenge both
indirect argument to show that such a status is a cuts across and is more basic than that between
consequence of the sorts of epistemic properties strong and weak foundationism. This being so, it
which are directly required to solve the regress will suffice to concentrate here on strong founda-
problem. But until such an argument is given tionism, leaving the application of the discussion
(and it is doubtful that it can be), the question of to weak foundation ism largely implicit.
whether basic beliefs are or can be certain, infal- The fundamental concept of strong founda-
lible, etc., will remain a relatively unimportant tionism is obviously the concept of a basic belief.
side-issue. It is by appeal to this concept that the threat of an
Indeed, many recent foundationists have felt infinite regress is to be avoided and empirical
that even the relatively modest version of strong knowledge given a secure foundation. But how
foundation ism outlined above is still too strong. can there be any empirical beliefs which are thus
Their alternative, still within the general aegis of basic? In fact, though this has not always been
the foundationist position, is a view which may noticed, the very idea of an epistemically basic
be called weak foundationism. Weak foundation- empirical belief is extremely paradoxical. For on
ism accepts the central idea of foundationism - what basis is such a belief to be justified, once
viz. that certain empirical beliefs possess a degree appeal to further empirical beliefs is ruled out?
of independent epistemic justification or warrant Chisholm's theological analogy, cited earlier, is
which does not derive from inference or coher- most appropriate: a basic belief is in effect an
ence relations. But the weak foundationist holds epistemological unmoved (or self-moved) mo~er.
that these foundational beliefs have only a quite It is able to confer justification on other beliefs,
low degree of warrant, much lower than that but apparently has no need to have justification
attributed to them by even modest strong foun- conferred on it. But is such a status any easier to
dationism and insufficient by itself to satisfy the understand in epistemology than it is in theology?
justification condition for knowledge or to qualify How can a belief impart epistemic "motion" to
them as acceptable justifying premises for other other beliefs unless it is itself in "motion"? And,

even more paradoxically, how can a belief These general remarks about epistemic justi-
epistemically"move" itself? fication apply in full measure to any strong foun-
This intuitive difficulty with the concept of dationist position and to its constituent account
a basic empirical belief may be elaborated and of basic beliefs. If basic beliefs are to provide a
clarified by reflecting a bit on the concept of secure foundation for empirical knowledge, if
epistemic justification. The idea of justification is inference from them is to be the sole basis for the
a generic one, admitting in principle of many justification of other empirical beliefs, then that
specific varieties. Thus the acceptance of an feature, whatever it may be, in virtue of which a
empirical belief might be morally justified, i.e. belief qualifies as basic must also constitute a
justified as morally obligatory by reference to good reason for thinking that the belief is true.
moral principles and standards; or pragmatically If we let "rp" represent this feature, then for a
justified, i.e. justified by reference to the desirable belief B to qualify as basic in an acceptable foun-
practical consequences which will result from dationist account, the premises of the following
such acceptance; or religiously justified, i.e. justi- justificatory argument must themselves be at
fied by reference to specified religious texts or least justified: 17
theological dogmas; etc. But none of these other
varieties of justification can satisfy the justifica- (i) Belief B has feature rp.
tion condition for knowledge. Knowledge requires (ii) Beliefs having feature rp are highly likely
epistemic justification, and the distinguishing to be true.
characteristic of this particular species of justifi- Therefore, B is highly likely to be true.
cation is, I submit,. its essential or internal rela-
tionship to the cognitive goal of truth. Cognitive Notice further that while either premise taken
doings are epistemically justified, on this concep- separately might turn out to be justifiable on an a
tion, only if and to the extent that they are aimed priori basis (depending on the particular choice
at this goal - which means roughly that one of rp), it seems clear that they could not both be
accepts all and only beliefs which one has good thus justifiable. For B is ex hypothesi an empirical
reason to think are true. IS To accept a belief in the belief, and it is hard to see how a particular empir-
absence of such a reason, however appealing or ical belief could be justified on a purely a priori
even mandatory such acceptance might be from basis. 18 And if we now assume, reasonably enough,
other standpoints, is to neglect the pursuit of that for B to be justified for a particular person (at
truth; such acceptance is, one might say, epistemi- a particular time) it is necessary, not merely that a
cally irresponsible. My contention is that the idea justification for B exist in the abstract, but that
of being epistemically responsible is the core of the person in question be in cognitive possession
the concept of epistemic justification. 16 of that justification, we get the result that B is not
A corollary of this conception of epistemic jus- basic after all since its justification depends on
tification is that a satisfactory defense of a particu- that of at least one other empirical belief. If this is
lar standard of epistemic justification must consist correct, strong foundation ism is untenable as a
in showing it to be truth-conductive, i.e. in show- solution to the regress problem (and an analo-
ing that accepting beliefs in accordance with its gous argument will show weak foundationism to
dictates is likely to lead to truth (and more likely be similarly untenable).
than any proposed alternative). Without such a The foregoing argument is, no doubt, exceed-
meta-justification, a proposed standard of epis- ingly obvious. But how is the strong founda-
temic justification lacks any underlying rationale. tionist to answer it? Prima facie, there seem to
Why after all should an epistemically responsible be only two general sorts of answer which are
inquirer prefer justified beliefs to unjustified ones, even remotely plausible, so long as the strong
if not that the former are more likely to be true? foundationist remains within the confines of
To insist that a certain belief is epistemically justi- the traditional conception of knowledge, avoids
fied, while confessing in the same breath that this tacitly embracing skepticism, and does not
fact about it provides no good reason to think that attempt the heroic task of arguing that an
it is true, would be to render nugatory the whole empirical belief could be justified on a purely a
concept of epistemic justification. priori basis. First, he might argue that although it

is indeed necessary for a belief to be justified true such that, given Bap, it must be the case that p"
and a fortiori for it to be basic that a justifying (p. 166). A similar view seems to be implicit in
argument of the sort schematized above be in Dretske's account of perceptual knowledge in
principle available in the situation, it is not always Seeing and Knowing, with the variation that Dretske
necessary that the person for whom the belief is requires for knowledge not only that the relation in
basic (or anyone else) know or even justifiably question obtain, but also that the putative knower
believe that it is available; instead, in the case of believe that it obtains - though not that this belief
basic beliefs at least, it is sufficient that the be justified. 20 In addition, it seems likely that vari-
premises for an argument of that general sort (or ous views of an ordinary-language stripe which
for some favored particular variety of such argu- appeal to facts about how language is learned either
ment) merely be true, whether or not that person to justify basic belief or to support the claim that
(or anyone else) justifiably believes that they are no justification is required would, if pushed, turn
true. Second, he might grant that it is necessary out to be positions of this general sort. Here I shall
both that such justification exist and that the mainly confine myself to Armstrong, who is the
person for whom the belief is basic be in cognitive only one of these philosophers who is explicitly
possession of it, but insist that his cognitive grasp concerned with the regress problem.
of the premises required for that justification There is, however, some uncertainty as to how
does not involve further empirical beliefs which views of this sort in general and Armstrong's view
would then require justification, but instead in particular are properly to be interpreted. On
involves cognitive states of a more rudimentary the one hand, Armstrong might be taken as offer-
sort which do not themselves require justifica- ing an account of how basic beliefs (and perhaps
tion: intuitions or immediate apprehensions. I will others as well) satisfy the adequate-justification
consider each of these alternatives in turn. condition for knowledge; while on the other
hand, he might be taken as simply repudiating the
traditional conception of knowledge and the
III associated concept of epistemic justification, and
offering a surrogate conception in its place - one
The philosopher who has come the closest to an which better accords with the "naturalistic"
explicit advocacy of the view that basic beliefs world-view which Armstrong prefers. 21 But it is
may be justified even though the person for only when understood in the former way that
whom they are basic is not in any way in cognitive externalism (to adopt Armstrong's useful term) is
possession of the appropriate justifying argument of any immediate interest here, since it is only on
is D. M. Armstrong. In his recent book, Belief, that interpretation that it constitutes a version of
Truth and Knowledge,19 Armstrong presents a foundation ism and offers a direct response to the
version of the epistemic regress problem (though anti-foundationist argument set out above. Thus
one couched in terms of knowledge rather than I shall mainly focus on this interpretation of
justification) and defends what he calls an externalism, remarking only briefly at the end of
"Externalist" solution: the present section on the alternative one.
Understood in this way, the externalist solu-
According to "Externalist" accounts of non- tion to the regress problem is quite simple: the
inferential knowledge, what makes a true person who has a basic belief need not be in pos-
non-inferential belief a case of knowledge is some session of any justified reason for his belief and
natural relation which holds between the belief- indeed, except in Dretske's version, need not even
state ... and the situation which makes the belief think that there is such a reason; the status of his
true. It is a matter of a certain relation holding belief as constituting knowledge (if true) depends
between the believer and the world. (p. 157) solely on the external relation and not at all on his
subjective view of the situation. Thus there are no
Armstrong's own candidate for this "natural rela- further empirical beliefs in need of justification
tion" is "that there must be a law-like connection and no regress.
between the state of affairs Bap [i.e. a's believing Now it is clear that such an externalist position
that p] and the state of affairs that makes 'p' succeeds in avoiding the regress problem and the

anti-foundationist argument. What may well be cognitive thermometer, for such an external
doubted, however, is whether this avoidance observer (and in fact the example of a thermo-
deserves to be considered a solution, rather than meter is exactly the analogy which Armstrong
an essentially ad hoc evasion, of the problem. employs to illustrate the relationship which is
Plainly the sort of "external" relation which supposed to obtain between the person who has
Armstrong has in mind would, if known, provide the belief and the external state of affairs (p.
a basis for a justifying argument along the lines 166ff.)). But P himself has no reason at all for
sketched earlier, roughly as follows: thinking that B is likely to be true. From his per-
spective, it is an accident that the belief is trueY
(i) Belief B is an instance of kind K. And thus his acceptance of B is no more rational
(ii) Beliefs of kind K are connected in a law- or responsible from an epistemic standpoint than
like way with the sorts of states of affairs would be the acceptance of a subjectively similar
which would make them true, and there- belief for which the external relation in question
fore are highly likely to be true. failed to obtainY
Therefore, B is highly likely to be true. Nor does it seem to help matters to move
from Armstrong's version of externalism, which
But precisely what generates the regress problem requires only that the requisite relationship
in the first place is the requirement that for a between the believer and the world obtain, to the
belief B to be epistemically justified for a given superficially less radical version apparently held
person P, it is necessary, not just that there be jus- by Dretske, which requires that P also believe that
tifiable or even true premises available in the situ- the external relation obtains, but does not require
ation which could in principle provide a basis for that this latter belief be justified. This view may
a justification of B, but that P himself know or at seem slightly less implausible, since it at least
least justifiably believe some such set of premises requires that the person have some idea, albeit
and thus be in a position to employ the corre- unjustified, of why B is likely to be true. But this
sponding argument. The externalist position change is not enough to save externalism. One
seems to amount merely to waiving this general way to see this is to suppose that the person
requirement in cases where the justification takes believes the requisite relation to obtain on some
a certain form, and the question is why this should totally irrational and irrelevant basis, e.g. as a
be acceptable in these cases when it is not accept- result of reading tea leaves or studying astrologi-
able generally. (If it were acceptable generally, cal charts. If B were an ordinary, non-basic belief,
then it would seem that any true belief would be such a situation would surely preclude its being
justified for any person, and the distinction justified, and it is hard to see why the result
between knowledge and true belief would col- should be any different for an allegedly basic
lapse.) Such a move seems rather analogous to belief.
solving a regress of causes by simply stipulating Thus it finally seems possible to make sense of
that although most events must have a cause, externalism only by construing the externalist as
events of a certain kind need not. simply abandoning the traditional notion of epis-
Whatever plausibility attaches to externalism temic justification and along with it anything
seems to derive from the fact that if the external resembling the traditional conception of know-
relation in question genuinely obtains, then P ledge. (As already remarked, this may be precisely
will not go wrong in accepting the belief, and it is, what the proponents of externalism intend to be
in a sense, not an accident that this is so. But it doing, though most of them are not very clear on
remains unclear how these facts are supposed to this point.) Thus consider Armstrong's final sum-
justify P's acceptance of B. It is clear, of course, mation of his conception of knowledge:
that an external observer who knew both that P
accepted B and that there was a law-like connec- Knowledge of the truth ofparticular matters offact
tion between such acceptance and the truth of B is a belief which must be true, where the "must" is
would be in a position to construct an argument a matter of law-like necessity. Such knowledge
to justify his own acceptance of B. P could thus is a reliable representation or "mapping" of
serve as a useful epistemic instrument, a kind of reality. (p. 220)

Nothing is said here of reasons or justification or regress problem is avoided for basic beliefs by an
evidence or having the right to be sure. Indeed appeal directly to the non-cognitive world; the
the whole idea, central to the western epistemo- crucial difference is that for the givenist, unlike
logical tradition, of knowledge as essentially the the externalist, the justifying state of affairs in the
product of reflective, critical, and rational inquiry world is allegedly apprehended in some way by the
has seemingly vanished without a trace. It is pos- believer.
sible of course that such an altered conception of The givenist position to be considered here is
knowledge may be inescapable or even in some significantly weaker than more familiar versions
way desirable, but it constitutes a solution to the of the doctrine of givenness in at least two differ-
regress problem or any problem arising out of ent respects. In the first place, the present version
the traditional conception of knowledge only in does not claim that the given (or, better, the
the radical and relatively uninteresting sense that apprehension thereof) is certain or even incorri-
to reject that conception is also to reject the prob- gible. As discussed above, these stronger claims
lems arising out of it. In this paper, I shall confine are inessential to the strong foundationist solu-
myself to less radical solutions. tion to the regress problem. If they have any
importance at all in this context it is only because,
as we shall see, they might be thought to be
IV entailed by the only very obvious intuitive picture
of how the view is supposed to work. In the
The externalist solution just discussed represents second place, givenism as understood here does
a very recent approach to the justification of basic not involve the usual stipulation that only one's
beliefs. The second view to be considered is, in private mental and sensory states can be given.
contrast, so venerable that it deserves to be called There mayor may not be other reasons for think-
the standard foundationist solution to the prob- ing that this is in fact the case, but such a restric-
lem in question. I refer of course to the traditional tion is not part of the position itself. Thus both
doctrine of cognitive givenness, which has played positions like that of C. 1. Lewis, for whom the
a central role in epistemological discussions at given is restricted to private states apprehended
least since Descartes. In recent years, however, the with certainty, and positions like that of Quinton,
concept of the given, like foundationism itself, for whom ordinary physical states of affairs are
has come under serious attack. One upshot of the given with no claim of certainty or incorrigibility
resulting discussion has been a realization that being involved, will count as versions of givenism.
there are many different notions of given ness, As already noted, the idea of givenness has
related to each other in complicated ways, which been roundly criticized in recent philosophical
almost certainly do not stand or fall together. discussion and widely dismissed as a piece of
Thus it will be well to begin by formulating the philosophical mythology. But much at least of
precise notion of givenness which is relevant in this criticism has to do with the claim of certainty
the present context and distinguishing it from on behalf of the given or with the restriction to
some related conceptions. private, subjective states. And some of it at least
In the context of the epistemic regress prob- has been mainly concerned with issues in the phi-
lem, givenness amounts to the idea that basic losophy of mind which are only distantly related
beliefs are justified by reference, not to further to our present epistemological concerns. Thus
beliefs, but rather to states of affairs in the world even if the objections offered are cogent against
which are "immediately apprehended" or "directly other and stronger versions of givenness, it
presented" or "intuited:' This justification by ref- remains unclear whether and how they apply
erence to non-cognitive states of affairs thus alleg- to the more modest version at issue here. The
edly avoids the need for any further justification possibility suggests itself that modest givenness
and thereby stops the regress. In a way, the basic may not be a myth, even if more ambitious varie-
gambit of givenism (as I shall call positions of this ties are, a result which would give the epistemo-
sort) thus resembles that of the externalist posi- logical foundationist all he really needs, even
tions considered above. In both cases the justifica- though he has usually, in a spirit of philosophical
tory appeal to further beliefs which generates the greed, sought considerably more. In what follows,

however, I shall sketch a line of argument which, must somehow constitute both an apprehension
if correct, will show that even modest givenism is of the state of affairs and a justification of that
an untenable position. 24 very apprehension, thus pulling itself up by its
The argument to be developed depends on a own cognitive bootstraps. One is reminded here
problem within the givenist position which is of Chisholm's claim that certain cognitive states
surprisingly easy to overlook. I shall therefore justify themselves,2s but that extremely paradoxi-
proceed in the following way. I shall first state the cal remark hardly constitutes an explanation of
problem in an initial way, then illustrate it by how this is possible.
showing how it arises in one recent version of If, on the other hand, an intuition is not a cog-
givenism, and finally consider whether any plau- nitive state and thus involves no cognitive grasp
sible solution is possible. (It will be useful for the of the state of affairs in question, then the need
purposes of this discussion to make two simplify- for a justification for the intuition is obviated, but
ing assumptions, without which the argument at the serious cost of making it difficult to see
would be more complicated, but not essentially how the intuition is supposed to justify the belief.
altered. First, I shall assume that the basic belief If the person in question has no cognitive grasp
which is to be justified by reference to the given or of that state of affairs (or of any other) by virtue
immediately apprehended state of affairs is just of having such an intuition, then how does the
the belief that this same state of affairs obtains. intuition give him a reason for thinking that his
Second, I shall assume that the given or immedi- belief is true or likely to be true? We seem again to
ately apprehended state of affairs is not itself a be back to an externalist position, which it was
belief or other cognitive state.) the whole point of the category of intuition or
Consider then an allegedly basic belief that-p givenness to avoid.
which is supposed to be justified by reference to a As an illustration of this problem, consider
given or immediately apprehended state of affairs Quinton's version of givenism, as outlined in his
that-po Clearly what justifies the belief is not the book The Nature of Things. 26 As noted above,
state of affairs simpliciter, for to say that would be basic beliefs may, according to Quinton, concern
to return to a form of externalism. For the given- ordinary perceptible states of affairs and need not
ist, what justifies the belief is the immediate appre- be certain or incorrigible. (Quinton uses the
hension or intuition of the state of affairs. Thus we phrase "intuitive belief" as I have been using
seem to have three items present in the situation: "basic belief" and calls the linguistic expression
the belief, the state of affairs which is the object of of an intuitive belief a "basic statements"; he also
the belief, and the intuition or immediate appre- seems to pay very little attention to the difference
hension of that state of affairs. The problem to be between beliefs and statements, shifting freely
raised revolves around the nature of the last of back and forth between them, and I will generally
these items, the intuition or immediate appre- follow him in this.) Thus "this book is red" might,
hension (hereafter I will use mainly the former in an appropriate context, be a basic statement
term). It seems to be a cognitive state, perhaps expressing a basic or intuitive belief. But how are
somehow of a more rudimentary sort than a such basic statements (or the correlative beliefs)
belief, which involves the thesis or assertion that-po supposed to be justified? Here Quinton's account,
Now if this is correct, it is easy enough to beyond the insistence that they are not justified
understand in a rough sort of way how an intui- by reference to further beliefs, is seriously unclear.
tion can serve to justify a belief with this same He says rather vaguely that the person is "aware"
assertive content. The problem is to understand (p. 129) or "directly aware" (p. 139) of the appro-
why the intuition, involving as it does the cogni- priate state of affairs, or that he has "direct knowl-
tive thesis that-p, does not itself require justifica- edge" (p. 126) of it, but he gives no real account of
tion. And if the answer is offered that the intuition the nature or epistemological status of this state
is justified by reference to the state of affairs that- of "direct awareness" or "direct knowledge;'
p, then the question will be why this would not though it seems clear that it is supposed to be a
require a second intuition or other apprehension cognitive state of some kind. (In particular, it is
of the state of affairs to justify the original one. not clear what "direct" means, over and above
For otherwise one and the same cognitive state "non -inferential.")27

The difficulty with Quinton's account comes seems to be implicit in most if not all givenist
out most clearly in his discussion of its relation to positions. But when stated thus baldly, this "solu-
the correspondence theory of truth: tion" to the problem seems hopelessly contrived
and ad hoc. If such a move is acceptable, one is
The theory of basic statements is closely con- inclined to expostulate, then once again any sort
nected with the correspondence theory of truth. of regress could be solved in similar fashion.
In its classical form that theory holds that to each Simply postulate a final term in the regress which
true statement, whatever its form may be, a fact is sufficiently similar to the previous terms to sat-
of the same form corresponds. The theory of isfy, with respect to the penultimate term, the
basic statements indicates the point at which sort of need or impetus which originally gener-
correspondence is established, at which the ated the regress; but which is different enough
system of beliefs makes its justifying contact with from previous terms so as not itself to require
the world. (p. 139) satisfaction by a further term. Thus we would
have semi-events, which could cause but need
And further on he remarks that the truth of basic not be caused; semi-explanatia, which could
statements "is directly determined by their corre- explain but need not be explained; and semi-
spondence with fact" (p. 143). (It is clear that beliefs, which could justify but need not be justi-
"determined" here means "epistemically deter- fied. The point is not that such a move is always
mined.") Now it is a familiar but still forceful ide- incorrect (though I suspect that it is), but simply
alist objection to the correspondence theory of that the nature and possibility of such a conven-
truth that if the theory were correct we could ient regress-stopper needs at the very least to be
never know whether any of our beliefs were true, clearly and convincingly established and
since we have no perspective outside our system explained before it can constitute a satisfactory
of beliefs from which to see that they do or do not solution to any regress problem.
correspond. Quinton, however, seems to suppose The main account which has usually been
rather blithely that intuition or direct awareness offered by givenists of such semi-cognitive states
provides just such a perspective, from which we is well suggested by the terms in which immedi-
can in some cases apprehend both beliefs and ate or intuitive apprehensions are described:
world and judge whether or not they correspond. "immediate," "direct," "presentation," etc. The
And he further supposes that the issue of justifi- underlying idea here is that of confrontation: in
cation somehow does not arise for apprehensions intuition, mind or consciousness is directly con-
made from this perspective, though without fronted with its object, without the intervention
giving any account of how or why this is so. of any sort of intermediary. It is in this sense that
My suggestion here is that no such account the object is given to the mind. The root meta-
can be given. As indicated above, the givenist is phor underlying this whole picture is vision:
caught in a fundamental dilemma: if his intui- mind or consciousness is likened to an immate-
tions or immediate apprehensions are construed rial eye, and the object of intuitive awareness is
as cognitive, then they will be both capable of that which is directly before the mental eye and
giving justification and in need of it themselves; if open to its gaze. If this metaphor were to be
they are non-cognitive, then they do not need jus- taken seriously, it would become relatively simple
tification but are also apparently incapable of to explain how there can be a cognitive state
providing it. This, at bottom, is why epistemo- which can justify but does not require justifica-
logical givenness is a myth. 28 tion. (If the metaphor is to be taken seriously
Once the problem is clearly realized, the only enough to do the foundationist any real good, it
possible solution seems to be to split the differ- becomes plausible to hold that the intuitive cog-
ence by claiming that an intuition is a semi-cog- nitive states which result would after all have to
nitive or quasi-cognitive state,29 which resembles be infallible. For if all need for justification is to
a belief in its capacity to confer justification, be precluded, the envisaged relation of confron-
while differing from a belief in not requiring jus- tation seemingly must be conceived as too inti-
tification itself. In fact, some such conception mate to allow any possibility of error. To the

extent that this is so, the various arguments seems to be no way to explain how a basic cogni-
which have been offered against the notion of tive state, whether called a belief or an intuition,
infallible cognitive states count also against this can be directly justified by the world without
version of givenism.) lapsing back into externalism - and from there
Unfortunately, however, it seems clear that the into skepticism. I shall conclude with three fur-
mental eye metaphor will not stand serious scru- ther comments aimed at warding off certain likely
tiny. The mind, whatever else it may be, is not an sorts of misunderstanding. First. It is natural in
eye or, so far as we know, anything like an eye. this connection to attempt to justify basic beliefs
Ultimately the metaphor is just far too simple to by appealing to experience. But there is a familiar
be even minimally adequate to the complexity of ambiguity in the term "experience," which in fact
mental phenomena and to the variety of condi- glosses over the crucial distinction upon which
tions upon which such phenomena depend. This the foregoing argument rests. Thus "experience"
is not to deny that there is considerable intuitive may mean either an experiencing (i.e., a cognitive
appeal to the confrontational model, especially as state) or something experienced (i.e., an object of
applied to perceptual consciousness, but only to cognition). And once this ambiguity is resolved,
insist that this appeal is far too vague in its import the concept of experience seems to be of no par-
to adequately support the very specific sorts of ticular help to the givenist. Second. I have con-
epistemological results which the strong founda- centrated, for the sake of simplicity, on Quinton's
tionist needs. In particular, even if empirical version of givenism in which ordinary physical
knowledge at some point involves some sort of states of affairs are among the things which are
confrontation or seeming confrontation, this by given. But the logic of the argument would be
itself provides no clear reason for attributing essentially the same if it were applied to a more
epistemic justification or reliability, let alone cer- traditional version like Lewis's in which it is pri-
tainty, to the cognitive states, whatever they may vate experiences which are given, and I cannot see
be called, which result. that the end result would be different - though it
Moreover, quite apart from the vicissitudes of might be harder to discern, especially in cases
the mental eye metaphor, there are powerful where the allegedly basic belief is a belief about
independent reasons for thinking that the attempt another cognitive state. Third. Notice carefully
to defend givenism by appeal to the idea of a that the problem raised here with respect to
semi-cognitive or quasi-cognitive state is funda- givenism is a logical problem (in a broad sense of
mentally misguided. The basic idea, after all, is to "logical"). Thus it would be a mistake to think
distinguish two aspects of a cognitive state, its that it can be solved simply by indicating some
capacity to justify other states and its own need sort of state which seems intuitively to have the
for justification, and then try to find a state which appropriate sorts of characteristics; the problem
possesses only the former aspect and not the is to understand how it is possible for any state to
latter. But it seems clear on reflection that these have those characteristics. (The mistake would be
two aspects cannot be separated, that it is one and analogous to one occasionally made in connec-
the same feature of a cognitive state, viz. its asser- tion with the free-will problem: the mistake of
tive content, which both enables it to confer justi- attempting to solve the logical problem of how an
fication on other states and also requires that it be action can be not determined but also not merely
justified itself. If this is right, then it does no good random by indicating a subjective act of effort or
to introduce semi-cognitive states in an attempt similar state, which seems intuitively to satisfy
to justify basic beliefs, since to whatever extent such a description.)
such a state is capable of conferring justification, Thus foundationism appears to be doomed by
it will to that very same extent require justifica- its own internal momentum. No account seems
tion. Thus even if such states do exist, they are of to be available of how an empirical belief can be
no help to the givenist in attempting to answer genuinely justified in an epistemic sense, while
the objection at issue here. 30 avoiding all reference to further empirical beliefs
Hence the givenist response to the anti- or cognitions which themselves would require
foundationist argument seems to fail. There justification. How then is the epistemic regress

problem to be solved? The natural direction to elimination are dangerous at best: there may be
look for an answer is to the coherence theory of further alternatives which have not yet been for-
empirical knowledge and the associated non- mulated; and the possibility still threatens that
linear conception of justification which were the epistemic regress problem may in the end be
briefly mentioned above. 3l But arguments by of aid and comfort only to the skeptic.


Roderick M. Chisholm, Theory of Knowledge making the belief highly likely to be true,
(Englewood Cliffs, NJ: Prentice-Hall, 1966), without trying to say exactly what this
p.30. means.
2 Anthony Quinton, The Nature of Things 4 See Edmund Gettier, "Is Justified True Belief
(London: Routledge and Kegan Paul, 1973), Knowledge?"thisvol.,ch. 15. Also Ackermann,
p. 119. This is an extremely venerable argu- Knowledge and Belief, ch. V, and the corre-
ment, which has played a central role in epis- sponding references.
temological discussion at least since Aristotle's 5 For simplicity, I will speak of inference rela-
statement of it in the Posterior Analytics, Book tions as obtaining between beliefs rather
I, ch. 2-3. (Some have found an anticipation than, more accurately, between the proposi-
of the argument in the Theaetetus at 209E- tions which are believed. "Inference" is to be
2 lOB, but Plato's worry in that passage appears understood here in a very broad sense; any
to be that the proposed definition of knowl- relation between two beliefs which allows
edge is circular, not that it leads to an infinite one, if accepted, to serve as a good reason for
regress of justification.) accepting the other will count as inferential.
3 "Adequately justified" because a belief could 6 It is difficult to give precise criteria for when a
be justified to some degree without being given reason is the reason for a person's hold-
sufficiently justified to qualify as knowledge ing a belief. G. Harman, in Thought (Princeton:
(if true). But it is far from clear just how Princeton University Press, 1973), argues that
much justification is needed for adequacy. for a person to believe for a given reason is for
Virtually all recent epistemologists agree that that reason to explain why he holds that belief.
certainty is not required. But the lottery par- But this suggestion, though heuristically
adox shows that adequacy cannot be under- useful, hardly yields a usable criterion.
stood merely in terms of some specified level 7 Thus it is a mistake to conceive the regress as
of probability. (For a useful account of the a temporal regress, as it would be if each jus-
lottery paradox, see Robert Ackermann, tifying argument had to be explicitly given
Knowledge and Belief (Garden City, NY: before the belief in question was justified.
Doubleday, 1972), pp. 39-50.) Armstrong, in 8 Obviously these views could be combined,
Belief, Truth and Knowledge (London: with different instances of the regress being
Cambridge University Press, 1973), argues handled in different ways. I will not consider
that what is required is that one's reasons for such combined views here. In general, they
the belief be "conclusive;' but the precise would simply inherit all of the objections
meaning of this is less than clear. Ultimately, pertaining to the simpler views.
it may be that the concept of knowledge is 9 Peirce seems to suggest a virtuous regress
simply too crude for refined epistemological view in "Questions concerning Certain
discussion, so that it may be necessary to Faculties Claimed for Man;' Collected Papers
speak instead of degrees of belief and corre- V, pp. 135-55. But the view is presented met-
sponding degrees of justification. I shall aphorically and it is hard to be sure exactly
assume (perhaps controversially) that the what it comes to or to what extent it bears on
proper solution to this problem will not the present issue.
affect the issues to be discussed here, and 10 The original statement of the non-linear
speak merely of the reasons or justification view was by Bernard Bosanquet in Implication

and Linear Inference (London, 1920). For counterexamples. But I shall ignore this
more recent discussions, see Harman, refinement here.
Thought; and Nicholas Rescher, "Founda- 18 On a Carnap-style a priori theory of proba-
tionalism, Coherentism, and the Idea of bility it could, of course, be the case that very
Cognitive Systematization;' The Journal of general empirical propositions were more
Philosophy 71 (1974), pp. 695-708. likely to be true than not, i.e. that the possi-
11 I have attempted to show how a coherence ble state-descriptions in which they are true
view might be defended against the most outnumber those in which they are false. But
standard of these objections III "The clearly this would not make them likely to be
Coherence Theory of Empirical Knowledge," true in a sense which would allow the
Philosophical Studies 30 (1976), pp. detached assertion of the proposition in
281-312. question (on pain of contradiction), and this
12 The presumption against a skeptical out- fact seems to preclude such justification
come is strong, but I think it is a mistake to from being adequate for knowledge.
treat it as absolute. If no non-skeptical theory 19 Chs 11-13. Bracketed page references in this
can be found which is at least reasonably section are to this book.
plausible in its own right, skepticism might 20 Fred I. Dretske, Seeing and Knowing (London:
become the only rational alternative. Routledge and Kegan Paul, 1969), ch. III,
13 For some useful distinctions among these especially pp. 126-39. It is difficult to be
terms, see William Alston, "Varieties of quite sure of Dretske's view, however, since
Privileged Access;' American Philosophical he is not concerned in this book to offer a
Quarterly 8 (1971), pp. 223-4l. general account of knowledge. Views which
14 For discussions of weak foundation ism, see are in some ways similar to those of
Bertrand Russell, Human Knowledge (New Armstrong and Dretske have been offered by
York: Simon and Schuster, 1949), part II, ch. Goldman and by Unger. See Alvin Goldman,
II, and part V, chs. 6 and 7; Nelson Goodman, "A Causal Theory of Knowing;' The Journal
"Sense and Certainty," Philosophical Review of Philosophy 64 (1967), pp. 357-72; and
61 (1952), pp. 160-7; Israel Scheffler, Science Peter Unger, "An Analysis of Factual
and Subjectivity (New York, 1967), chapter V; Knowledge," The Journal of Philosophy 65
and Roderick Firth, "Coherence, Certainty, (1968), pp. 157-70. But both Goldman and
and Epistemic Priority;' The Journal of Unger are explicitly concerned with the
Philosophy 61 (1964), pp. 545-57. Gettier problem and not at all with the
15 How good a reason must one have? regress problem, so it is hard to be sure how
Presumably some justification accrues from their views relate to the sort of externalist
any reason which makes the belief even min- view which is at issue here.
imally more likely to be true than not, but 21 On the one hand, Armstrong seems to argue
considerably more than this would be that it is not a requirement for knowledge
required to make the justification adequate that the believer have "sufficient evidence"
for knowledge. (See note 3, above.) (The for his belief, which sounds like a rejection
James-Clifford controversy concerning the of the adequate-justification condition. On
"will to believe" is also relevant here. I am the other hand, he seems to want to say that
agreeing with Clifford to the extent of saying the presence of the external relation makes it
that epistemic justification requires some rational for a person to accept a belief, and
positive reason in favor of the belief and not he seems (though this is not clear) to have
just the absence of any reason against.) epistemic rationality in mind; and there
16 For a similar use of the notion of epistemic appears to be no substantial difference
irresponsibility, see Ernest Sosa, "How Do between saying that a belief is epistemically
You Know?" American Philosophical Quarterly rational and saying that it is epistemically
11(1974), p. 117. justified.
17 In fact, the premises would probably have to 22 One way to put this point is to say that
be true as well, in order to avoid Gettier-type whether a belief is likely to be true or whether

in contrast it is an accident that it is true "On the Logical Positivists' Theory of Truth:'
depends significantly on how the belief is Analysis, 2 (1934-5), pp. 49-59. The Hempel
described. Thus it might be true of one and paper is in part a reply to a foundationist cri-
the same belief that it is "a belief connected tique of Neurath by Schlick in "The
in a law-like way with the state of affairs Foundation of Knowledge," also translated
which it describes" and also that it is "a belief in Ayer, Logical Positivism, pp. 209-27.
adopted on the basis of no apparent evi- Schlick replied to Hempel in "Facts and
dence"; and it might be likely to be true on Propositions," and Hempel responded in
the first description and unlikely to be true "Some Remarks on 'Facts' and Propositions,"
on the second. The claim here is that it is the both in Analysis 2 (1934-5), pp. 65-70 and
believer's own conception which should be 93-6, respectively. Though the Neurath-
considered in deciding whether the belief is Hempel argument conflates issues having to
justified. (Something analogous seems to be do with truth and issues having to do with
true in ethics: the moral worth of a person's justification in a confused and confusing
action is correctly to be judged only in terms way, it does bring out the basic objection to
of that person's subjective conception of givenism.
what he is doing and not in light of what 25 Chisholm, "Theory of Knowledge," in
happens, willy-nilly, to result from it.) Chisholm et aI., Philosophy (Englewood
23 Notice, however, that if beliefs standing in Cliffs, NJ: Prentice-Hall, 1964), pp. 270 ff.
the proper external relation should happen 26 Bracketed page references in this section will
to possess some subjectively distinctive fea- be to this book.
ture (such as being spontaneous and highly 27 Quinton does offer one small bit of clarifica-
compelling to the believer), and if the tion here, by appealing to the notion of
believer were to notice empirically, that ostensive definition and claiming in effect
beliefs having this feature were true a high that the sort of awareness involved in the
proportion of the time, he would then be in intuitive justification of a basic belief is the
a position to construct a justification for a same as that involved in a situation of osten-
new belief of that sort along the lines sive definition. But such a comparison is of
sketched at the end of section II. But of little help, for at least two reasons. First, as
course a belief justified in that way would no Wittgenstein, Sellars, and others have argued,
longer be basic. the notion of ostensive definition is itself
24 I suspect that something like the argument seriously problematic. Indeed, an objection
to be given here is lurking somewhere quite analogous to the present one against
in Sellars' "Empiricism and the Philosophy the notion of a basic belief could be raised
of Mind" (reprinted in Sellars, Science, against the notion of an ostensive definition;
Perception, and Reality (London: Routledge and this objection, if answerable at all, could
and Kegan Paul, 1963), pp. 127-96), but it is only be answered by construing the aware-
difficult to be sure. A more recent argument ness involved in ostension in such a way as to
by Sellars which is considerably closer on the be of no help to the foundationist in the
surface to the argument offered here is con- present discussion. Second, more straight-
tained in "The Structure of Knowledge:' his forwardly, even if the notion of ostensive
Machette Foundation Lectures given at the definition were entirely unobjectionable,
University of Texas in 1971, in Hector-Nerl there is no need for the sort of awareness
Casteneda (ed.), Action, Knowledge, and involved to be justified. If all that is at issue is
Reality: Critical Studies in Honor of Wilfrid learning the meaning of a word (or acquir-
Sellars (Indianapolis, 1975), Lecture III, sec- ing a concept), then justification is irrelevant.
tions III-IV. A similar line of argument was Thus the existence of ostensive definitions
also offered by Neurath and Hempel. See would not show how there could be basic
Otto Neurath, "Protocol Sentences:' tr. in beliefs.
A. J. Ayer (ed.), Logical Positivism (New York, 28 Notice, however, that to reject an epistemo-
1959), pp. 199-208; and Carl G. Hempel, logical given does not necessarily rule out

other varieties of givenness which may have 29 Compare the Husserlian notion of a "pre-
importance for other philosophical issues. In predicative awareness:'
particular, there may still be viable versions 30 It is interesting to note that Quinton seems
of givenness which pose an obstacle to mate- to offer an analogous critique of givenness in
rialist views in the philosophy of mind. For an earlier paper, "The Problem of Perception,"
useful distinctions among various versions reprinted in Robert J. Swartz (ed. ),Perceiving,
of givenness and a discussion of their rele- Sensing, and Knowing (Garden City, NY:
vance to the philosophy of mind, see James Doubleday, 1965), pp. 497-526; cf. especially
w. Cornman, "Materialism and Some Myths p.503.
about Some Givens," The Monist 56 (1972), 31 For a discussion of such a coherence theory,
pp.215-33. see my paper cited i,l note 11, above.
A Coherence Theory
of Truth and Knowledge

Donald Davidson

In this paper I defend what may as well be called are given by objective truth conditions there is a
a coherence theory of truth and knowledge. The question how we can know that the conditions
theory I defend is not in competition with a cor- are satisfied, for this would appear to require a
respondence theory, but depends for its defense confrontation between what we believe and real-
on an argument that purports to show that coher- ity; and the idea of such a confrontation is absurd.
ence yields correspondence. But if coherence is a test of truth, then coherence
The importance of the theme is obvious. If is a test for judging that objective truth condi-
coherence is a test of truth, there is a direct con- tions are satisfied, and we no longer need to
nection with epistemology, for we have reason to explain meaning on the basis of possible confron-
believe many of our beliefs cohere with many tation. My slogan is: correspondence without
others, and in that case we have reason to believe confrontation. Given a correct. epistemology, we
many of our beliefs are true. When the beliefs are can be realists in all departments. We can accept
true, then the primary conditions for knowledge objective truth conditions as the key to meaning,
would seem to be satisfied. a realist view of truth, and we can insist that
Someone might try to defend a coherence theory knowledge is of an objective world independent
of truth without defending a coherence theory of of our thought or language.
knowledge, perhaps on the ground that the holder Since there is not, as far as I know, a theory
of a coherent set of beliefs might lack a reason to that deserves to be called "the" coherence theory,
believe his beliefs coherent. This is not likely, but let me characterize the sort of view I want to
it may be that someone, though he has true defend. It is obvious that not every consistent set
beliefs, and good reasons for holding them, does of interpreted sentences contains only true sen-
not appreciate the relevance of reason to belief. tences, since one such set might contain just the
Such a one may best be viewed as having knowl- consistent sentence 5 and another just the nega-
edge he does not know he has: he thinks he is a tion of S. And adding more sentences, while
skeptic. In a word, he is a philosopher. maintaining consistency, will not help. We can
Setting aside aberrant cases, what brings truth imagine endless state-descriptions - maximal
and knowledge together is meaning. If meanings consistent descriptions - which do not describe
our world.
My coherence theory concerns beliefs, or sen-
Originally published in Ernest LePore (ed.), Truth and tences held true by someone who understands
Interpretation: Perspectives on the Philosophy of Donald them. I do not want to say, at this point, that every
Davidson (New York: Blackwell, 1989), pp. 307-19. possible coherent set of beliefs is true (or contains

mostly true beliefs). I shy away from this because it of the true theory (if there is one) of evidential
is so unclear what is possible. At one extreme, it support must be in there? Since no person has a
might be held that the range of possible maximal completely consistent body of convictions, coher-
sets of beliefs is as wide as the range of possible ence with which beliefs creates a presumption of
maximal sets of sentences, and then there would truth? Some of these problems will be put in
be no point to insisting that a defensible coherence better perspective as I go along.
theory concerns beliefs and not propositions or It should be clear that I do not hope to define
sentences. But there are other ways of conceiving truth in terms of coherence and belief. Truth is
what it is possible to believe which would justify beautifully transparent compared to belief and
saying not only that all actual coherent belief sys- coherence, and I take it as primitive. Truth,
tems are largely correct but that all possible ones as applied to utterances of sentences, shows the
are also. The difference between the two notions disquotational feature enshrined in Tarski's
of what it is possible to believe depends on what Convention T, and that is enough to fix its domain
we suppose about the nature of belief, its interpre- of application. Relative to a language or a speaker,
tation, its causes, its holders, and its patterns. of course, so there is more to truth than
Beliefs for me are states of people with intentions, Convention T; there is whatever carries over from
desires, sense organs; they are states that are caused language to language or speaker to speaker. What
by, and cause, events inside and outside the bodies Convention T, and the trite sentences it declares
of their entertainers. But even given all these con- true, like" 'Grass is green' spoken by an English
straints, there are many things people do believe, speaker, is true if and only if grass is green", reveal
and many more that they could. For all such cases, is that the truth of an utterance depends on just
the coherence theory applies. two things: what the words as spoken mean, and
Of course some beliefs are false. Much of the how the world is arranged. There is no further
point of the concept of belief is the potential gap relativism to a conceptual scheme, a way of view-
it introduces between what is held to be true and ing things, a perspective. Two interpreters, as
what is true. So mere coherence, no matter how unlike in culture, language and point of view as
strongly coherence is plausibly defined, can not you please, can disagree over whether an utter-
guarantee that what is believed is so. All that a ance is true, but only if they differ on how things
coherence theory can maintain is that most of the are in the world they share, or what the utterance
beliefs in a coherent total set of beliefs are true. means.
This way of stating the position can at best be I think we can draw two conclusions from
taken as a hint, since there is probably no useful these simple reflections. First, truth is corre-
way to count beliefs, and so no clear meaning to spondence with the way things are. (There is no
the idea that most of a person's beliefs are true. straightforward and non-misleading way to state
A somewhat better way to put the point is to say this; to get things right, a detour is necessary
there is a presumption in favor of the truth of a through the concept of satisfaction in terms of
belief that coheres with a significant mass of which truth is characterized. l ) So if a coherence
belie Every belief in a coherent total set of beliefs theory of truth is acceptable, it must be consistent
is justified in the light of this presumption, much with a correspondence theory. Second, a theory
as every intentional action taken by a rational of knowledge that allows that we can know the
agent (one whose choices, beliefs and desires truth must be a non-relativized, non-internal
cohere in the sense of Bayesian decision theory) is form of realism. So if a coherence theory of
justified. So to repeat, if knowledge is justified knowledge is acceptable, it must be consistent
true belief, then it would seem that all the true with such a form of realism. My form of realism
beliefs of a consistent believer constitute knowl- seems to be neither Hilary Putnam's internal real-
edge. This conclusion, though too vague and ism nor his metaphysical realism. 2 It is not inter-
hasty to be right, contains an important core of nal realism because internal realism makes truth
truth, as I shall argue. Meanwhile I merely note relative to a scheme, and this is an idea I do not
the many problems asking for treatment: what think is intelligible. 3 A major reason, in fact, for
exactly does coherence demand? How much of accepting a coherence theory is the unintelligibil-
inductive practice should be included, how much ity of the dualism of a conceptual scheme and

a "world" waiting to be coped with. But my and there is no way to get outside our beliefs and
realism is certainly not Putnam's metaphysical our language so as to find some test other than
realism, for it is characterized by being "radically coherence."4 About this I am, as you see, in agree-
non-epistemic", which implies that all our best ment with Rorty. Where we differ, if we do, is on
researched and established thoughts and theories whether there remains a question how, given that
may be false. I think the independence of belief we cannot "get outside our beliefs and our lan-
and truth requires only that each of our beliefs guage so as to find some test other than coher-
may be false. But of course a coherence theory ence", we nevertheless can have knowledge of, and
cannot allow that all of them can be wrong. talk about, an objective public world which is not
But why not? Perhaps it is obvious that the of our own making. I think this question does
coherence of a belief with a substantial body of remain, while I suspect that Rorty doesn't think
belief enhances its chance of being true, provided so. If this is his view, then he must think I am
there is reason to suppose the body of belief is making a mistake in trying to answer the ques-
true, or largely so. But how can coherence alone tion. Nevertheless, here goes.
supply grounds for belief? Mayhap the best we It will promote matters at this point to review
can do to justify one belief is to appeal to other very hastily some of the reasons for abandoning
beliefs. But then the outcome would seem to be the search for a basis for knowledge outside the
that we must accept philosophical skepticism, no scope of our beliefs. By "basis" here I mean
matter how unshaken in practice our beliefs specifically an epistemological basis, a source of
remalll. justification.
This is skepticism in one of its traditional The attempts worth taking seriously attempt
garbs. It asks: Why couldn't all my beliefs hang to ground belief in one way or another on the tes-
together and yet be comprehensively false about timony of the senses: sensation, perception, the
the actual world? Mere recognition of the fact given, experience, sense data, the passing show.
that it is absurd or worse to try to confront our All such theories must explain at least these two
beliefs, one by one, or as a whole, with what they things: what, exactly, is the relation between sen-
are about does not answer the question nor show sation and belief that allows the first to justify the
the question unintelligible. In short, even a mild second? and, why should we believe our sensa-
coherence theory like mine must provide a skep- tions are reliable, that is, why should we trust our
tic with a reason for supposing coherent beliefs senses?
are true. The partisan of a coherence theory can't The simplest idea is to identify certain beliefs
allow assurance to come from outside the system with sensations. Thus Hume seems not to have
of belief, while nothing inside can produce sup- distinguished between perceiving a green spot and
port except as it can be shown to rest, finally or at perceiving that a spot is green. (An ambiguity in
once, on something independently trustworthy. the word "idea" was a great help here.) Other phi-
It is natural to distinguish coherence theories losophers noted Hume's confusion, but tried to
from others by reference to the question whether attain the same results by reducing the gap
or not justification can or must come to an end. between perception and judgement to zero by
But this does not define the positions, it merely attempting to formulate judgements that do not
suggests a form the argument may take. For there go beyond stating that the perception or sensation
are coherence theorists who hold that some beliefs or presentation exists (whatever that may mean).
can serve as the basis for the rest, while it would Such theories do not justify beliefs on the basis of
be possible to maintain that coherence is not sensations, but try to justify certain beliefs by
enough, although giving reasons never comes to claiming that they have exactly the same epistemic
an end. What distinguishes a coherence theory is content as a sensation. There are two difficulties
simply the claim that nothing can count as a with such a view: first, if the basic beliefs do not
reason for holding a belief except another belief. exceed in content the corresponding sensation
Its partisan rejects as unintelligible the request for they cannot support any inference to an objective
a ground or source of justification of another ilk. world; and second, there are no such beliefs.
As Rorty has put it, "nothing counts as justifica- A more plausible line is to claim that we cannot
tion unless by reference to what we already accept, be wrong about how things appear to us to be.

If we believe we have a sensation, we do; this is to answer our second question: What justifies
held to be an analytic truth, or a fact about how the belief that our senses do not systematically
language is used. deceive us? For even if sensations justify belief in
It is difficult to explain this supposed connec- sensation, we do not yet see how they justify belief
tion between sensations and some beliefs in a way in external events and objects.
that does not invite skepticism about other minds, Quine tells us that science tells us that "our
and in the absence of an adequate explanation, only source of information about the external
there should be a doubt about the implications of world is through the impact of light rays and
the connection for justification. But in any case, it molecules upon our sensory surfaces".5 What
is unclear how, on this line, sensations justify the worries me is how to read the words "source" and
belief in those sensations. The point is rather that "information': Certainly it is true that events and
such beliefs require no justification, for the exist- objects in the external world cause us to believe
ence of the belief entails the existence of the sen- things about the external world, and much, if not
sation, and so the existence of the belief entails its all, of the causality takes a route through the sense
own truth. Unless something further is added, we organs. The notion of information, however,
are back to another form of coherence theory. applies in a non-metaphorical way only to the
Emphasis on sensation or perception in mat- engendered beliefs. So "source" has to be read
ters epistemological springs from the obvious simply as "cause" and "information" as "true
thought: sensations are what connect the world belief" or "knowledge". Justification of beliefs
and our beliefs, and they are candidates for justi- caused by our senses is not yet in sight. 6
fiers because we often are aware of them. The The approach to the problem of justification
trouble we have been running into is that the jus- we have been tracing must be wrong. We have
tification seems to depend on the awareness, been trying to see it this way: a person has all his
which is just another belief. beliefs about the world - that is, all his beliefs.
Let us try a bolder tack. Suppose we say that How can he tell if they are true, or apt to be true?
sensations themselves, verbalized or not, justify Only, we have been assuming, by connecting his
certain beliefs that go beyond what is given in beliefs to the world, confronting certain of his
sensation. So, under certain conditions, having beliefs with the deliverances of the senses one by
the sensation of seeing a green light flashing may one, or perhaps confronting the totality of his
justify the belief that a green light is flashing. The beliefs with the tribunal of experience. No such
problem is to see how the sensation justifies the confrontation makes sense, for of course we can't
belief. Of course if someone has the sensation of get outside our skins to find out what is causing
seeing a green light flashing, it is likely, under cer- the internal happenings of which we are aware.
tain circumstances, that a green light is flashing. Introducing intermediate steps or entities into
We can say this, since we know of his sensation, the causal chain, like sensations or observations,
but he can't say it, since we are supposing he is serves only to make the epistemological problem
justified without having to depend on believing more obvious. For if the intermediaries are merely
he has the sensation. Suppose he believed he causes, they don't justify the beliefs they cause,
didn't have the sensation. Would the sensation while if they deliver information, they may be
still justify him in the belief in an objective flash- lying. The moral is obvious. Since we can't swear
ing green light? intermediaries to truthfulness, we should allow
The relation between a sensation and a belief no intermediaries between our beliefs and their
cannot be logical, since sensations are not beliefs objects in the world. Of course there are causal
or other propositional attitudes. What then is the intermediaries. What we must guard against are
relation? The answer is, I think, obvious: the rela- epistemic intermediaries.
tion is causal. Sensations cause some beliefs and There are common views of language that
in this sense are the basis or ground of those encourage bad epistemology. This is no accident,
beliefs. But a causal explanation of a belief does of course, since theories of meaning are connected
not show how or why the belief is justified. with epistemology through attempts to answer
The difficulty of transmuting a cause into a the question how one determines that a sentence
reason plagues the anticoherentist again ifhe tries is true. If knowing the meaning of a sentence

(knowing how to give a correct interpretation of forms of verification ism, makes for skepticism.
it) involves, or is, knowing how it could be recog- For clearly a person's sensory stimulations could
nized to be true, then the theory of meaning raises be just as they are and yet the world outside very
the same question we have been struggling with, different. (Remember the brain in the vat.)
for giving the meaning of a sentence will demand Quine's way of doing without meanings is
that we specify what would justify asserting it. subtle and complicated. He ties the meanings of
Here the coherentist will hold that there is no use some sentences directly to patterns of stimulation
looking for a source of justification outside of (which also constitute the evidence, Quine thinks,
other sentences held true, while the foundational- for assenting to the sentence), but the meanings
ist will seek to anchor at least some words or sen- of further sentences are determined by how they
tences to non-verbal rocks. This view is held, I are conditioned to the original, or observation
think, both by Quine and by Michael Dummett. sentences. The facts of such conditioning do not
Dummett and Quine differ, to be sure. In par- permit a sharp division between sentences held
ticular, they disagree about holism, the claim that true by virtue of meaning and sentences held true
the truth of our sentences must be tested together on the basis of observation. Quine made this
rather than one by one. And they disagree also, point by showing that if one way of interpreting
and consequently, about whether there is a useful a speaker's utterances was satisfactory, so were
distinction between analytic and synthetic sen- many others. This doctrine of the indeterminacy
tences, and about whether a satisfactory theory of of translation, as Quine called it, should be
meaning can allow the sort of indeterminacy viewed as neither mysterious nor threatening. It
Quine argues for. (On all these points, I am is no more mysterious than the fact that tempera-
Quine's faithful student.) ture can be measured in Centigrade or Fahrenheit
But what concerns me here is that Quine and (or any linear transformation of those numbers).
Dummett agree on a basic principle, which is that And it is not threatening because the very proce-
whatever there is to meaning must be traced back dure that demonstrates the degree of indetermi-
somehow to experience, the given, or patterns of nacy at the same time demonstrates that what is
sensory stimulation, something intermediate determinate is all we need.
between belief and the usual objects our beliefs In my view, erasing the line between the ana-
are about. Once we take this step, we open the lytic and synthetic saved philosophy of language
door to skepticism, for we must then allow that a as a serious subject by showing how it could be
very great many - perhaps most - of the sentences pursued without what there cannot be: determi-
we hold to be true may in fact be false. It is ironi- nate meanings. I now suggest also giving up the
cal. Trying to make meaning accessible has made distinction between observation sentences and
truth inaccessible. When meaning goes epistemo- the rest. For the distinction between sentences
logical in this way, truth and meaning are neces- belief in whose truth is justified by sensations and
sarily divorced. One can, of course, arrange a sentences belief in whose truth is justified only by
shotgun wedding by redefining truth as what we appeal to other sentences held true is as anathema
are justified in asserting. But this does not marry to the coherentist as the distinction between
the original mates. beliefs justified by sensations and beliefs justified
Take Quine's proposal that whatever there is to only by appeal to further beliefs. Accordingly, I
the meaning (information value) of an observa- suggest we give up the idea that meaning or
tion sentence is determined by the patterns of knowledge is grounded on something that counts
sensory stimulation that would cause a speaker to as an ultimate source of evidence. No doubt
assent to or dissent from the sentence. This is a meaning and knowledge depend on experience,
marvellously ingenious way of capturing what is and experience ultimately on sensation. But this
appealing about verificationist theories without is the "depend" of causality, not of evidence or
having to talk of meanings, sense-data, or sensa- justification.
tions; for the first time it made plausible the idea I have now stated my problem as well as I can.
that one could, and should, do what I call the The search for an empirical foundation for mean-
theory of meaning without need of what Quine ing or knowledge leads to skepticism, while a
calls meanings. But Quine's proposal, like other coherence theory seems at a loss to provide any

reason for a believer to believe that his beliefs, if other levels. Since the propositional attitudes
coherent, are true. We are caught between a false are deeply interlocked, we cannot learn the
answer to the skeptic, and no answer. nature of one by first winning understanding of
The dilemma is not a true one. What is needed another. As interpreters, we work our way into
to answer the skeptic is to show that someone the whole system, depending much on the pattern
with a (more or less) coherent set of beliefs has a of interrelationships.
reason to suppose his beliefs are not mistaken in Take for example the interdependence of
the main. What we have shown is that it is absurd belief and meaning. What a sentence means
to look for a justifying ground for the totality of depends partly on the external circumstances
beliefs, something outside this totality which we that cause it to win some degree of conviction;
can use to test or compare with our beliefs. The and partly on the relations, grammatical, logical
answer to our problem must then be to find a or less, that the sentence has to other sentences
reason for supposing most of our beliefs are true held true with varying degrees of conviction.
that is not a form of evidence. Since these relations are themselves translated
My argument has two parts. First I urge that a directly into beliefs, it is easy to see how meaning
correct understanding of the speech, beliefs, depends on belief. Belief, however, depends
desires, intentions and other propositional atti- equally on meaning, for the only access to the
tudes of a person leads to the conclusion that fine structure and individuation of beliefs is
most of a person's beliefs must be true, and so through the sentences speakers and interpreters
there is a legitimate presumption that anyone of speakers use to express and describe beliefs. If
of them, if it coheres with most of the rest, is true. we want to illuminate the nature of meaning and
Then I go on to claim that anyone with thoughts, belief, therefore, we need to start with something
and so in particular anyone who wonders whether that assumes neither. Quine's suggestion, which I
he has any reason to suppose he is generally right shall essentially follow, is to take prompted assent
about the nature of his environment, must know as basic, the causal relation between assenting to a
what a belief is, and how in general beliefs are to sentence and the cause of such assent. This is a
be detected and interpreted. These being perfectly fair place to start the project of identifying beliefs
general facts we cannot fail to use when we com- and meanings, since a speaker's assent to a sen-
municate with others, or when we try to commu- tence depends both on what he means by the
nicate with others, or even when we merely think sentence and on what he believes about the
we are communicating with others, there is a world. Yet it is possible to know that a speaker
pretty strong sense in which we can be said to assents to a sentence without knowing either
know that there is a presumption in favor of the what the sentence, as spoken by him, means, or
overall truthfulness of anyone's beliefs, including what belief is expressed by it. Equally obvious is
our own. So it is bootless for someone to ask for the fact that once an interpretation has been
some further reassurance; that can only add to his given for a sentence assented to, a belief has been
stock of beliefs. All that is needed is that he recog- attributed. If correct theories of interpretation
nize that belief is in its nature veridical. are not unique (do not lead to uniquely correct
Belief can be seen to be veridical by considering interpretations), the same will go for attribu-
what determines the existence and contents of a tions of belief, of course, as tied to acquiescence
belief. Belief, like the other so-called propositional in particular sentences.
attitudes, is supervenient on facts of various sorts, A speaker who wishes his words to be under-
behavioral, neuro-physiological, biological and stood cannot systematically deceive his would-be
physical. The reason for pointing this out is not to interpreters about when he assents to sentences -
encourage definitional or nomological reduction that is, holds them true. As a matter of principle,
of psychological phenomena to something more then, meaning, and by its connection with mean-
basic, and certainly not to suggest epistemologi- ing, belief also, are open to public determination.
cal priorities. The point is rather understanding. I shall take advantage of this fact in what follows
We gain one kind of insight into the nature of the and adopt the stance of a radical interpreter when
propositional attitudes when we relate them sys- asking about the nature of belief. What a fully
tematically to one another and to phenomena on informed interpreter could learn about what a

speaker means is all there is to learn; the same would add the same for first-order quantification
goes for what the speaker believes. 7 theory. This leads directly to the identification
The interpreter's problem is that what he is of the logical constants, as well as to assigning a
assumed to know - the causes of assents to sen- logical form to all sentences.
tences of a speaker - is, as we have seen, the prod- Something like charity operates in the inter-
uct of two things he is assumed not to know, pretation of those sentences whose causes of
meaning and belief. If he knew the meanings he assent come and go with time and place: when the
would know the beliefs, and if he knew the beliefs interpreter finds a sentence of the speaker the
expressed by sentences assented to, he would speaker assents to regularly under conditions
know the meanings. But how can he learn both at he recognizes, he takes those conditions to be the
once, since each depends on the other? truth conditions of the speaker's sentence. This is
The general lines of the solution, like the prob- only roughly right, as we shall see in a moment.
lem itself, are owed to Quine. I will, however, Sentences and predicates less directly geared to
introduce some changes into Quine's solution, as easily detected goings-on can, in Quine's canon,
I have into the statement of the problem. The be interpreted at will, given only the constraints
changes are directly relevant to the issue of episte- of interconnections with sentences conditioned
mological skepticism. directly to the world. Here I would extend the
I see the aim of radical interpretation (which principle of charity to favor interpretations that
is much, but not entirely, like Quine's radical as far as possible preserve truth: I think it makes
translation) as being to produce a Tarski-style for mutual understanding, and hence for better
characterization of truth for the speaker's lan- interpretation, to interpret what the speaker
guage, and a theory of his beliefs. (The second accepts as true when we can. In this matter, r have
follows from the first plus the presupposed less choice than Quine, because r do not see how
knowledge of sentences held true.) This adds little to draw the line between observation sentences
to Quine's program of translation, since transla- and theoretical sentences at the start. There are
tion of the speaker's language into one's own plus several reasons for this, but the one most relevant
a theory of truth for one's own language add up to the present topic is that this distinction is ulti-
to a theory of truth for the speaker. But the shift mately based on an epistemological consideration
to the semantic notion of truth from the syntactic of a sort r have renounced: observation sentences
notion of translation puts the formal restrictions are directly based on something like sensation -
of a theory of truth in the foreground, and patterns of sensory stimulation - and this is an
emphasizes one aspect of the close relation idea I have been urging leads to skepticism.
between truth and meaning. 'Nithout the direct tie to sensation or stimulation,
The principle of charity plays a crucial role the distinction between observation sentences
in Quine's method, and an even more crucial role and others can't be drawn on epistemologically
in my variant. In either case, the principle directs significant grounds. The distinction between sen-
the interpreter to translate or interpret so as to tences whose causes to assent come and go with
read some of his own standards of truth into the observable circumstances and those a speaker
pattern of sentences held true by the speaker. The clings to through change remains however, and
point of the principle is to make the speaker offers the possibility of interpreting the words
intelligible, since too great deviations from con- and sentences beyond the logical.
sistency and correctness leave no common The details are not here to the point. What
ground on which to judge either conformity or should be clear is that if the account I have given
difference. From a formal point of view, the prin- of how belief and meaning are related and under-
ciple of charity helps solve the problem of the stood by an interpreter, then most of the sen-
interaction of meaning and belief by restraining tences a speaker holds to be true - especially the
the degrees of freedom allowed belief while ones he holds to most stubbornly, the ones most
determining how to interpret words. central to the system of his beliefs - most of these
We have no choice, Quine has urged, but to sentences are true, at least in the opinion of the
read our own logic into the thoughts of a speaker; interpreter. For the only, and therefore unim-
Quine says this for the sentential calculus, and I peachable, method available to the interpreter

automatically puts the speaker's beliefs in accord to be largely correct and consistent by objective
with the standards of logic of the interpreter, and standards. We may also, if we want, let the omnis-
hence credits the speaker with plain truths of cient interpreter turn his attention to the fallible
logic. Needless to say there are degrees of logical interpreter of the fallible speaker. It turns out that
and other consistency, and perfect consistency is the fallible interpreter can be wrong about some
not to be expected. What needs emphasis is only things, but not in general; and so he cannot share
the methodological necessity for finding consist- universal error with the agent he is interpreting.
encyenough. Once we agree to the general method of interpre-
Nor, from the interpreter's point of view, is tation I have sketched, it becomes impossible cor-
there any way he can discover the speaker to be rectly to hold that anyone could be mostly wrong
largely wrong about the world. For he interprets about how things are.
sentences held true (which is not to be distin- There is, as I noted above, a key difference
guished from attributing beliefs) according to the between the method of radical interpretation I
events and objects in the outside world that cause am now recommending, and Quine's method of
the sentence to be held true. radical translation. The difference lies in the
What I take to be the important aspect of this nature of the choice of causes that govern inter-
approach is apt to be missed because the approach pretation. Quine makes interpretation depend on
reverses our natural way of thinking of commu- patterns of sensory stimulation, while I make it
nication derived from situations in which under- depend on the external events and objects the
standing has already been secured. Once sentence is interpreted as being about. Thus
understanding has been secured we are able, Quine's notion of meaning is tied to sensory
often, to learn what a person believes quite inde- criteria, something he thinks that can be treated
pendently of what caused him to believe it. This also as evidence. This leads Quine to give epis-
may lead us to the crucial, indeed fatal, conclu- temic significance to the distinction between
sion that we can in general fix what someone observation sentences and others, since observa-
means independently of what he believes and tion sentences are supposed, by their direct
independently of what caused the belief. But if I conditioning to the senses, to have a kind of extra-
am right, we can't in general first identify beliefs linguistic justification. This is the view against
and meanings and then ask what caused them. which I argued in the first part of my paper, urging
The causality plays an indispensable role in deter- that sensory stimulations are indeed part of the
mining the content of what we say and believe. causal chain that leads to belief, but cannot, with-
This is a t~1Ct we can be led to recognize by taking out confusion, be considered to be evidence, or a
up, as we have, the interpreter's point of view. source of justification, for the stimulated beliefs.
It is an artifact of the interpreter's correct What stands in the way of global skepticism of
interpretation of a person's speech and attitudes the senses is, in my view, the fact that we must, in
that there is a large degree of truth and consist- the plainest and methodologically most basic
ency in the thought and speech of an agent. But cases, take the objects of a belief to be the causes
this is truth and consistency by the interpreter's of that belief. And what we, as interpreters, must
standards. Why couldn't it happen that speaker take them to be is what they in fact are.
and interpreter understand one another on the Communication begins where causes converge:
basis of shared but erroneous beliefs? This can, your utterance means what mine does if belief in
and no doubt often does, happen. But it cannot its truth is systematically caused by the same
be the rule. For imagine for a moment an inter- events and objects. s
preter who is omniscient about the world, and The difficulties in the way of this view are
about what does and would cause a speaker to obvious, but I think they can be overcome. The
assent to any sentence in his (potentially unlim- method applies directly, at best, only to occasion
ited) repertoire. The omniscient interpreter, using sentences - the sentences assent to which is caused
the same method as the fallible interpreter, finds systematically by common changes in the world.
the fallible speaker largely consistent and correct. Further sentences are interpreted by their condi-
By his own standards, of course, but since these tioning to occasion sentences, and the appearance
are objectively correct, the fallible speaker is seen in them of words that appear also in occasion

sentences. Among occasion sentences, some will that mayor may not jibe with reality. But beliefs
vary in the credence they command not only in are also identified, directly and indirectly, by their
the face of environmental change, but also in the causes. What an omniscient interpreter knows a
face of change of credence awarded related sen- fallible interpreter gets right enough if he under-
tences. Criteria can be developed on this basis to stands a speaker, and this is just the complicated
distinguish degrees of observationality on inter- causal truth that makes us the believers we are,
nal grounds, without appeal to the concept of a and fixes the contents of our beliefs. The agent
basis for belief outside the circle of beliefs. has only to reflect on what a belief is to appreciate
Related to these problems, and easier still to that most of his basic beliefs are true, and among
grasp, is the problem of error. For even in the sim- his beliefs, those most securely held and that
plest cases it is clear that the same cause (a rabbit cohere with the main body of his beliefs are the
scampers by) may engender different beliefs in most apt to be true. The question, how do I know
speaker and observer, and so encourage assent to my beliefs are generally true? thus answers itself,
sentences which cannot bear the same interpreta- simply because beliefs are by nature generally true.
tion. It is no doubt this fact that made Quine turn Rephrased or expanded, the question becomes,
from rabbits to patterns of stimulation as the key how can I tell whether my beliefs, which are by
to interpretation. Just as a matter of statistics, I'm their nature generally true, are generally true?
not sure how much better one approach is than All beliefs are justified in this sense: they are sup-
the other. Is the relative frequency with which ported by numerous other beliefs (otherwise they
identical patterns of stimulation will touch off wouldn't be the beliefs they are), and have a pre-
assent to "Gavagai" and "Rabbit" greater than the sumption in favor of their truth. The presumption
relative frequency with which a rabbit touches off increases the larger and more significant the body
the same two responses in speaker and interpreter? of beliefs with which a belief coheres, and there
Not an easy question to test in a convincing way. being no such thing as an isolated belief, there is no
But let the imagined results speak for Quine's belief without a presumption in its favor. In this
method. Then I must say, what I must say in any respect, interpreter and interpreted differ. From the
case, the problem of error cannot be met sentence interpreter's point of view, methodology enforces a
by sentence, even at the simplest level. The best general presumption of truth for the body of beliefs
we can do is cope with error holistically, that is, as a whole, but the interpreter does not need to pre-
we interpret so as to make an agent as intelligible sume each particular belief of someone else is true.
as possible, given his actions, his utterances and The general presumption applied to others does
his place in the world. About some things we will not make them globally right, as I have emphasized,
find him wrong, as the necessary cost of finding but provides the background against which to
him elsewhere right. As a rough approximation, accuse them of error. But from each person's own
finding him right means identifying the causes vantage point, there must be a graded presumption
with the objects of his beliefs, giving special weight in favor of each of his own beliefs.
to the simplest cases, and countenancing error We cannot, alas, draw the picturesque and pleas-
where it can be best explained. ant conclusion that all true beliefs constitute knowl-
Suppose I am right that an interpreter must so edge. For though all of a believer's beliefs are to
interpret as to make a speaker or agent largely some extent justified to him, some may not be justi-
correct about the world. How does this help the fied enough, or in the right way, to constitute
person himself who wonders what reason he has knowledge. The general presumption in favor of
to think his beliefs are mostly true? How can he the truth of belief serves to rescue us from a stand-
learn about the causal relations between the real ard form of skepticism by showing why it is impos-
world and his beliefs that lead the interpreter to sible for all our beliefs to be false together. This
interpret him as being on the right track? leaves almost untouched the task of specifying the
The answer is contained in the question. In conditions of knowledge. I have not been concerned
order to doubt or wonder about the provenance with the canons of evidential support (if such there
of his beliefs an agent must know what belief is. be), but to show that all that counts as evidence or
This brings with it the concept of objective truth, justification for a belief must come from the same
for the notion of a belief is the notion of a state totality of belief to which it belongs.


See my "True to the Facts", The Journal of must rest ultimately on sensory evidence." In
Philosophy (1960), pp. 216-34. The Roots of Reference (Illinois: Open Court
2 Hilary Putnam, Meaning and the Moral Publishing Company, 1974), pp. 37-8, Quine
Sciences (London: Routledge and Kegan Paul, says "observations" are basic "both in the sup-
1978), p. 125. port of theory and in the learning of lan-
3 See my "On the Very Idea of a Conceptual guage", and then goes on, "What are
Scheme", in Proceedings and Addresses of the observations? They are visual, auditory, tac-
American Philosophical Association (1974), tual, olfactory. They are sensory, evidently,
pp.5-20. and thus subjective. ... Should we say then
4 Richard Rorty, Philosophy and the Mirror of that the observation is not the sensation ... ?
Nature (Princeton: Princeton University Press, No ... " Quine goes on to abandon talk of
1979),p.178. observations for talk of observation sentences.
5 W. V. Quine, "The Nature of Natural But of course observation sentences, unlike
Knowledge", in S. Guttenplan (ed.), Mind and observations, cannot play the role of evidence
Language (Clarendon Press, Oxford, 1975), unless we have reason to believe they are true.
p.68. 7 I now think it is essential, in doing radical
6 Many other passages in Quine suggest that interpretation, to include the desires of the
Quine hopes to assimilate sensory causes to speaker from the start, so that the springs of
evidence. In Word and Object (Massachusetts: action and intention, namely both belief and
MIT Press, 1960), p. 22, he writes that "surface desire, are related to meaning. But in the
irritations ... exhaust our clues to an external present talk it is not necessary to introduce
world". In Ontological Relativity (New York: this further factor.
Columbia University Press, 1969), p. 75, we 8 It is clear that the causal theory of meaning
find that "The stimulation of his sensory has little in common with the causal theories
receptors is all the evidence anybody has had of reference of Kripke and Putnam. Those
to go on, ultimately, in arriving at his picture theories look to causal relations between
of the world." On the same page: "Two cardi- names and objects of which speakers may well
nal tenets of empiricism remain unassaila- be ignorant. The chance of systematic error is
ble .... One is that whatever evidence there is thus increased. My causal theory does the
for science is sensory evidence. The other ... is reverse by connecting the cause of a belief
that all inculcation of meanings of words, with its object.
A Foundherentist Theory of
Empirical Justification

Susan Haack

Let us remember how common the folly is, of going from one faulty extreme into
the opposite. 1

Does the evidence presented establish beyond a often messy, ambiguous, misleading, inquiry is
reasonable doubt that the defendant did it? Given often untidy, inconclusive, biased by the inquir-
the evidence recently discovered by space scien- ers' interests; but it doesn't follow, as the cynics
tists, am I justified in believing there was once apparently suppose, that standards of good evi-
bacterial life on Mars? Is scientific evidence espe- dence and well-conducted inquiry are local, con-
cially authoritative, and if so, why? Should we ventional, or mythical. And an even half-way
take those advertisements claiming that the adequate understanding of the complexities of
Holocaust never happened seriously, and if not, real-life evidence and the untidiness of real-life
why not? .. Questions about what makes evidence inquiry requires a re-examination of some of
better or worse, about what makes inquiry better those comfortably familiar dichotomies on which
or worse conducted, about disinterestedness and recent epistemology has relied - the logical versus
partiality, are of real, daily - and sometimes of the causal, internalism versus externalism, apri-
life-and-death - consequence. orism versus naturalism, foundationalism versus
Of late, however, cynicism about the very coherentism.
legitimacy of such questions has become the Though the other dichotomies will also come
familiar philosophical theme of a whole chorus of under scrutiny, the main theme here will be that
voices, from enthusiasts of the latest develop- foundational ism and coherentism - the tradi-
ments in neuroscience, to radical self-styled tionally rival theories of justified belief - do not
neo-pragmatists, radical feminists and multicul- exhaust the options, and that an intermediate
turalists, and followers of (by now somewhat theory is more plausible than either. I call it
dated) Paris fashions. "foundherentism:'
This cynicism is unwarranted; but dealing
with it requires something a bit more radical than
epistemological business-as-usual. Evidence is I The Case for Foundherentism

Foundationalist theories of empirical justifica-

Originally published in Louis Pojman (ed.), The Theory tion hold that an empirical belief is justified if
of Knowledge: Classical and Contemporary Readings, and only if it is either a basic belief justified by
2nd edn (Belmont, CA: Wadsworth, 1999), pp. 283-93. the subject's experience,2 or else a derived belief

justified, directly or indirectly, by the support of As the two styles of theory have evolved, with
basic beliefs. Coherentist theories of empirical each party trying to overcome the difficulties the
justification hold that a belief is justified if and other thinks insuperable, they have come closer
only if it belongs to a coherent set of beliefs. In together.
short, foundationalism requires a distinction of Strong foundationalism requires that basic
basic versus derived beliefs and an essentially beliefs be fully justified by the subject's experi-
one-directional notion of evidential support, ence; pure foundationalism requires that derived
while coherentism holds that beliefs can be justi- beliefs be justified exclusively by the support,
fied only by mutual support among themselves. direct or indirect, of basic beliefs. But weak foun-
The merit of foundationalism is that it acknow- dationalism requires only that basic beliefs be
ledges that a person's experience - what he sees, justified to some degree by experience; and
hears, etc. - is relevant to how justified he is in his impure foundationalism, though requiring all
beliefs about the world; its drawback is that it derived beliefs to get some support from basic
requires a privileged class of basic beliefs justified beliefs, allows mutual support among derived
by experience alone but capable of supporting the beliefs to raise their degree of justification.
rest of our justified beliefs, and ignores the perva- Uncompromisingly egalitarian forms of
sive interdependence among a person's beliefs. coherentism hold that only overall coherence
The merit of coherentism is that it acknowledges matters, so that every belief in a coherent set is
that pervasive interdependence, and requires no equally justified. But moderated, inegalitarian
distinction of basic and derived beliefs; its draw- forms of coherentism give a subject's beliefs about
back is that it allows no role for the subject's expe- his present experience a distinguished initial
rience. status, or give a special standing to beliefs which
Foundationlists, naturally, are keenly aware of are spontaneous rather than inferential in origin.
the problems with coherentism. How could one In a way, these moderated forms of founda-
possibly be justified in believing there's a dog in tionalism and coherentism lean in the right direc-
the yard, they ask, if what one sees, hears, smells, tion. But the leaning destabilizes them.
etc., plays no role? And isn't the coherentist's talk Weak foundationalism concedes that basic
of mutual support among beliefs just a euphe- beliefs need not be fully justified by experience
mism for what is really a vicious circle in which alone; but then what reason remains to deny that
what supposedly justifies the belief that p is the they could get more (or less) justified by virtue of
belief that q, and what justifies the belief that q their relations to other beliefs? Impure founda-
the belief that r .. . and what justifies the belief that tionalism concedes that there can be mutual sup-
z is the belief that p? port among derived beliefs; but then what reason
Coherentists, naturally, are no less keenly aware remains to insist that more pervasive mutual sup-
of the problems with foundationalism. What sense port is unacceptable? And weak, impure founda-
does it make to suppose that someone could have tionalism allows both that basic beliefs are less
a justified belief that there's a dog in the yard, they than fully justified by experience, and that derived
ask, except in the context of the rest of his beliefs beliefs may be mutually supportive; but now the
about dogs, etc.? Besides, why should we suppose insistence that derived beliefs can give no support
that there are any beliefs both justified by experi- to basic beliefs looks arbitrary, and the distinction
ence alone and capable of supporting the rest of of basic and derived beliefs pointless. 3
our justified beliefs? After all, foundationalists Moderated, inegalitarian coherentism con-
can't even agree among themselves whether the cedes that some beliefs are distinguished by their
basic beliefs are about observable physical objects, perceptual content or "spontaneous" origin; but
along the lines of "there's a dog;' or are about the isn't this implicitly to concede that justification is
subject's experience, along the lines of "it now not after all a relation exclusively among beliefs,
seems to me that I see what looks like a dog" or "I that input from experience is essential?
am appeared to brownly." And anyway, only prop- Not surprisingly, these fancier forms of
ositions, not events, can stand in logical relations foundationalism and compromising kinds of
to other propositions; so how could a subject's coherent ism, though more sophisticated than
experience justify those supposedly basic beliefs? their simpler ancestors, tend to be ambiguous

and unstable. On the foundationalist side, for difference between legitimate mutual support
example, under pressure of just the kinds of dif- and vicious circularity - my version will rely on
ficulty my analysis identifies, C. 1. Lewis moves an analogy between the structure of evidence and
from a pure to an impure foundationalism and a crossword puzzle.
then, briefly, to a kind of proto-foundherentism. 4 Of course, the viability of the foundherentist
And on the coherentist side, under pressure of approach doesn't depend on my being completely
just the kind of difficulty my analysis identifies, successful in articulating it. No doubt there could
BonJour tries to guarantee experiential input by be other versions of foundherentism falling
adding an "Observation Requirement" - which, within these general contours but differing in
however, is ambiguous; on one interpretation it is their details.
genuinely coherentist, but doesn't allow the rele- I take as my starting point the following vague,
vance of experience, and on the other it allows the but very plausible, formulation: ''A is morelless
relevance of experience, but isn't genuinely co her- justified, at t, in believing that p, depending on
entist. 5 (BonJour now acknowledges that, after how good his evidence is."
all, coherentism won't do.)6 By starting from here I take for granted, first,
Neither of the traditionally rival theories can that justification comes in degrees: a person may
be made satisfactory without sacrificing its dis- be more or less justified in believing something.
tinctive character. The obvious conclusion - (I also assume that a person may be more justified
though those still wedded to the old dichotomy in believing some things than he is in believing
will doubtless continue to resist it - is that we others.)
need a new approach which allows the relevance I also take for granted, second, that the con-
of experience to empirical justification, but with- cepts of evidence and justification are internally
out postulating any privileged class of basic connected: how justified a person is in believing
beliefs or requiring that relations of support be something depends on the quality of his evidence
essentially one-directional: in other words, a with respect to that belief.
foundherentist theory. I assume, third, that justification is personal: one
person may be more justified in believing some-
thing than another is in believing the same thing -
II Explication of Foundherentism because one person's evidence may be better than
another's. (But though justification is personal, it
The details get complicated, but the main ideas is not subjective. How justified A is in believing that
are simple. p depends on how good his, Xs, evidence is. But
A foundherentist account will acknowledge how justified A is in believing that p doesn't depend
(like foundationalism) that how justified a person on how good A thinks his evidence is; and anyone
is in an empirical belief must depend in part on who believed the same thing on the same evidence
his experience - my version will give a role both would be justified to the same degree.)
to sensory experience, and to introspective aware- And I assume, fourth, that justification is
ness of one's own mental states. As coherentists relative to a time: a person may be more justified
point out, though experience can stand in causal in believing something at one time than at
relations to beliefs, it can't stand in logical relations another - because his evidence at one time may
to propositions. But what this shows is not that be better than his evidence at another.
experience is irrelevant to empirical justification,
but that justification is a double-aspect concept, ''A is more/less justified, at t, in believing that p,
partly causal as well as partly logical in character. depending on how good his evidence is:' The main
A foundherentist account will acknowledge tasks, obviously, are to explain "his evidence" and
(like coherentism) that there is pervasive mutual "how good." The double-aspect character of the
support among a person's justified beliefs. As concept of justification is already in play; for "his:'
foundationalists point out, a belief can't be justi- in "his evidence:' is a causal notion, while "how
fied by a vicious circle of reasons. But what this good" is logical, or quasi-logical, in character.
shows is not that mutual support is illegitimate, The concept of justification is causal as well as
but that we need a better understanding of the logical across the board 7 - its causal aspect is not

restricted to experiential evidence alone. Quite In this vector of forces [the causal nexus of Xs
generally, how justified someone is in believing S-belief that p], besides Xs present experience
something depends not only on what he believes, and present memory traces of his past experience,
but on why he believes it. For example: if two and other S-beliefs of his, such factors as his
people both believe the accused is innocent, one wishes, hopes, and fears will often playa role. But
because he has evidence that she was a hundred Xs desire not to believe ill of his students, say, or
miles from the scene of the crime at the relevant his being under the influence of alcohol, though
time, the other because he thinks she has an they may affect whether or with what degree of
honest face, the former is more justified than the confidence he believes that Grabit stole the book,
latter. In short, degree of justification depends on aren't themselves part of his evidence with respect
the quality of the evidence that actually causes the to that proposition.
belief in question. So "Xs S-evidence with respect to p" will refer
The word "belief" is ambiguous: sometimes it to those experiential and belief-states of Xs which
refers to a mental state, someone's believing belong, at the time in question, to the causal nexus
something [an S-beliefl;8 sometimes it refers to of Xs S-belief that p. The phrase "with respect to"
the content of what is believed, a proposition signals the inclusion of both positive, sustaining,
[a C-beliefl. "Xs evidence" needs to be tied some- and negative, inhibiting, evidence [respectively,
how to what causes Xs S-belief, but must also be Xs S-evidence for p, and Xs S-evidence against pl.
capable of standing in logical or quasi-logical Xs S-evidence with respect to p will include other
relations to the C-belief, the proposition believed. beliefs of his [Xs S-reasons with respect to p l; and
The idea is to begin by characterizing Xs S- his perceptions, his introspective awareness of his
evidence with respect to p - this will be a set of own mental goings-on, and memory traces of his
states of A causally related to his S-belief that p; earlier perceptual and introspective states [Xs
and then to use this as the starting point of a experiential S-evidence with respect to pl.
characterization of Xs C-evidence with respect to The part about memory needs amplifying. Xs
p - this will be a set of propositions capable of experiential S-evidence may include present
standing in logical or quasi-logical relations to memory traces of past experience - such as his
the C-belief that p. remembering seeing his car-keys on the dresser.
If A initially came to believe that the rock- This corresponds to the way we talk of Xs
rabbit is the closest surviving relative of the ele- remembering seeing, hearing, reading, etc. We
phant because a fellow-tourist told him he read also talk of 1\s remembering that p, meaning that
this somewhere, and later still believes it, but now he earlier came to believe that p and has not for-
because he has learned all the relevant biological gotten it. How justified A is in such persisting
details, he is more justified at the later time than at beliefs will depend on how good his evidence
the earlier. So, if they are different, "Xs S-evidence is - his evidence at the time in question, that is.
with respect to p" should relate to the causes of A person's evidence for persisting beliefs will
Xs S-belief that p at the time in question rather normally include memory traces of past percep-
than to what prompted it in the first place. tual experience; my belief that my high-school
What goes on in people's heads is very com- English teacher's name was "Miss Wright," for
plicated. There will likely be some factors inclin- instance, is now sustained by my remembering
ing A towards believing that p, and others pulling hearing and seeing the name used by myself and
against it. Perhaps, e.g., A believes that Tom others.
Grabit stole the book because his seeing Grabit Testimonial evidence, in a broad sense - what
leave the library with a shifty expression and a person reads, what others tell him - enters the
a suspicious bulge under his sweater exerts a picture byway of his hearing or seeing, or remem-
stronger positive pull than his belief that it is bering hearing or seeing, what someone else says
possible that Tom Grabit has a light-fingered or writes. Of course, Xs hearing B say that p won't
identical twin exerts in the opposite direction. contribute to his, Xs, believing that p, unless A
Both sustaining and inhibiting factors are rele- understands B's language. But if A believes that p
vant to degree of justification, so both will be in part because B told him that p, how justified
included in Xs S-evidence. A is in believing that p will depend in part on how

justified A is in thinking B honest and reliable. experiential C-evidence in terms of propositions

But I anticipate. to the effect that A is in the sort of perceptual state
a normal subject would be in when seeing this or
J\s S-evidence with respect to p is a set of states of that in these or those circumstances. For example,
A causally related to his S-belief that p. But in the if J\s experiential S-evidence with respect to p is
part of the theory that explains what makes evi- his perceptual state, its looking to him as it would
dence better or worse, "evidence" will have to to a normal observer seeing a female cardinal bird
mean "C-evidence:' and refer to a set of proposi- at a distance of forty feet in poor light, the corre-
tions. The two aspects interlock: J\s C-evidence sponding experiential C-evidence will be a propo-
with respect to p will be a set of propositions, and sition to the effect that A is in the kind of perceptual
how good it is will depend on those propositions' state a normal observer would be in when looking
logical or quasi-logical relations to p; but which at a female cardinal bird in those circumstances.
propositions J\s C-evidence with respect to p Built into my account of experiential evidence
consists of depends on which of J\s S-beliefs and is a conception of perception as, in a certain sense,
perceptual, etc., states belong to the causal nexus direct. This is not to deny that perception involves
of the S-belief in question. complicated neurophysiological goings-on. Nor
A's C-reasons with respect to p, obviously is it to deny that the judgments causally sustained
enough, should be the C-beliefs, i.e., the proposi- by the subject's experience are interpretative, that
tions, which are the contents of his S-reasons. For they depend on his background beliefs as well -
example, if one of J\s S-reasons with respect to p which, on the contrary, is a key foundherentist
is his S-belief that female cardinal birds are brown, thought. It is only to assert that in normal percep-
the corresponding C-reason will be the proposi- tion we interact with physical things and events
tion that female cardinal birds are brown. around us, which look a certain way to all normal
But what about J\s experiential C-evidence? My observers under the same circumstances.
proposal is that "J\s experiential C-evidence with You may be wondering why I include the sub-
respect to p" refers to propositions to the effect that ject's sensory and introspective experience as evi-
A is in the perceptual/introspective/memory states deIKe, but not, say, his extra-sensory perceptual
which constitute his experiential S-evidence with experience. Well, the task here is descriptive - to
respect to p. Since a perceptual, etc., state can't be articulate explicitly what is implicit when we say that
part of the causal nexus of J\s S-belief that p unless A has excellent reasons for believing that p, that B is
A is in that state, these propositions are all true. But guilty of wishful thinking, that C has jumped to an
they need not be propositions that A believes." unjustified conclusion, and so on. As those phrases
So J\s experiential C-evidence has a distinctive "excellent reasons" and "guilty of wishful thinking"
status. J\s C-reasons may be true or may be false, indicate, his other belief, should be included as part
and A may be more or less justified, or not justi- of a subject's evidence, but his wishes should not.
fied at all, in believing them. But J\s experiential Actually, I think it most unlikely there is such a thing
C-evidence consists of propositions all of which as ESP; but it is excluded because - unlike sensory
are, ex hypothesi, true, and with respect to which experience, for which we even have the phrase, "the
the question of justification doesn't arise. (This is evidence of the senses" - it has no role in the implicit
the foundherentist way of acknowledging that the conception of evidence I am trying to make explicit.
ultimate evidence for empirical beliefs is experi- The concepts of better and worse evidence, of
ence - very different from the forced and unnatu- more and less justified belief, are evaluative; so,
ral way in which foundationalism tries to after the descriptive task of explication, there will
acknowledge it, by requiring basic beliefs justified be the ratificatory question, whether our stand-
by experience alone.) ards of better and worse evidence really are, as we
In line with the way we ordinarily talk about the hope and believe they are, indicative of truth. But
evidence of the senses - "Why do I think there's a that comes later.
cardinal in the oak tree? Well, I can see the thing;
that distinctive profile is clear, though the light's The present task is to explicate "how good" in
not too good, and it's quite far away, so I can't really "how good J\s C-evidence is." What factors raise,
see the color" - I suggest a characterization of J\s and what lower, degree of justification?

Foundationalists often think of the structure Think, for example, of a detective whose evidence
of evidence on the model of a mathematical is: the murder was committed by a left-handed
proof - a model which, understandably, makes person; either Smith or Brown did it; Smith is
them leery of the idea of mutual support. My right-handed; Brown is right-handed. Though
approach will be informed by the analogy of a this deductively implies that Smith did it, it cer-
crossword puzzle - where, undeniably, there is tainly isn't conclusive evidence for that belief (let
pervasive mutual support among entries, but, alone conclusive evidence for the belief that Smith
equally undeniably, no vicious circle. The clues did it and conclusive evidence for the belief that
are the analogue of experiential evidence, already- Brown did it and conclusive evidence for the
completed intersecting entries the analogue of belief that extra-terrestrials did it!).
reasons. As how reasonable a crossword entry is Deductive implication is necessary but not
depends both on the clues and on other intersect- sufficient for conclusiveness. Evidence E is conclu-
ing entries, the idea is, so how justified an empiri- sive for p just in case the result of adding p to E [the
cal belief is depends on experiential evidence and p-extrapolation of E] is consistent, and the result
reasons working together. of adding not-p to E [the not-p-extrapolation
Perhaps needless to say, an analogy is only an of E] is inconsistent. E is conclusive against p
analogy, not an argument. Its role is only to sug- just in case its p-extrapolation is inconsistent and
gest ideas, which then have to stand on their own its not-p-extrapolation consistent. But if E itself
feet. And there are always disanalogies; there will is inconsistent, both its p-extrapolation and its
be nothing in my theory analogous to the solu- not-p-extrapolation are also inconsistent, so E is
tion to to day's crossword which appears in tomor- indifferent with respect to p.
row's newspaper, for instance, nor any analogue Often, though, evidence is not conclusive
of the designer of a crossword. either way, nor yet inconsistent and hence indif-
But the analogy does suggest a very plausible ferent, but supports the belief in question, or its
multi-dimensional answer to the question, what negation, to some degree. Suppose the detective's
makes a belief more or less justified? How reason- evidence is: the murder was committed by a left-
able a crossword entry is depends on how well handed person; either Smith or Brown did it;
it is supported by the clue and any already- Smith is left-handed; Brown is left-handed; Smith
completed intersecting entries; how reasonable recently saw the victim, Mrs Smith, in a romantic
those other entries are, independent of the entry restaurant holding hands with Brown. Though
in question; and how much of the crossword has not conclusive, this evidence is supportive to
been completed. How justified A is in believing some degree of the belief that Smith did it - for, if
that p, analogously, depends on how well the he did, we have some explanation of why.
belief in question is supported by his experiential The example suggests that supportiveness
evidence and reasons [supportiveness]; how jus- depends on whether and how much adding p to E
tified his reasons are, independent of the belief in makes a better explanatory story. But a better
question [independent security]; and how much explanatory story than what? Conclusiveness is a
of the relevant evidence his evidence includes matter of the superiority of p over its negation
[comprehensiveness]. with respect to consistency. But if p is potentially
On the first dimension, 1\s C-evidence may be explanatory of E or some component of E, it is
conclusive for p, conclusive against p, supportive- not to be expected that not-p will be too. So I con-
but-not-conclusive of p, undermining-but-not- strue supportiveness as depending on the superi-
conclusive against p, or indifferent with respect to ority ofp over its rivals with respect to explanatory
pi with respect to not-po integration; where a rival of p is any proposition
Foundationalists often take for granted that adding which to E improves its explanatory inte-
evidence is conclusive just in case it deductively gration to some degree, and which, given E, is
implies the proposition in question; but this isn't incompatible with p.
quite right. Inconsistent premisses deductively The word "integration" was chosen to indicate
imply any proposition whatever; but inconsistent that E may support p either because p explains E
evidence isn't conclusive evidence for anything - or some component of E, or vice versa - that there
let alone conclusive evidence for everything! is "mutual reinforcement between an explanation

and what it explains."IO (So the concept of explan- believing that p may depend in part on how justi-
atory integration is closer kin to the coherentist fied he is in believing that q - independent of the
concept of explanatory coherence than to the support given his belief that q by his belief that p.
foundationalist concept of inference to the best And, though "justified" appears on the right-
explanation. ) hand side of the independent security clause,
Usually, as conclusiveness of evidence is taken there is no danger of an infinite regress - any
to be the province of deductive logic, supportive- more than with a crossword puzzle. As in the
ness of evidence is taken to be the province of case of a crossword eventually we reach the clues,
inductive logic. But at least if "logic" is taken in its so with empirical justification eventually we
now-usual narrow sense, as depending on form reach experiential evidence. And experiential
alone, this looks to be a mistake. Explanation C-evidence does not consist of other C-beliefs of
requires generality, kinds, laws - a motive for the the subject, but of propositions all of which are,
murder, a mechanism whereby smoking causes ex hypothesi, true, and with respect to which the
cancer, and so forth. If so, explanatoriness, and question of justification doesn't arise. This is not
hence supportiveness, requires a vocabulary to deny that, as crossword clues may be cryptic,
which classifies things into real kinds; and hence experiential evidence may be ambiguous or mis-
depends on content, not on form alone. (Hempel leading; on the contrary, my account of experien-
drew the moral, many years ago now, from the tial C-evidence is intended to recognize that it
"grue" paradox. lI ) But there is supportive-but- often is. It is only to say that the question of justi-
not -conclusive evidence, even if there is no formal fication arises with respect to a person's beliefs,
inductive logic. but not with respect to his experiences.
Supportiveness alone does not determine As how reasonable a crossword entry is
degree of justification, which also depends on depends not only on how well it is supported by
independent security and comprehensiveness. the clue and other intersecting entries, and on
Suppose our detective's evidence is: the murder how reasonable those other entries are, but also
was committed by a left-handed person; either on how much of the crossword has been com-
Smith or Brown did it; Smith is right-handed, but pleted, so degree of justification depends not only
Brown left-handed. The detective's evidence is on supportiveness and independent security, but
conclusive that Brown did it; nevertheless, he is also on comprehensiveness - on how much of the
not well-justified in believing this unless, among relevant evidence the subject's evidence includes.
other things, he is justified in believing that the Comprehensiveness promises to be even
murder was committed by a left-handed person, tougher to spell out than supportiveness and
that either Smith or Brown did it, etc. independent security; the crossword analogy isn't
The idea of independent security is easiest to much help here, and neither is the nearest ana-
grasp in the context of the crossword analogy. In logue in the literature, the total evidence require-
a crossword, how reasonable an entry is depends ment on inductions, which refers, not to the
in part on its fit with intersecting entries, and totality of relevant evidence, but to the totality of
hence on how reasonable those entries are, inde- relevant available evidence - and then there is the
pendently of the entry in question. Similarly, how further problem that relevance itself comes in
justified a person is in believing something degrees.
depends in part on how well it is supported by his I am assuming, however, that (degree of) rel-
other beliefs, and hence on how justified he is in evance is an objective matter. Naturally, whether I
believing those reasons, independently of the think your handwriting is relevant to your trust-
belief in question. worthiness depends on whether I believe in
It is that last phrase - in my theory as with a graphology; but whether it is relevant depends on
crossword puzzle - that averts the danger of a whether graphology is true.
vicious circle. The reasonableness of the entry for 3 As this reveals, though relevance, and hence
down may depend in part on the reasonableness of comprehensiveness, is objective, judgments of
the intersecting entry for 5 across - independent of relevance, and hence judgments of comprehen-
the support given to the entry for 5 across by the siveness, are perspectival, i.e., they depend on the
entry for 3 down. Similarly, how justified A is in background beliefs of the person making them.

The same goes for judgments of supportiveness Xs particular business to know whether p, and
and independent security. How supportive you or how important it is to be right about whether p;
I judge E to be with respect to p, for example, will perhaps it also runs together strictly epistemo-
depend on what rivals of p we happen to be able logical with ethical concerns. This vague concept
to think of; but how supportive E is of p does not. [complete justification] is useful for practical pur-
Quality of evidence is objective, but judgments of poses - and for the statement of Gettier-type
quality of evidence are perspectival. paradoxes. In other philosophical contexts, how-
Because quality of evidence is multi-dimen- ever, "A is completely justified in believing that p"
sional, we should not necessarily expect a linear is used in a context-neutralized, optimizing way,
ordering of degrees of justification; e.g., Xs evi- requiring conclusiveness, maximal independent
dence with respect to p might be strongly sup- security, and full comprehensiveness of evidence
portive but weak on comprehensiveness, while [COMPLETE justification].
his evidence with respect to q might be strong on The account sketched here has been personal,
comprehensiveness but only weakly supportive. i.e., focussed firmly on our friend A. But this is
Nor, a fortiori, does it look realistic to aspire to not to deny that in even the most ordinary of our
anything as ambitious as a numerical scale of everyday beliefs we rely extensively on testimonial
degrees of justification. But something can be evidence. And where the sciences are concerned,
said about what is required for A to be justified to reliance on others' evidence - and hence on the
any degree in believing that p. interpretation of others' words and judgments of
One necessary condition is that there be such others' reliability - is absolutely pervasive. (This
a thing as Xs C-evidence with respect to p. If Xs reveals that not only the social sciences but also
S-belief that p is caused simply by a blow to the the natural sciences presuppose the possibility of
head, or by one of those belief-inducing pills phi- interpreting others' utterances: think, e.g., of an
10sophers are fond of imagining, A isn't justified astronomer's reliance on others' reports of obser-
to any degree in believing that p. Since it is the vations.)
justification of empirical beliefs that is at issue, Anyhow, thinking about evidence in the sci-
another necessary condition is that Xs C-evidence ences prompts me to ask whether it is possible to
should include some experiential C-evidence - extrapolate from my account of "A is more/less
present experiential evidence, or memory traces justified in believing that p" to a concept of justi-
of what he earlier saw, heard, read, etc. This is my fication applicable to groups of people. It might
analogue of BonJour's Observation Requirement, be feasible to do this by starting with the degree
obviously much more at home in foundherent- of justification of a hypothetical subject whose
ism than his requirement was in his coherentist evidence includes all the evidence of each member
theory. (It is not meant to rule out the possibility of the group, and then discount this by some
that some of a person's beliefs may not be sus- measure of the degree to which each member of
tained directly by experiential evidence, not even the group is justified in believing that other mem-
by memory traces, but rely on other beliefs and bers are competent and honest.
their experiential evidence - as in an unconven-
tional crossword some entries might have no
clues of their own but rely on other entries and III The Ratification of Foundherentism
their cluesY) A third necessary condition is that
Xs C-evidence with respect to p should meet Thus far the task has been to articulate our stand-
minimal conditions of supportiveness, independ- ards of better and worse evidence, of more and
ent security, and comprehensiveness; e.g., it less justified belief. But what do I mean by "our"?
should be better than indifferent in terms of sup- And what assurance can I give that a belief's being
portiveness. Jointly, these necessary conditions justified, by those standards, is any indication that
look to be sufficient. it is true?
What about the upper end of the scale? Our When I speak of "our" standards of better and
ordinary use of phrases like "A is completely justi- worse evidence, I emphatically do not mean to
fied in believing that p" is vague and context- suggest that these standards are local or paro-
dependent, depending inter alia on whether it is chial, accepted in "our;' as opposed to "their,"

community. Rather, I see these standards - essen- of standards does not, in and of itself, imply rela-
tially, how well a belief is anchored in experience tivity of standards. 13 So those epistemic relativists
and how tightly it is woven into an explanatory who have inferred that, since judgments of justi-
mesh of beliefs - as rooted in human nature, in fication vary from community to community,
the cognitive capacities and limitations of all there can be no objectively correct standards of
normal human beings. better and worse evidence, have committed a non
It is sure to be objected that the evidential sequitur as well as relying on a dubious premiss.
standards of different times, cultures, communi- As for those who have succumbed to epistemic
ties, or scientific paradigms differ radically. But relativism because they have given up on the con-
I think this supposed variability is at least an cept of truth, I have room here only to say that
exaggeration, and quite possibly altogether an theirs seems to me an entirely factitious despair.14
illusion, the result of mistaking the perspectival In any case, all that will be required of the concept
character of judgments of evidential quality for of truth in what follows is that a proposition or
radical divergence in standards of better and statement is true just in case things are as it says.
worse evidence. Supposing - as I believe, and so do you - that
Because judgments of the quality of evidence we humans are fallible, limited but inquiring
are perspectival, people with radically different creatures who live in a world which is largely
background beliefs can be expected to differ sig- independent of us and what we believe about it,
nificantly in their judgments of degree of justifi- but in which there are kinds, laws, regularities;
cation. It doesn't follow that there are no shared and supposing - as I believe, and so do you - that
standards of evidence. If we think of the con- our senses are a source, though by no means an
straints of experiential anchoring and explana- infallible source, of information about things
tory integration rather than of specific judgments and events in the world around us, and intro-
of the relevance, supportiveness, etc., of this or spection a source, though by no means an infal-
that evidence, I believe we will find commonality lible source, of information about our own
rather than divergence. mental goings-on; then, if any indication of how
Again, the point is easier to see in the context things are is possible for us, how well our beliefs
of the crossword analogy. Suppose you and I are are anchored in our experience and knit into an
both doing the same crossword puzzle, and have explanatory mesh is such an indication. (And
filled in some long central entry differently. You supposing - as I believe, and so, probably, do
think, given your solution to that long central you - we have no other sourc~s of information
entry, that the fact that 14 down ends in a "T" is about the world and ourselves, no ESP or clair-
evidence in its favor; I think, given my solution to voyance or etc., then this is the only indication
that long central entry, that the fact that it ends in we can have of how things are.)
a "D" is evidence in its favor. Nevertheless, we are That last paragraph was nothing like an a
both trying to fit the entry to its clue and to other priori ratification of foundherentism; for those
already-completed entries. Now suppose you and "supposing" clauses are empirical in character.
I are both on an appointments committee. You Assumptions about human cognitive capacities
think the way this candidate writes his "g"s indi- and limitations are built into our standards of evi-
cates that he is not to be trusted; I think graphol- dential quality; so the truth-indicativeness of
ogy is bunk and scoff at your "evidence:' Because those standards depends on the truth of those
of a disagreement in background beliefs, we disa- empirical assumptions. But neither was that last
gree about what evidence is relevant. Nevertheless, paragraph much like the appeals to psychology or
we are both trying to assess the supportiveness, cognitive science on which some epistemological
independent security, and comprehensiveness of naturalists of a more extreme stripe than mine
the evidence with respect to the proposition that propose to rely; for the assumptions referred to in
the candidate is trustworthy. my "supposing" clauses, though empirical, are of
But even if I am wrong about this, even if such generality as to be rather philosophical than
there really are radically divergent standards of scientific in character.
evidential quality, it wouldn't follow that there Those assumptions would surely be presup-
are no objective indications of truth; variability posed by any conceivable scientific experiment.

But they are well integrated with what the sci- external world at all? After pointing out that
ences of cognition have to tell us about the mech- since, ex hypothesi, his machinations would be
anisms of perception and introspection, and of absolutely undetectable, if there were an Evil
when and why they are more or less reliable, and Demon no truth-indication would be possible
with what the theory of evolution suggests about for us - only that my claim is a conditional one:
how we came to have the sort of information- that, if any truth-indication is possible for us, the
detecting apparatus we do. As one would hope, foundherentist criteria are truth-indicative. (I
the epistemological part of my crossword - the could discharge the antecedent, and arrive at a
part where the entries are themselves about cross- categorical conclusion, by adopting a definition
words - interlocks snugly with other parts. of truth along Peircean lines, as the opinion that
But what am I to say to those readers familiar would survive all possible experiential evidence
with Descartes' failed attempt to prove "what I and the fullest logical scrutiny; but I prefer the
clearly and distinctly perceive is true," who are more cautious, and more realist, strategy.)
bound to suspect that I must be arguing in a circle? Determined skeptics won't be persuaded; but
After pointing out that I have not offered a ratifi- determined skeptics never are! And the rest of
catory argument in which some premiss turns out you may notice that foundherentism enables us
to be identical with the conclusion, nor an argu- to sidestep another dichotomy which has - if
ment relying on a certain mode of inference to you'll pardon the pun - bedeviled recent episte-
arrive at the conclusion that this very mode of mology: either a hopeless obsession with hyper-
inference is a good one - only that, to borrow bolic skepticism, or a hopeless relativism or
Peirce's words, by now "the reader will, I trust, be tribalism preoccupied with "our (local, parochial)
too well-grounded in logic to mistake mutual epistemic practices." Foundherentism, I believe,
support for a vicious circle of reasoning:'ls provides a more realistic picture of our epistemic
And what am I to say to readers worried condition - a robustly fallibilist picture which,
about the Evil Demon, who are bound to object without sacrificing objectivity, acknowledges
that I have not ruled out the possibility that our something of how complex and confusing evi-
senses are not a source of information about the dence can be.


This brief statement of foundherentism is 3 (1996), pp. 611-57, and from the debate with
based primarily on my Evidence and Inquiry: BonJour in Synthese 112.1 (July 1997), pp. 7-35.
Towards Reconstruction in Epistemology (Oxford: Thomas Reid, Essays on the Intellectual Powers
Blackwell, 1993), especially chapters 1, 4, and (l785), in R. E. Beanblossom and K. Lehrer
lO. I have also drawn on material from earlier (eds), Thomas Reid: Inquiry and Essays,
articles of mine, especially "Theories of Know- (Indianapolis, IN: Hackett, 1983), vol. VI, p. 4.
ledge: an Analytic Framework," Proceedings of the 2 I restrict my attention here to experientialist
Aristotelian Society LXXXIII (1982-3), pp. 143-57 forms of foundationalism, ignoring, e.g., foun-
(where foundherentism was first introduced), dationalist theories of a priori knowledge.
"c. I. Lewis," in American Philosophy, ed. Marcus 3 My characterization of foundationalism is
Singer, Royal Institute of Philosophy Lecture quite standard; cf. for example, Alston's in
Series, 19 (Cambridge: Cambridge University J. Dancy and E. Sosa (eds), Companion to
Press, 1985), pp. 215-39, and "Rebuilding the Ship Epistemology (Oxford: Blackwell, 1992),
While Sailing on the Water," in R. Barrett and p. 144, or Sosa's in "The Raft and the Pyramid,"
R. Gibson (eds), Perspectives on Quine (Oxford: this vol., ch. 13. But matters have been con-
Blackwell, 1990), pp. 111-27 (where some of the fused because, in "Can Empirical Knowledge
key ideas of foundherentism were developed). Have a Foundation?", this vol., ch. lO, and The
I have drawn as well on material from the sympo- Structure ofEmpirical Knowledge (Cambridge,
sium on Evidence and Inquiry published in MA: Harvard University Press, 1986), p. 28,
Philosophy and Phenomenological Research LVI. BonJour uses "weak foundationalism" to refer

a style of theory which is both weak and (Indianapolis, IN: Bobbs-Merrill, 2nd edn,
impure, in my sense, and in addition allows 1965), pp. 59-83; C. G. Hempel, "Postscript
mutual support among basic beliefs and - on Confirmation" (1964), in Aspects of
apparently - allows "basic" beliefs to get sup- Scientific Explanation (New York: Free Press,
port from "derived" beliefs. As my scare quotes 1965), pp. 47-52.
indicate, once one-directionality has been so 12 In case a desperate foundationalist is tempted
completely abandoned it is unclear that the to try seizing on this in hopes of salvaging
theory really qualifies as foundationalist at all; the derived/basic distinction, let me point
certainly the basic!derived distinction has out that beliefs without direct experiential
become purely proforma. See also Haack, evidence could contribute to the support of
"Reply to BonJour:' Synthese 112.1 (July beliefs with direct experiential evidence; and
1997), pp. 25-35. that this maneuver would identify no plausi-
4 See Evidence and Inquiry, ch. 2, for details. ble kind of belief as basic! as derived - think,
5 See ibid., ch. 3, for details. e.g., of a scientist whose belief that electrons
6 Laurence BonJour, "Haack on Justification are composed thus and so is sustained by
and Experience:' Synthese 112.1 (July 1997), what he sees in the bubble chamber.
pp.13-15. 13 See also Susan Haack, "Reflections on
7 An idea I first began to work out in Relativism: From Momentous Tautology to
"Epistemology With a Knowing Subject:' Seductive Contradiction," NoCts Supplement
Review of Metaphysics XXXIII.2 (1979), pp. (1996), pp. 297-315, and in James
309-36. E. Tomberlin (ed.), Philosophical Perspectives,
8 Expressions introduced in square brackets 10: Metaphysics (Oxford: Blackwell, 1996),
are my new, technical terms, or special, tech- pp. 297-315; reprinted in Haack, Manifesto
nical uses of familiar terms. of a Passionate Moderate: Unfashionable
9 So my theory is not straightforwardly exter- Essays (Chicago: University of Chicago Press,
nalist, since Ns S-evidence must consist of 1998), pp. 149-66.
states of A - states, furthermore, of which A 14 I have more to say in "Confessions of an
can be aware; but neither is it straightfor- Old-Fashioned Prig," in Manifesto of a
wardly internalist, since Ns experiential C- Passionate Moderate, pp. 7-30.
evidence consists of propositions A need not 15 C. S. Peirce, Collected Papers, eds C. Hartshorne,
believe or even conceive. P. Weiss, and A. Burks (Cambridge, MA:
10 W. V. Quine and J. Ullian, The Web of Belief Harvard University Press, 1931-58),6.315.
(New York: Random House, 1970), p. 79.
11 N. Goodman, "The New Riddle ofInduction"
(1953), in Fact, Fiction and Forecast
The Raft and the Pyramid

Ernest So sa

Contemporary epistemology must choose and, second, that the relevant sort of
between the solid security of the ancient founda- justification is that which pertains to
tionalist pyramid and the risky adventure of the knowledge: epistemic (or theoretical) justi-
new coherentist raft. Our main objective will be fication. Someone seriously ill may have
to understand, as deeply as we can, the nature of two sorts of justification for believing he
the controversy and the reasons for and against will recover: the practical justification that
each of the two options. But first of all we take derives from the contribution such belief
note of two underlying assumptions. will make to his recovery and the theoreti-
cal justification provided by the lab results,
the doctor's diagnosis and prognosis, and
1 Two Assumptions so on. Only the latter is relevant to the
question whether he knows.
(Al) Not everything believed is known, but
nothing can be known without being at
least believed (or accepted, presumed, 2 Knowledge and Criteria
taken for granted, or the like) in some
broad sense. What additional requirements a. There are two key questions of the theory of
must a belief fill in order to be knowledge? knowledge:
There are surely at least the following two: (i) What do we know?
(a) it must be true, and (b) it must be justi- (ii) How do we know?
fied (or warranted, reasonable, correct, or The answer to the first would be a list of bits
the like). of knowledge or at least of types of knowl-
(A2) Let us assume, moreover, with respect to edge: of the self, of the external world, of
the second condition Al(b): first, that it other minds, and so on. An answer to the
involves a normative or evaluative property; second would give criteria (or canons, meth-
ods, principles, or the like) that would explain
how we know whatever it is that we do know.
Originally published in Midwest Studies in Philosophy,
Vol. 5: Studies in Epistemology (Minneapolis: University b. In developing a theory of knowledge, we can
of Minnesota Press, 1980), pp. 3-25; an appendix to begin either with a(i) or with a(ii). Particular-
this paper is drawn from Ernest Sosa, "How Do You ism would have us begin with an answer to
Know?" American Philosophical Quarterly 11 (1974), a(i) and only then take up a(ii) on the basis of
pp.113-22. that answer. Quite to the contrary, methodism

would reverse that order. The particularist plumbed by Freud, nor even to Chomsky's. Nor
thus tends to be antiskeptical on principle. need we recall the labyrinths inhabited by states-
But the methodist is as such equally receptive men and diplomats, nor the rich patterns of some
to skepticism and to the contrary. Hume, for novels or theories. We need look no further than
example, was no less a methodist than the most common, everyday beliefs. Take, for
Descartes. Each accepted, in effect, that only instance, the belief that driving tonight will be dan-
the obvious and what is proved deductively gerous. Brief reflection should reveal that any of us
on its basis can possibly be known. with that belief will join to it several other closely
c. What, then, is the obvious? For Descartes it is related beliefs on which the given belief depends
what we know by intuition, what is clear and for its existence or (at least) its justification. Among
distinct, what is indubitable and credible with such beliefs we could presumably find some or all
no fear of error. Thus for Descartes basic knowl- of the following: that the road will be icy or snowy;
edge is always an infallible belief in an indubita- that driving on ice or snow is dangerous; that it will
be truth. All other knowledge must stand on rain or snow tonight; that the temperature will be
that basis through deductive proof. Starting below freezing; appropriate beliefs about the
from such criteria (canons, methods, etc.), forecast and its reliability; and so on.
Descartes concluded that knowledge extended How must such beliefs be interrelated in order
about as far as his contemporaries believed. I to help justify my belief about the danger of driv-
Starting from similar criteria, however, Hume ing tonight? Here foundationalism and coherent-
concluded that both science and common sense ism disagree, each offering its own metaphor. Let
made claims far beyond their rightful limits. us have a closer look at this dispute, starting with
d. Philosophical posterity has rejected Descartes's foundationalism.
theory for one main reason: that it admits too Both Descartes and Hume attribute to human
easily as obvious what is nothing of the sort. knowledge an architectonic structure. There is a
Descartes's reasoning is beautifully simple: nonsymmetric relation of physical support such
God exists; no omnipotent perfectly good that any two floors of a building are tied by that rela-
being would descend to deceit; but if our tion: one of the two supports (or at least helps sup-
common sense beliefs were radically false, that port) the other. And there is, moreover, a part with a
would represent deceit on His part. Therefore, special status: the foundation, which is supported by
our common sense beliefs must be true or at none of the floors while supporting them all.
least cannot be radically false. But in order to With respect to a body of knowledge K (in
buttress this line of reasoning and fIll in someone's possession), foundationalism implies
details, Descartes appeals to various principles that K can be divided into parts K1, K, ... such that
that appear something less than indubitable. there is some nonsymmetric relation R (analo-
e. For his part, Hume rejects all but a minuscule gous to the relation of physical support) which
portion of our supposed common sense orders those parts in such a way that there is
knowledge. He establishes first that there is no one - call it F - that bears R to every other part
way to prove such supposed knowledge on the while none of them bears R in turn to F.
basis of what is obvious at any given moment According to foundational ism, each piece of
through reason or experience. And he con- knowledge lies on a pyramid such as that shown
cludes, in keeping with this methodism, that in in Figure 13.1. The nodes of such a pyramid (for a
point of fact there really is no such knowledge. proposition P relative to a subject S and a time t)
must obey the following requirements:

3 Two Metaphors: The Raft

and the Pyramid

Both metaphors concern the body or system of

knowledge in a given mind. But the mind is of
course a more complex marvel than is sometimes
supposed. Here I do not allude to the depths Figure 13.1

a. The set of all nodes that succeed (directly) no part is untouchable, we must stand on some in
any given node must serve jointly as a base order to replace or repair others. Not every part
that properly supports that node (for S at t). can go at once.
b. Each node must be a proposition that S is According to the new metaphor, what justifies
justified in believing at t. a belief is not that it be an infallible belief with an
c. If a node is not self-evident (for S at t), it indubitable object, nor that it have been proved
must have successors (that serve jointly as deductively on such a basis, but that it cohere
a base that properly supports that node). with a comprehensive system of beliefs.
d. Each branch of an epistemic pyramid
must terminate.
4 A Coherentist Critique of
For the foundationalist Descartes, for instance, Foundationalism
each terminating node must be an indubitable
proposition that S believes at t with no possibility What reasons do coherentists offer for their total
of error. As for the nonterminal nodes, each of rejection of foundationalism? The argument that
them represents inferential knowledge, derived follows below summarizes much of what is alleged
by deduction from more basic beliefs. against foundational ism. But first we must distin-
Such radical foundationalism suffers from a guish between subjective states that incorporate a
fatal weakness that is twofold: (a) there are not so propositional attitude and those that do not.
many perfectly obvious truths as Descartes A propositional attitude is a mental state of some-
thought; and (b) once we restrict ourselves to one with a proposition for its object: beliefs,
what is truly obvious in any given context, very hopes, and fears provide examples. By way of
little of one's supposed common sense knowledge contrast, a headache does not incorporate any
can be proved on that basis. If we adhere to such such attitude. One can of course be conscious of a
radical foundationalism, therefore, we are just headache, but the headache itself does not consti-
wrong in thinking we know so much. tute or incorporate any attitude with a proposi-
Note that in citing such a "fatal weakness" of tion for its object. With this distinction in the
radical foundationalism, we favor particularism background, here is the antifoundationalist argu-
as against the methodism of Descartes and Hume. ment, which has two lemmas - a(iv) and b(iii) -
For we reject the methods or criteria of Descartes and a principal conclusion.
and Hume when we realize that they plunge us in
a deep skepticism. If such criteria are incompati- a. (i) If a mental state incorporates a
ble with our enjoyment of the rich body of knowl- propositional attitude, then it does
edge that we commonly take for granted, then as not give us direct contact with reality,
good particularists we hold on to the knowledge e.g., with pure experience, unfiltered
and reject the criteria. by concepts or beliefs.
If we reject radical foundationalism, however, (ii) If a mental state does not give us
what are we to put in its place? Here epistemol- direct contact with reality, then it
ogy faces a dilemma that different epistemolo- provides no guarantee against error.
gists resolve differently. Some reject radical (iii) If a mental state provides no guaran-
foundational ism but retain some more moderate tee against error, then it cannot serve
form of foundationalism in favor of a radically as a foundation for knowledge.
different coherentism. Coherentism is associated (iv) Therefore, if a mental state incorpo-
with idealism - of both the German and the rates a propositional attitude, then it
British variety - and has recently acquired new cannot serve as a foundation for
vigor and interest. knowledge.
The coherentists reject the metaphor of the b. (i) If a mental state does not incorporate
pyramid in favor of one that they owe to the pos- a propositional attitude, then it is an
itivist Neurath, according to whom our body of enigma how such a state can provide
knowledge is a raft that floats free of any anchor support for any hypothesis, raising its
or tie. Repairs must be made afloat, and though credibility selectively by contrast with

its alternatives. (If the mental state has If we accept such examples, they show us a source
no conceptual or propositional con- of justification that serves as such without
tent, then what logical relation can it incorporating a propositional attitude.
possibly bear to any hypothesis? Belief As for premise a(iii), it is already under suspi-
in a hypothesis would be a proposi- cion from our earlier exploration of premise b(i).
tional attitude with the hypothesis A mental state M can be nonpropositional and
itself as object. How can one depend hence not a candidate for so much as truth, much
logically for such a belief on an experi- less infallibility, while it serves, in spite of that, as
ence with no propositional content?) a foundation of knowledge. Leaving that aside, let
(ii) If a mental state has no proposi- us suppose that the relevant mental state is indeed
tional content and cannot provide propositional. Must it then be infallible in order
logical support for any hypothesis, to serve as a foundation of justification and
then it cannot serve as a foundation knowledge? That is so far from being obvious that
for knowledge. it seems more likely false when compared with an
(iii) Therefore, if a mental state does not analogue in ethics. With respect to beliefs, we may
incorporate a propositional attitude, distinguish between their being true and their
then it cannot serve as a foundation being justified. Analogously, with respect to
for knowledge. actions, we may distinguish between their being
c. Every mental state either does or does not optimal (best of all alternatives, all things consid-
incorporate a propositional attitude. ered) and their being (subjectively) justified. In
d. Therefore, no mental state can serve as a practical deliberation on alternatives for action, is
foundation for knowledge. (From a(iv), it inconceivable that the most eligible alternative
b(iii), and c.) not be objectively the best, all things considered?
Can there not be another alternative - perhaps a
According to the coherentist critic, foundational- llJost repugnant one worth little if any considera-
ism is run through by this dilemma. Let us take a tion - that in point of fact would have a much
closer look. 2 better total set of consequences and would thus
In the first place, what reason is there to think, be better, all things considered? Take the physi-
in accordance with premise b(i), that only proposi- cian attending to Frau Hitler at the birth of little
tional attitudes can give support to their own kind? Adolf. Is it not possible that if he had acted less
Consider practices - e.g., broad policies or cus- morally, that would have proved better in the full-
toms. Could not some person or group be justified ness of time? And if that is so in ethics, may not its
in a practice because of its consequences: that is, likeness hold good in epistemology? Might there
could not the consequences of a practice make it a not be justified (reasonable, warranted) beliefs
good practice? But among the consequences of a that are not even true, much less infallible? That
practice may surely be found, for example, a more seems to me not just a conceivable possibility, but
just distribution of goods and less suffering than indeed a familiar fact of everyday life, where
there would be under its alternatives. And neither observational beliefs too often prove illusory but
the more just distribution nor the lower degree of no less reasonable for being false.
suffering is a propositional attitude. This provides If the foregoing is on the right track, then the
an example in which propositional attitudes (the antifoundationalist is far astray. What has led him
intentions that sustain the practice) are justified by there?
consequences that are not propositional attitudes. As a diagnosis of the antifoundationalist
That being so, is it not conceivable that the justifi- argument before us, and more particularly of its
cation of belief that matters for knowledge be anal- second lemma, I would suggest that it rests on an
ogous to the objective justification by consequences Intellectualist Model of Justification.
that we find in ethics? According to such a model, the justification of
Is it not possible, for instance, that a belief that belief (and psychological states generally) is para-
there is something red before one be justified in sitical on certain logical relations among proposi-
part because it has its origins in one's visual experi- tions. For example, my belief (i) that the streets
ence of red when one looks at an apple in daylight? are wet, is justified by my pair of beliefs (ii) that it

is raining, and (iii) that if it is raining, the streets may perhaps be said to justify belief in its
are wet. Thus we have a structure such as this: existence parasitically on the fact that P
logically implies P! The Intellectualist
B(Q) is justified by the fact that B(Q) IS Model seems either so trivial as to be dull,
grounded on (B(P), B(P~Q)). or else sharp enough to cut equally against
both foundationalism and coherentism.
And according to an Intellectualist Model, this is b. If (i) only propositional attitudes can jus-
parasitical on the fact that tify such propositional attitudes as belief,
and if (ii) to do so they must in turn be
P and (P~Q) together logically imply Q. justified by yet other propositional atti-
tudes, it seems clear that (iii) there is no
Concerning this attack on foundationalism I hope of constructing a complete episte-
will argue (a) that it is useless to the coherentist, mology, one which would give us, in theory,
since if the antifoundationalist dilemma impales an account of what the justification of any
the foundationalist, a form of it can be turned justified belief would supervene on. For (i)
against the coherentist to the same effect; (b) that and (ii) would rule out the possibility of a
the dilemma would be lethal not only to founda- finite regress of justification.
tionalism and coherentism but also to the very c. If only propositional attitudes can justify
possibility of substantive epistemology; and (c) propositional attitudes, and if to do so
that a form of it would have the same effect on they must in turn be justified by yet other
normative ethics. propositional attitudes, it seems clear that
there is no hope of constructing a com-
a. According to coherentism, what justifies a plete normative ethics, one which would
belief is its membership in a coherent and give us, in theory, an account of what the
comprehensive set of beliefs. But whereas justification of any possible justified
being grounded on B(P) and B(PcQ) is action would supervene upon. For the
a property of a belief B(Q) that yields justification of an action presumably
immediately the logical implication of Q depends on the intentions it embodies
and P and (PcQ) as the logical source of and the justification of these, and here we
that property's justificatory power, the are already within the net of propositional
property of being a member of a coherent attitudes from which, for the Intellectualist,
set is not one that immediately yields any there is no escape.
such implication.
It may be argued, nevertheless, (i) that It seems fair to conclude that our coherentist
the property of being a member of a takes his antifoundationalist zeal too far. His anti-
coherent set would supervene in any foundationalist argument helps expose some val-
actual instance on the property of being a uable insights but falls short of its malicious
member of a particular set a that is in fact intent. The foundationalist emerges showing
coherent, and (ii) that this would enable no serious damage. Indeed, he now demands
us to preserve our Intellectualist Model, equal time for a positive brief in defense of his
since (iii) the justification of the member position.
belief B( Q) by its membership in a would
then be parasitical on the logical relations
among the beliefs in a which constitute 5 The Regress Argument
the coherence of that set of beliefs, and
(iv) the justification of B(Q) by the fact a. The regress argument in epistemology con-
that it is part of a coherent set would then cludes that we must countenance beliefs that
be indirectly parasitical on logical rela- are justified in the absence of justification by
tions among propositions after all. other beliefs. But it reaches that conclusion
But if such an indirect form of parasit- only by rejecting the possibility in principle of
ism is allowed, then the experience of pain an infinite regress of justification. It thus opts

for foundational beliefs justified in some non- Let us focus not so much on the giving of
inferential way by ruling out a chain or pyra- justification as on the having of it. Can there
mid of justification that has justifiers, and be a belief that is justified in part by other
justifiers of justifiers, and so on without end. beliefs, some of which are in turn justified by
One may well find this too short a route to yet other beliefs, and so on without end? Can
foundationalism, however, and demand more there be an endless regress of justification?
compelling reasons for thus rejecting an infi- c. There are several familiar objections to such a
nite regress as vicious. We shall find indeed regress:
that it is not easy to meet this demand. (i) Objection: "It is incompatible with human
b. We have seen how even the most ordinary limitations. No human subject could
of everyday beliefs is the tip of an iceberg. harbor the required infinity of beliefs;'
A closer look below the surface reveals a com- Reply: It is mere presumption to fathom
plex structure that ramifies with no end in with such assurance the depths of the
sight. Take again my belief that driving will be mind, and especially its unconscious
dangerous tonight, at the tip of an iceberg, and dispositional depths. Besides, our
(I), as presented in Figure 13.2. The immedi- object here is the nature of epistemic
ate cause of my belief that driving will be haz- justification in itself and not only that of
ardous tonight is the sound of raindrops on such justification as is accessible to
the windowpane. All but one or two members humans. Our question is not whether
of the underlying iceberg are as far as they can humans could harbor an infinite iceberg
be from my thoughts at the time. In what of justification. Our question is rather
sense, then, do they form an iceberg whose tip whether any mind, no matter how deep,
breaks the calm surface of my consciousness? could do so. Or is it ruled out in princi-
Here I will assume that the members of (I) ple by the very nature of justification?
are beliefs of the subject, even if unconscious (ii) Objection: "An infinite regress is indeed
or subconscious, that causally buttress and ruled out in principle, for if justification
thus justify his prediction about the driving were thus infinite how could it possibly
conditions. end?"
Can the iceberg extend without end? It Reply: (i) If the end mentioned is tem-
may appear obvious that it cannot do so, and poral, then why must there be such an
one may jump to the conclusion that any end? In the first place, the subject may
piece of knowledge must be ultimately be eternal. Even if he is not eternal,
founded on beliefs that are not (inferentially) moreover, why must belief acquisition
justified or warranted by other beliefs. This is and justification occur seriatim? What
a doctrine of epistemic foundationalism. precludes an infinite body of beliefs


Driving will be dangerous tonight.

The road will be Driving on snow or

icy or snowy. ice is dangerous.

It will rain or The temperature will

snow all night. be below freezing.
It is raining hard
already and the sky
is overcast.
The forecast estimates
a 100 percent probability
of rain or snow all night.
It is near freezing already
and the forecast
calls for a sharp drop
in temperature.

Figure 13.2

acquired at a single stroke? Human limi- a mind powerful enough to believe every
tations may rule this out for humans, but member of the following sequence:
we have yet to be shown that it is pre-
cluded in principle, by the very nature of (01) There is at least one perfect
justification. (ii) If the end mentioned is number> 100
justificatory, on the other hand, then to There are at least two perfect
ask how justification could possibly end numbers> 100
is just to beg the question. There are at least three perfect
(iii) Objection: "Let us make two assump- numbers> 100
tions: first, that S's belief of q justifies
his belief of p only if it works together If such a believer has no other belief
with a justified belief on his part that q about perfect numbers save the belief
provides good evidence for p; and, that a perfect number is a whole number
second, that if S is to be justified in equal to the sum of its whole factors,
believingp on the basis of his belief of q then surely he is not justified in believing
and is to be justified in believing q on that there are perfect numbers greater
the basis of his belief of r, then S must than 100. He is quite unjustified in
be justified in believing that r provides believing any of the members of sequence
good evidence for p via q. These assump- (a1), in spite of the fact that a challenge
tions imply that an actual regress of to any can be met easily by appeal to its
justification requires belief in an infi- successor. Thus it cannot be allowed after
nite proposition. Since no one (or at all that justification extend infinitely, and
least no human) can believe an infinite an infinite regress is ruled out:'
proposition, no one (no human) can be Reply: We must distinguish between
a subject of such an actual regress.") regresses of justification that are actual
Reply: Neither of the two assumptions is and those that are merely potential.
beyond question, but even granting them The difference is not simply that an
both, it may still be doubted that the con- actual regress is composed of actual
clusion follows. It is true that each finitely beliefs. For even if all members of the
complex belief of form "r provides good regress are actual beliefs, the regress
evidence for p via q[ ... q,," will omit how may still be merely potential in the fol-
some members of the full infinite regress lowing sense: while it is true that if any
are epistemically tied to belief of p. But member were justified then its prede-
that seems irrelevant given the fact that cessors would be, still none is in fact
for each member r of the regress, such that justified. Anyone with our series of
r is tied epistemically to belief of p, there is beliefs about perfect numbers in the
a finite belief of the required sort ("r pro- absence of any further relevant infor-
vides good evidence for p via q[.. .qn") that mation on such numbers would pre-
ties the two together. Consequently there sumably be the subject of such a merely
is no apparent reason to suppose - even potential justificatory regress.
granted the two assumptions - that an (v) Objection: "But defenders of infinite
infinite regress will require a single belief justificatory regresses cannot distin-
in an infinite proposition, and not just an guish thus between actual regresses
infinity of beliefs in increasingly complex and those that are merely potential.
finite propositions. There is no real distinction to be drawn
(iv) Objection: "But if it is allowed that justifi- between the two. For if any regress
cation extend infinitely, then it is ever justifies the belief at its head, then
too easy to justify any belief at all or every regress must always do so. But
too many beliefs altogether. Take, for obviously not every regress does so (as
instance, the belief that there are perfect we have seen by examples), and hence
numbers greater than 100. And suppose no regress can do SO."4

Reply: One can in fact distinguish between rather, whether there can be an actual infi-
actual justificatory regresses and merely nite regress of justification, and the fact
potential ones, and one can do so both that a belief at the head of a potential
abstractly and by examples. regress might still fail to be justified despite
What an actual regress has that a its position does not settle this question.
merely potential regress lacks is the prop- For even if there can be a merely potential
erty of containing only justified beliefs as regress with an unjustified belief at its
members. What they both share is the head, that leaves open the possibility of an
property of containing no member with- infinite regress, each member of which is
out successors that would jointly justify it. justified by its immediate successors work-
Recall our regress about perfect num- ing jointly, where every member of the
bers greater than 100; i.e., there is at least regress is in addition actually justified.
one; there are at least two; there are at least
three; and so on. Each member has a suc-
cessor that would justify it, but no member 6 The Relation of Justification and
is justified (in the absence of further infor- Foundationalist Strategy
mation external to the regress). That is
therefore a merely potential infinite The foregoing discussion is predicated on a simple
regress. As for an actual regress, I see no conception of justification such that a set of
compelling reason why someone (if not a beliefs f3 conditionally justifies (would justify) a
human, then some more powerful mind) belief X iff, necessarily, if all members of f3 are jus-
could not hold an infinite series of actually tified then X is also justified (if it exists). The fact
justified beliefs as follows: that on such a conception of justification actual
endless regresses - such as (0"2) - seem quite pos-
(0"2) There is at least one even number sible blocks a straightforward regress argument in
There are at least two even numbers favor of foundations. For it shows that an actual
There are at least three even infinite regress cannot be dismissed out of hand.
numbers Perhaps the foundationalist could introduce
some relation of justification - presumably more
It may be that no one could be the subject complex and yet to be explicated - with respect to
of such a series of justified beliefs unless he which it could be argued more plausibly that an
had a proof that there is a denumerable actual endless regress is out of the question.
infinity of even numbers. But even if that There is, however, a more straightforward
should be so, it would not take away the strategy open to the foundationalist. For he need
fact of the infinite regress of potential justi- not object to the possibility of an endless regress
fiers, each of which is actually justified, and of justification. His essential creed is the more
hence it would not take away the fact of the positive belief that every justified belief must be
actual endless regress of justification. at the head of a terminating regress. Fortunately,
The objection under discussion is con- to affirm the universal necessity of a terminating
fused, moreover, on the nature of the issue regress is not to deny the bare possibility of a
before us. Our question is not whether nonterminating regress. For a single belief can
there can be an infinite potential regress, trail at once regresses of both sorts: one terminat-
each member of which would be justified ing and one not. Thus the proof of the denumer-
by its successors, such that the belief at its ably infinite cardinality of the set of evens may
head is justified in virtue of its position provide for a powerful enough intellect a termi-
there, at the head of such a regress. The nating regress for each member of the endless
existence and even the possibility of a series of justified beliefs:
single such regress with a belief at its head
that was not justified in virtue of its posi- (0"2) There is at least one even number
tion there would of course settle that There are at least two even numbers
question in the negative. Our question is, There are at least three even numbers

At the same time, it is obvious that each member (iii) every belief that is justified IS so III
of (0"2) lies at the head of an actual endless regress virtue of (i) or (ii) above.
of justification, on the assumption that each
member is conditionally justified by its successor, Classical foundationalism is a type of formal
which is in turn actually justified. foundationalism with respect to epistemic
"Thank you so much;' the foundationalist justification.
may sneer, "but I really do not need that kind of Both of the foregoing theories - simple-
help. Nor do I need to be reminded of my essen- minded hedonism in ethics, and classical
tial creed, which I know as well as anyone. Indeed foundationalism in epistemology - are of
my rejection of endless regresses of justification is course flawed. But they both remain exam-
only a means of supporting my view that every ples of formal foundationalist theories.
justified belief must rest ultimately on founda- b. One way of arguing in favor of formal founda-
tions, on a terminating regress. You reject that tionalism in epistemology is to formulate a
strategy much too casually, in my view, but I will convincing formal foundationalist theory of
not object here. So we put that strategy aside. And justification. But classical foundationalism in
now, my helpful friend, just what do we put in its epistemology no longer has for many the
place?" attraction that it had for Descartes, nor has any
Fair enough. How then could one show the other form of epistemic foundationalism won
need for foundations if an endless regress is not general acceptance. Indeed epistemic founda-
ruled out? tionalism has been generally abandoned, and
its advocates have been put on the defensive by
the writings of Wittgenstein, Quine, Sellars,
7 Two Levels of Foundationalism Rescher, Aune, Harman, Lehrer, and others. It
is lamentable that in our headlong rush away
a. We need to distinguish, first, between two from foundationalism we have lost sight of the
forms of foundational ism: one formal, the different types of foundationalism (formal vs.
other substantive. A type of formal founda- substantive) and of the different grades of each
tionalism with respect to a normative or eval- type. Too many of us now see it as a blur to be
uative property J is the view that the decried and avoided. Thus our present attempt
conditions (actual and possible) within which to bring it all into better focus.
J would apply can be specified in general, e. If we cannot argue from a generally accepted
perhaps recursively. Substantive foundational- foundationalist theory, what reason is there
ism is only a particular way of doing so, and to accept formal foundationalism? There is
coherentism is another. no reason to think that the conditions (actual
Simpleminded hedonism is the view that: and possible) within which an object is spher-
ical are generally specifiable in nongeometric
(i) every instance of pleasure is good, terms. Why should we think that the condi-
(ii) everything that causes something good tions (actual and possible) within which a
is itself good, and belief is epistemically justified are generally
(iii) everything that is good is so in virtue of specifiable in nonepistemic terms?
(i) or (ii) above. So far as I can see, the main reason for
accepting formal foundationalism in the
Simpleminded hedonism is a type of formal absence of an actual, convincing formal
foundationalism with respect to the good. foundationalist theory is the very plausible
Classical foundationalism in epistemology idea that epistemic justification is subject to
is the view that: the supervenience that characterizes norma-
tive and evaluative properties generally.
(i) every infallible, indubitable belief is Thus, if a car is a good car, then any physical
justified, replica of that car must be just as good. If it
(ii) every belief deductively inferred from is a good car in virtue of such properties as
justified beliefs is itself justified, and being economical, little prone to break down,

etc., then surely any exact replica would how epistemic justification supervenes on the
share all such properties and would thus be nonepistemic in a theory of remarkable sim-
equally good. Similarly, if a belief is epistem- plicity: a belief is justified if it has a place
ically justified, it is presumably so in virtue within a system of beliefs that is coherent and
of its character and its basis in perception, comprehensive.
memory, or inference (if any). Thus any It is a goal of ethics to explain how the
belief exactly like it in its character and its ethical rightness of an action supervenes on
basis must be equally well justified. Epistemic what is not ethically evaluative or normative.
justification is supervenient. The justifica- Similarly, it is a goal of epistemology to
tion of a belief supervenes on such proper- explain how the epistemic justification of a
ties of it as its content and its basis (if any) in belief supervenes on what is not epistemically
perception, memory, or inference. Such a evaluative or normative. If coherentism aims
doctrine of supervenience may itself be con- at this goal, that imposes restrictions on the
sidered, with considerable justice, a grade of notion of coherence, which must now be con-
foundationalism. For it entails that every ceived innocent of epistemically evaluative or
instance of justified belief is founded on a normative admixture. Its substance must
number of its nonepistemic properties, such therefore consist of such concepts as explana-
as its having a certain basis in perception, tion, probability, and logical implication -
memory, and inference, or the like. with these conceived, in turn, innocent of
But there are higher grades of founda- normative or evaluative content.
tionalism as well. There is, for instance, the e. We have found a surprising kinship between
doctrine that the conditions (actual and coherentism and substantive foundational-
possible) within which a belief would be ism, both of which turn out to be varieties
epistemically justified can be specified in of a deeper foundationalism. This deeper
general, perhaps recursively (and by refer- foundationalism is applicable to any norma-
ence to such notions as perception, memory, tive or evaluative property , and it comes in
and inference). three grades. The first or lowest is simply the
A higher grade yet of formal foundation- supervenience of : the idea that whenever
alism requires not only that the conditions something has its having it is founded on
for justified belief be specifiable, in general, certain others of its properties which fall
but that they be specifiable by a simple, com- into certain restricted sorts. The second is
prehensive theory. the explicable supervenience of : the idea
d. Simpleminded hedonism is a formal founda- that there are formulable principles that
tionalist theory of the highest grade. If it is explain in quite general terms the condi-
true, then in every possible world goodness tions (actual and possible) within which
supervenes on pleasure and causation in a applies. The third and highest is the easily
way that is recursively specifiable by means of explicable supervenience of : the idea that
a very simple theory. there is a simple theory that explains the
Classical foundationalism in epistemology conditions within which applies. We have
is also a formal foundationalist theory of the found the coherentist and the substantive
highest grade. If it is true, then in every pos- foundationalist sharing a primary goal: the
sible world epistemic justification supervenes development of a formal foundationalist
on infallibility cum indubitability and deduc- theory of the highest grade. For they both
tive inference in a way that is recursively want a simple theory that explains precisely
specifiable by means of a very simple theory. how epistemic justification supervenes, in
Surprisingly enough, coherentism may also general, on the nonepistemic. This insight
turn out to be formal foundationalism of the gives us an unusual viewpoint on some
highest grade, provided only that the concept of recent attacks against foundationalism. Let
coherence is itself both simple enough and free us now consider as an example a certain
of any normative or evaluative admixture. simple form of argument distilled from the
Given these provisos, coherentism explains recent antifoundationalist literature. 5

8 Doxastic Ascent Arguments question is justified. A belief B is made founda-

tional by having some such nonepistemic prop-
Several attacks on foundationalism turn on a sort erty that yields its justification. Take my belief
of"doxastic ascent" argument that calls for closer that I am in pain in a context where it is caused by
scrutiny.6 Here are two examples: my being in pain. The property that my belief
then has, of being a self-attribution of pain caused
A. A belief B is foundationally justified for S by one's own pain is, let us suppose, a nonepis-
in virtue of having property F only if S is temic property that yields the justification of any
justified in believing (1) that most at least belief that has it. So my belief that I am in pain is
of his beliefs with property F are true, and in that context foundationally justified. Along
(2) that B has property F. But this means with my belief that I am in pain, however, there
that belief B is not foundational after all, come other beliefs that are equally well justified,
and indeed that the very notion of (empir- such as my belief that someone is in pain. Thus I
ical) foundational belief is incoherent. am foundationally justified in believing that I am
It is sometimes held, for example, that in pain only if I am justified in believing that
perceptual or observational beliefs are someone is in pain. Those who object to founda-
often justified through their origin in the tionalism as in A or B above are hence mistaken
exercise of one or more of our five senses in thinking that their premises would refute foun-
in standard conditions of perception. The dationalism. The fact is that they would not touch
advocate of doxastic ascent would raise a it. For a belief is no less foundationally justified
vigorous protest, however, for in his view for having its justification yoked to that of another
the mere fact of such sensory prompting closely related belief.
is impotent to justify the belief prompted. The advocate of arguments like A and B must
Such prompting must be coupled with the apparently strengthen his premises. He must appar-
further belief that one's senses work well ently claim that the beliefs whose justification is
in the circumstances, or the like. For we entailed by the foundationally justified status of
are dealing here with knowledge, which belief B must in some sense function as a neces-
requires not blind faith but reasoned trust. sary source of the justification of B. And this would
But now surely the further belief about of course preclude giving B foundationally justi-
the reliability of one's senses itself cannot fied status. For if the being justified of those beliefs
rest on blind faith but requires its own is an essential part of the source of the justification
backing of reasons, and we are off on the of B, then it is ruled out that there be a wholly
regress. non-epistemic source ofB's justification.
B. A belief B of proposition P is foundation- That brings us to a second point about A and
ally justified for S only if S is justified in B, for it should now be clear that these cannot be
believing that there are no factors present selectively aimed at foundationalism. In particu-
that would cause him to make mistakes lar, they seem neither more nor less valid objec-
on the matter of the proposition P. But, tions to coherentism than to foundationalism, or
again, this means that belief B is not foun- so I will now argue about each of them in turn.
dational after all and indeed that the
notion of (empirical) foundational belief A'. A belief X is justified for S in virtue of
is incoherent. membership in a coherent set only if S is
justified in believing (1) that most at
From the vantage point of formal foundation- least of his beliefs with the property of
alism, neither of these arguments seems persua- thus cohering are true, and (2) that X has
sive. In the first place, as we have seen, what makes that property.
a belief foundational (formally) is its having a
property that is nonepistemic (not evaluative in Any coherentist who accepts A seems bound to
the epistemic or cognitive mode), and does not accept A'. For what could he possibly appeal to as
involve inference from other beliefs, but guaran- a relevant difference? But A' is a quicksand of end-
tees, via a necessary principle, that the belief in less depth. (How is he justified in believing A' (1)?

Partly through justified belief that it coheres? And subject: explanatory relations, perhaps, or
what would justify this? And so on ... ) relations of probability or logic.
According to substantive foundational ism,
B'. A belief X is justified for S only if S is jus- as it is to be understood here, there are ulti-
tified in believing that there are no fac- mate sources of justification other than rela-
tors present that would cause him to tions among beliefs. Traditionally these
make mistakes on the subject matter of additional sources have pertained to the spe-
that belief. cial content of the belief or its special relations
to the subjective experience of the believer.
Again, any coherentist who accepts B seems bound b. The view that justification is a matter of rela-
to accept B'. But this is just another road to the tions among beliefs is open to an objection
quicksand. (For S is justified in believing that from alternative coherent systems or detach-
there are no such factors only if. .. and so on.) ment from reality, depending on one's per-
Why are such regresses vicious? The key is spective. From the latter perspective the body
again, to my mind, the doctrine of supervenience. of beliefs is held constant and the surround-
Such regresses are vicious because they would be ing world is allowed to vary, whereas from the
logically incompatible with the supervenience of former perspective it is the surrounding world
epistemic justification on such nonepistemic facts that is held constant while the body of beliefs
as the totality of a subject's beliefs, his cognitive is allowed to vary. In either case, according to
and experiential history, and as many other non- the coherentist, there could be no effect on
epistemic facts as may seem at all relevant. The the justification for any belief.
idea is that there is a set of such nonepistemic
facts surrounding a justified belief such that no Let us sharpen the question before us as fol-
belief could possibly have been surrounded by lows. Is there reason to think that there is at least
those very facts without being justified. Advocates one system B', alternative to our actual system of
of A or B run afoul of such supervenience, since beliefs B, such that B' contains a belief X with the
they are surely committed to the more general following properties:
views derivable from either A or B by deleting
"foundationally" from its first sentence. In each (i) in our present nonbelief circumstances we
case the more general view would then preclude would not be justified in having belief X even
the possibility of supervenience, since it would if we accepted along with that belief (as our
entail that the source of justification always total system of beliefs ) the entire belief system
includes an epistemic component. B' in which it is embedded (no matter how
acceptance of B' were brought about); and
(ii) that is so despite the fact that belief X coheres
9 Coherentism and Substantive within B' at least as fully as does some actual
Foundationalism justified belief of ours within our actual
belief system B (where the justification of
a. The notions of coherentism and substantive that actual justified belief is alleged by the
foundationalism remain unexplicated. We coherentist to derive solely from its coher-
have relied so far on our intuitive grasp of ence within our actual body of beliefs B).
them. In this section we shall consider rea-
sons for the view that substantive founda- The coherentist is vulnerable to counterexam-
tionalism is superior to coherentism. To ples of this sort right at the surface of his body of
assess these reasons, we need some more beliefs, where we find beliefs with minimal coher-
explicit account of the difference between ence, whose detachment and replacement with
the two. contrary beliefs would have little effect on the
By coherentism we shall mean any view coherence of the body. Thus take my belief that I
according to which the ultimate sources of have a headache when I do have a splitting head-
justification for any belief lie in relations ache, and let us suppose that this does cohere
among that belief and other beliefs of the within my present body of beliefs. (Thus I have

no reason to doubt my present introspective other beliefs of mine as that the table has the same
beliefs, and so on. And if my belief does not shape as the piece of paper before me, which is
cohere, so much the worse for coherentism, since oblong, and a different shape than the window
my belief is surely justified.) Here then we have a frame here, which is square, and so on. So far as I
perfectly justified or warranted belief. And yet can see, however, there is no insurmountable
such a belief may well have relevant relations of obstacle to replacing that whole set of coherent
explanation, logic, or probability with at most a beliefs with an equally coherent set as follows:
small set of other beliefs of mine at the time: say, that the table before me is square, that the table
that I am not free of headache, that I am in pain, has the same shape as the square window frame,
that someone is in pain, and the like. If so, then an and a different shape than the piece of paper,
equally coherent alternative is not far to seek. Let which is oblong, and so on. The important points
everything remain constant, including the split- are (a) that this replacement may be made with-
ting headache, except for the following: replace out changing the rest of one's body of beliefs or
the belief that I have a headache with the belief any aspect of the world beyond, including one's
that I do not have a headache, the belief that I am present visual experience of something oblong,
in pain with the belief that I am not in pain, the not square, as one looks at the table before one;
belief that someone is in pain with the belief that and (b) that it is so, in part, because of the fact (c)
someone is not in pain, and so on. I contend that that the subject need not have any beliefs about
my resulting hypothetical system of beliefs would his present sensory experience.
cohere as fully as does my actual system of beliefs, Some might be tempted to respond by alleging
and yet my hypothetical belief that I do not have a that one's present experience is self-intimating,
headache would not therefore be justified. What i.e., always necessarily taken note of and reflected
makes this difference concerning justification in one's beliefs. Thus if anyone has visual experi-
between my actual belief that I have a headache ence of something oblong, then he believes that
and the hypothetical belief that I am free of head- he has such experience. But this would involve a
ache, each as coherent as the other within its own further important concession by the coherentist
system, if not the actual splitting headache? But to the radical foundationalist, who would have
the headache is not itself a belief nor a relation been granted two of his most cherished doctrines:
among beliefs and is thus in no way constitutive the infallibility of introspective belief and the
of the internal coherence of my body of beliefs. self-intimation of experience.
Some might be tempted to respond by alleging
that one's belief about whether or not one has a
headache is always infallible. But since we could 10 The Foundationalist's Dilemma
devise similar examples for the various sensory
modalities and propositional attitudes, the response The antifoundationalist zeal of recent years has
given for the case of headache would have to be left several forms of foundationalism standing.
generalized. In effect, it would have to cover These all share the conviction that a belief can be
"perip heral" beliefs generally - beliefs at the periph- justified not only by its coherence within a com-
ery of one's body of beliefs, minimally coherent prehensive system but also by an appropriate
with the rest. These peripheral beliefs would all be combination of observational content and origin
said to be infallible. That is, again, a possible in the use of the senses in standard conditions.
response, but it leads to a capitulation by the coher- What follows presents a dilemma for any founda-
entist to the radical foundationalist on a crucial tionalism based on any such idea.
issue that has traditionally divided them: the infal-
libility of beliefs about one's own subjective states. a. We may surely suppose that beings with
What is more, not all peripheral beliefs are observational mechanisms radically
about one's own subjective states. The direct real- unlike ours might also have knowledge of
ist is probably right that some beliefs about our their environment. (That seems possible
surroundings are uninferred and yet justified. even if the radical difference in observa-
Consider my present belief that the table before tional mechanisms precludes overlap in
me is oblong. This presumably coheres with such substantive concepts and beliefs.)

Table 13.1 one may then base various generaliza-

Human tions, rules of thumb, and maxims of
Extraterrestrial being
public health, nutrition, legislation, eti-
Visual experience experience quette, hygiene, and so on. But these are
Experience of Experience of something F all then derived generalizations which rest
something red for their validity on the fundamental prin-
Belief that there is Belief that there is ciple. Similarly, one may also ask, with
something red something F before one
respect to the generalizations advanced by
before one
our foundationalist, whether these are
proposed as fundamental principles or as
derived maxims or the like. This sets him
b. Let us suppose that there is such a being, face to face with a dilemma, each of whose
for whom experience of type cfJ (of which alternatives is problematic. If his propos-
we have no notion) has a role with respect als are meant to have the status of second-
to his beliefs of type cfJ analogous to the ary or derived maxims, for instance, then
role that our visual experience has with it would be quite unphilosophical to stop
respect to our visual beliefs. Thus we might there. Let us turn, therefore, to the other
have a schema such as that in Table 13.l. alternative.
c. It is often recognized that our visual expe- f. On reflection it seems rather unlikely that
rience intervenes in two ways with respect epistemic principles for the justification
to our visual beliefs: as cause and as justi- of observational beliefs by their origin in
fication. But these are not wholly inde- sensory experience could have a status
pendent. Presumably, the justification of more fundamental than that of derived
the belief that something here is red generalizations. For by granting such
derives at least in part from the fact that principles fundamental status we would
it originates in a visual experience of open the door to a multitude of equally
something red that takes place in normal basic principles with no unifying factor.
circumstances. There would be some for vision, some for
d. Analogously, the extraterrestrial belief hearing, etc., without even mentioning the
that something here has the property of corresponding extraterrestrial principles.
being F might be justified partly by the g. It may appear that there is after all an idea,
fact that it originates in a cfJ experience of however, that unifies our multitude of
something F that takes place in normal principles. For they all involve sensory
circumstances. experience and sensible characteristics.
e. A simple question presents the founda- But what is a sensible characteristic?
tionalist's dilemma: regarding the epis- Aristotle's answer appeals to examples:
temic principle that underlies our colors, shapes, sounds, and so on. Such a
justification for believing that something notion might enable us to unify percep-
here is red on the basis of our visual expe- tual epistemic principles under some
rience of something red, is it proposed as more fundamental principle such as the
a fundamental principle or as a derived following.
generalization? Let us compare the famous
Principle of Utility of value theory, accord- If a is a sensible characteristic, then the belief
ing to which it is best for that to happen that there is something with a before one is
which, of all the possible alternatives in the (prima facie) justified if it is based on a visual
circumstances, would bring with it into experience of something with a in condi-
the world the greatest balance of pleasure tions that are normal with respect to a.
over pain, joy over sorrow, happiness over
unhappiness, content over discontent, or h. There are at least two difficulties with such
the like. Upon this fundamental principle a suggestion, however, and neither one

can be brushed aside easily. First, it is not extraterrestrial foundations? If such more
clear that we can have a viable notion of abstract principles are in fact accessible,
sensible characteristics on the basis of then the less general principles that define
examples so diverse as colors, shapes, tones, the human foundations and those that
odors, and so on. Second, the authority of define the extraterrestrial foundations are
such a principle apparently derives from both derived principles whose validity
contingent circumstances concerning the depends on that of the more abstract
reliability of beliefs prompted by sensory principles. In this the human and extra-
experiences of certain sorts. According to terrestrial epistemic principles would
the foundationalist, our visual beliefs are resemble rules of good nutrition for an
justified by their origin in our visual expe- infant and an adult. The infant's rules
rience or the like. Would such beliefs be would of course be quite unlike those
equally well justified in a world where valid for the adult. But both would still be
beliefs with such an origin were nearly based on a more fundamental principle
always false? that postulates the ends of well-being and
I. In addition, finally, even if we had a viable good health. What more fundamental
notion of such characteristics, it is not principles might support both human
obvious that fundamental knowledge of and extraterrestrial knowledge in the way
reality would have to derive causally or that those concerning good health and
otherwise from sensory experience of well-being support rules of nutrition for
such characteristics. How could one both the infant and adult?
impose reasonable limits on extraterres-
trial mechanisms for noninferential acqui-
sition of beliefs? Is it not possible that 11 Reliabilism: An Ethics of Moral
such mechanisms need not always func- Virtues and an Epistemology of
tion through sensory experience of any Intellectual Virtues
sort? Would such beings necessarily be
denied any knowledge of the surround- In what sense is the doctor attending Frau Hitler
ings and indeed of any contingent spatio- justified in performing an action that brings with it
temporal fact? Let us suppose them to far less value than one of its accessible alternatives?
possess a complex system of true beliefs According to one promising idea, the key is to be
concerning their surroundings, the struc- found in the rules that he embodies through stable
tures below the surface of things, exact dispositions. His action is the result of certain stable
details of history and geography, all virtues, and there are no equally virtuous alterna-
constituted by concepts none of which tive dispositions that, given his cognitive limitations,
corresponds to any of our sensible charac- he might have embodied with equal or better total
teristics. What then? Is it not possible that consequences, and that would have led him to
their basic beliefs should all concern fields infanticide in the circumstances. The important
of force, waves, mathematical structures, move for our purpose is the stratification of justifi-
and numerical assignments to variables in cation. Primary justification attaches to virtues and
several dimensions? This is no doubt an other dispositions, to stable dispositions to act,
exotic notion, but even so it still seems through their greater contribution of value when
conceivable. And if it is in fact possible, compared with alternatives. Secondary justification
what then shall we say of the noninferen- attaches to particular acts in virtue of their source
tial beliefs of such beings? Would we have in virtues or other such justified dispositions.
to concede the existence of special epis- The same strategy may also prove fruitful in
temic principles that can validate their epistemology. Here primary justification would
noninferential beliefs? Would it not be apply to intellectual virtues, to stable dispositions
preferable to formulate more abstract for belief acquisition, through their greater con-
principles that can cover both human and tribution toward getting us to the truth. Secondary

justification would then attach to particular then how can it possibly serve as a founda-
beliefs in virtue of their source in intellectual tion for belief? How can one infer or justify
virtues or other such justified dispositions.? anything on the basis of a state that, having
That raises parallel questions for ethics and no propositional content, must be logically
epistemology. We need to consider more carefully dumb? An analogy with ethics suggests a
the concept of a virtue and the distinction reason to reject this dilemma. Other reasons
between moral and intellectual virtues. In episte- are also advanced and discussed.
mology, there is reason to think that the most 5. The regress argument. In defending his
useful and illuminating notion of intellectual position, the foundationalist often attempts
virtue will prove broader than our tradition to rule out the very possibility of an infi-
would suggest and must give due weight not only nite regress of justification (which leads
to the subject and his intrinsic nature but also to him to the necessity for a foundation).
his environment and to his epistemic community. Some of his arguments to that end are
This is a large topic, however, to which I hope examined.
some of us will turn with more space, and insight, 6. The relation of justification and foundation-
than I can now command. alist strategy. An alternative foundationalist
strategy is exposed, one that does not require
ruling out the possibility of an infinite
Summary regress of justification.
7. Two levels of foundationalism. Substantive
1. Two assumptions: (AI) that for a belief to con- foundationalism is distinguished from
stitute knowledge it must be (a) true and (b) formal foundationalism, three grades of
justified; and (A2) that the justification rele- which are exposed: first, the superveni-
vant to whether or not one knows is a sort of ence of epistemic justification; second, its
epistemic or theoretical justification to be dis- explicable supervenience; and, third, its
tinguished from its practical counterpart. supervenience explicable by means of a
2. Knowledge and criteria. Particularism is dis- simple theory. There turns out to be a sur-
tinguished from methodism: the first gives prising kinship between coherentism and
priority to particular examples of knowledge substantive foundationalism, both of
over general methods or criteria, whereas the which aim at a formal foundationalism of
second reverses that order. The methodism of the highest grade, at a theory of the great-
Descartes leads him to an elaborate dogma- est simplicity that explains how epistemic
tism whereas that of Hume leads him to a justification supervenes on nonepistemic
very simple skepticism. The particularist is, of factors.
course, antiskeptical on principle. 8. Doxastic ascent arguments. The distinction
3. Two metaphors: the raft and the pyramid. For between formal and substantive founda-
the foundationalist every piece of knowledge tionalism provides an unusual viewpoint on
stands at the apex of a pyramid that rests on some recent attacks against foundational-
stable and secure foundations whose stability ism. We consider doxastic ascent arguments
and security do not derive from the upper as an example.
stories or sections. For the coherentist a body 9. Coherentism and substantive foundational-
of knowledge is a free- floating raft every plank ism. It is argued that substantive founda-
of which helps directly or indirectly to keep tionalism is superior, since coherentism is
all the others in place, and no plank of which unable to account adequately for the epis-
would retain its status with no help from the temic status of beliefs at the "periphery" of a
others. body of beliefs.
4. A coherentist critique of foundationalism. No 10. The foundationalist's dilemma. All founda-
mental state can provide a foundation for tionalism based on sense experience is sub-
empirical knowledge. For if such a state is ject to a fatal dilemma.
propositional, then it is fallible and hence no 11. Reliabilism. An alternative to foundational-
secure foundation. But if it is not propositional, ism of sense experience is sketched.

Appendix8 blamed on the investigator. In the present exam-

ple, blame is out of place. By hypothesis, Magoo
What one is rationally justified in believing obvi- conducts impeccable "inquiry" both in arriving
ously depends on the data in one's possession. at his data and on the basis of his data. But he
But what data one has can depend on how much still falls short of knowledge, despite his war-
and how well one investigates. Consider, there- ranted, correct belief. His shortcoming is sub-
fore, the following possibility. What if A is standard equipment, for which we may suppose
rationally justified in believing x given his body him to be blameless. Hence something other
of data D whereas B is not rationally justified in
than epistemic justification or correct belief can
believing x given his body of data D z, where D z help determine what one knows or does not
includes D j but is much more extensive as a know. Even if one correctly believes that p with
result of Xs irresponsible negligence and B's com- full rational justification and free of irrational
mendable thoroughness? The present account or neglectful unbelief, one may still be in no
might unfortunately grant A knowledge while position to know, because of faulty cognitive
denying it to B, for Xs neglect so far has no bearing equipment.
on any epistemic pyramid. In all of the foregoing cases, someone misses
We have considered a situation where some- or is liable to miss available information which
one lacks knowledge owing to his misuse of his may be highly relevant and important and may
cognitive equipment, either by letting it idle make a difference to what he can conclude on the
when it should be functioning or by busily question in hand. In each case, moreover, he
employing it dysfunctionally. Another situation seems culpable or discredited in some sense:
where someone lacks knowledge despite having he would seem less reliable than otherwise for his
rationally justified correct belief might be called role in any such case. But there appear to be situ-
the Magoo situation - where S lacks adequate ations where again someone misses available
equipment to begin with (relative to the ques- information with no culpability or discredit.
tion in hand: whether p ).9 It is because of this Harman gives an example where S reads in a
type of lack that despite his extensive experi- newspaper that some famous person has been
ence with cable cars, Mr Magoo does not know assassinated, but does not read the next edition,
that his cable car will arrive safely when, where all reports of the assassination are denied
unknown to him, bombs are raining all around by highly authoritative and trustworthy people.
it. Of course, even if you have less than 20-20 If practically the whole country reads the next
vision you can still know that there is an ele- edition and people don't know what to believe,
phant in front of you when you see one there. does S alone know of the assassination, provided
So not just any defect will make your equip- the next edition is in fact a pack of lies?!O I sup-
ment inadequate for a judgment on the ques- pose we would be inclined to say that he does not
tion whether p. I would venture that it must be know (especially if had he read the next edition,
a defect that prevents you from acquiring infor- he would not have known what to believe). But
mation that (i) a normal inquirer in the epis- what if only two or three people get a chance to
temic community would acquire in that read the next edition before it is recalled by the
situation and (ii) makes a difference to what newspaper? Should we now say that out of mil-
you can reasonably conclude on the question lions who read the first story and mourn the
whether p (or at least to how reasonably you loved leader not one knows of his death? I sup-
can draw the conclusion). pose we would be inclined to say that the fake
The possibility of inadequate cognitive equip- edition and the few deceived by it make no differ-
ment requires a further and more striking depar- ence concerning what everybody else knows. It
ture from the traditional conception of seems plausible to conclude that knowledge has a
knowledge. Despite having warranted correct further "social aspect," that it cannot depend on
belief, someone may lack knowledge owing to one's missing or blinking what is generally
his neglectful data-collection. There lack of known.
knowledge could be traced back to epistemic Our departures from the traditional concep-
irresponsibility, to substandard performance tion of knowledge put in relief the relativity of

knowledge to an epistemic community. This is (from the point of view of a K) to a subject only if
brought out most prominently by the require- both he is rationally justified in believing it and he
ment that inquirers have at least normal cognitive is in a position to know (from the K point of
equipment (e.g., normal perceptual apparatus, view) whether it is true. It may be (and not just
where that is relevant). But our new require- appear) evident to Magoo from his point of view
ment - that inquirers not lack or blink generally that he will reach the other side safely, but it seems
known relevant information - also brings out the wrong to say of Magoo as he steps into the cable
relativity. A vacationer in the woods may know car with bombs raining all around that it is quite
that p well enough for an average vacationer, but evident to him that he will arrive safely. It seems
he won't have the kind of knowledge his guide wrong for whom to say this? For one of us, natu-
has. A guide would scornfully deny that the ten- rally; that is, for a normal human from his point
derfoot really knows that p. Relative to the epis- of view. And since a normal human could not
temic community of guides (for that area) the help seeing and hearing the bombs, from the
tenderfoot lacks relevant generally known infor- human point of view Magoo is not in a position
mation, and misses relevant data that the average to know that he will arrive safely, inasmuch as he
guide would grasp in the circumstances. is missing relevant information that a normal
These departures from the traditional account human would gather in the circumstances. Hence
may make better sense if we reflect that the hon- Magoo does not have human knowledge that he
orific term "knowledgeable" is to be applied only will arrive safely, for it is not evident to him from
to those who are reliable sources of information, the human point of view that he will so arrive.
surely an important category for a language-
using, social species. Consider this account:
We have now taken note of two types of situa- (A) S knows that p iff
tion where correct, fully warranted belief falls (a) it is true thatp;
short of knowledge owing to no neglect or faulty (b) S believes that p; and
reasoning or false belief. Despite commendable (c) there is a non-defective epistemic pyra-
thoroughness and impeccable reasoning mid for S and the proposition that p.
unspoiled by falsehood, one may still fail to be "in
a position to know," owing either to faulty cogni- Every node of such a pyramid must be true and
tive equipment or to missed generally known evident. And for every node n that has successors,
information. I am not suggesting that these are the successors must serve as grounds that give the
the only ways to be out of position to know. I have subject S rational warrant for believing n. What
no complete list of epistemic principles describ- now seems too narrow about this account emerges
ing ways of arriving at a position to know or of with the explanation of what a pyramid of know1-
being blocked from such a position. My sugges- edge is, and of what the evident is. For in this
tion is only that there are such principles, and explanation what is evident to S is identified with
that in any case we must go beyond the traditional what S is rationally justified in believing. But it
emphasis by epistemologists on warrant and rea- now seems plain that for x to be evident to S, two
soning as determinants of knowledge. Despite the conditions must be satisfied: (i) that S be ration-
importance of warranted correct belief in deter- ally justified in believing x, and (ii) that S be in a
mining what we know, the Gettier examples show position to know whether x is true. And we must
that it is not alone enough to guarantee knowl- also take note of the relativity of knowledge to an
edge. What is more, warranted correct belief sup- epistemic community. Let us therefore replace
ported by reasoning unspoiled by falsehood seems (A) with the following:
immune to Gettier examples, but it still falls short
of knowledge, as we have seen. (B) S knows (from the Kpoint of view) thatp iff
My conclusion is that to understand knowl- (a) it is true thatp;
edge we must enrich our traditional repertoire of (b) S believes that p; and
epistemic concepts with the notion of being in a (c) there is a non-defective epistemic pyra-
position to know (from the point of view of a K, e.g., mid (from the K point of view) for S
a human being). Thus a proposition is evident and the proposition that p.

Every node of such a pyramid must now be true terms of which pyramids were defined. What fol-
and evident from the K point of view. lows is an attempt to solve these problems by
Normally when epistemologists discuss switching pyramids upside down into trees.
knowledge (of the colors and shapes of surround- Let us emphasize, however, that this will not
ing objects, of one's own or one's neighbor's commit one to a picture of knowledge according
mental states, and so on), they plainly do so from to which there is a bedrock of self-evident propo-
the human point of view. But other points of view sitions. It is perfectly consistent with the present
are possible even in ordinary conversation. The theory that part of what makes any proposition
expert/layman distinction is replicable in many evident is its coherence with a network of mutu-
different contexts, and with each replication we ally supporting propositions. Since there is bound
have a new epistemically relevant distinction in to be a multitude of such coherent networks, how-
points of view, with expert knowledge on one side ever, a non-arbitrary narrowing of the field must
and layman knowledge on the other. be supported by something other than coherence.
Neither Magoo nor the newspaper reader who We turn finally to an account (C) according to
alone has not seen the new edition is in a position which S knows that p provided that both (a) S cor-
to know (from the human point of view) about rectly believes that p,l1 and (b) there is a set of
the relevant subject matter. Thus we can under- propositions that fully and non -defectively renders
stand their ignorance and, by parity of reasoning, it evident to S that p (where a set "non-defectively
the ignorance of all those who are out of position renders it evident to S that p" if and only if it does
to know that p because they lack either adequate so without attributing to S any false belief). 12
cognitive equipment or relevant information Supposing this account correct, every bit of
that is generally known to those who have taken knowledge has a tree like that shown in Figure 13.3.
an epistemic stand on the question whether p Note that each node of such a tree is a proposi-
(where to suspend judgment is to take an epis- tion. Thus the "root" node is the-proposition-
temic stand, whereas to be totally oblivious to the that-PI' and the first terminal node (from the left)
matter is not). the- proposition-that-Pll1. 13
What it is for S's belief that p to be fully grounded There is an important difference between these
has been explained by means of our epistemic pyra- trees and our earlier pyramids. Except for terminal
mids. That answer points in the right direction, but nodes, every node of a tree is an epistemic proposi-
it can be made more precise: e.g., by clarifying the tion, whereas not a single node of a pyramid need
grounding relation. Moreover, we have found that a be epistemic at all. Pyramids display propositions
fully grounded correct belief is not necessarily that are evident to A (not propositions that such
knowledge, and this for at least two reasons: (i) it and such other propositions are evident to S), and
may rest directly or indirectly on some false ground, they also show which propositions ground (for S)
and (ii) the believer may not be in a position to any proposition for which S has grounds. Trees
know. display true epistemic propositions concerning S
We have tried to allow for these possibilities by and they also show what "makes these proposi-
broadening epistemic pyramids, by making room tions true" via epistemic principles. A tree must do
for our new epistemic notion ofbeing-in-a-posi- this for every epistemic proposition that consti-
tion-to-know, and by noting that to support tutes one of its nodes. That is to say, trees contain
knowledge epistemic pyramids must be non- no epistemic terminal nodes. It is in this sense that
defective, i.e., must contain no false nodes. But trees provide complete epistemic explanations of
pyramids are objectionable for other reasons as the truth of their root nodes.
well: (i) they may mislead by suggesting that ter-
minal nodes provide a "foundation" in one or RIII P111 P112 P121
another undesirable sense, or by suggesting that
terminal nodes must come first in time, so that RII Pll
one may later build on them; (ii) more seriously,
there is an unacceptable vagueness in the very
idea of such a pyramid, which derives mainly
from the vagueness of the "grounding" relation in Figure 13.3


But Descartes's methodism was at most par- theory of knowing." From our viewpoint,
tial. James Van Cleve has supplied the mate- this effort is better understood not as an
rials for a convincing argument that the way attempt to define propositional knowledge,
out of the Cartesian circle is through a par- but as an attempt to formulate fundamental
ticularism of basic knowledge. See James principles of justification.
Van Cleve, "Foundationalism, Epistemic Cf. the work ofD. Armstrong, Belief, Truth
Principles, and the Cartesian Circle:' The and Knowledge (Cambridge: Cambridge
Philosophical Review 88 (1979), pp. 55-91 and University Press, 1973), and that of F. Dretske,
E. Sosa and J. Kim (eds), Epistemology: An A. Goldman, and M. Swain, whose relevant
Anthology (Malden, MA; Oxford: Blackwell, already published work is included in
2000), pp. 242-60. But this is, of course, com- G. Pappas and M. Swain (eds), Essays on
patible with methodism on inferred knowl- Knowledge and Justification (Ithaca and
edge. Whether Descartes subscribed to such London, 1978). But the theory is still under
methodism is hard (perhaps impossible) to development by Goldman and by Swain,
determine, since in the end he makes room who have reached general conclusions about
for all the kinds of knowledge required by it similar to those suggested here, though not
particularism. But his language when he necessarily - so far as I know - for the same
introduces the method of hyperbolic doubt, reasons or in the same overall context.
and the order in which he proceeds, suggest 8 From "How Do You Know?" American
that he did subscribe to such methodism. Philosophical Quarterly 11, 2 (1974),
2 Cf. Laurence BonJour, "The Coherence Theory pp.113-22.
of Empirical Knowledge:' Philosophical Studies 9 The Magoo situation is the situation of that
30, pp. 281-312, and, especially, Michael unfortunate nearsighted and hearing-
Williams, Groundless Belief (Oxford: Blackwell, impaired cartoon character who fortunately
1977); and Bonjour, "Can Empirical Knowledge escapes disaster at every turn.
Have a Foundation?," this vol., ch. lO. lO Gilbert Harman, "Induction:' in Marshall
3 Cf. Richard Foley, "Inferential Justification Swain (ed.), Induction, Acceptance and
and the Infinite Regress:' American Rational Belief (Dordrecht: Reidel, 1970),
Philosophical Quarterly 15 (1978), pp. 311-16. esp. Sect. IV, pp. 95-7.
4 Cf. John Post, "Infinite Regresses of 11 Whether knowledge entails belief at all is of
Justification and of Explanation," Philosophical course a vexed question of long standing,
Studies 34 (1980). but there is no room for it here. A helpful
5 The argument of this whole section is devel- and interesting discussion is found in Keith
oped in greater detail in my paper "The Lehrer's "Belief and Knowledge,"
Foundations of Foundationalism:' Nous 14, Philosophical Review 77 (1968), pp. 491-9.
pp. 547--65. 12 In what follows, the relativity of knowledge
6 For some examples of the influence of doxas- to an epistemic community is left implicit, as
tic ascent arguments, see Wilfrid Sellars's it normally is in ordinary thought and
writing in epistemology: "Empiricism and the speech.
Philosophy of Mind," in Science, Perception 13 Strictly speaking, what we have here is obvi-
and Reality (London: Routledge and Kegan ously a partial tree schema. For convenience,
Paul, 1963), esp. section VIII, and particularly however, I speak of trees even when I mean
p. 168. Also 1. T. Oakley, "An Argument for partial tree schemata. Also, it should not be
Skepticism Concerning Justified Belief' thought that every tree must have exactly
American Philosophical Quarterly 13 (1976), three ranks (RI, RII, and RIII). On the con-
pp. 221-8; and BonJour, "Can Empirical trary, a tree may have any number of ranks,
Knowledge Have a Foundation?" so long as it has more than one.
7 This puts in a more traditional perspective
the contemporary effort to develop a "causal

Human Knowledge
and the Infinite
Regress of Reasons

Peter D. Klein

Introduction provide well considered reasons for rejecting

foundationalism. Of course, if there are no con-
The purpose of this paper is to ask you to con- vincing reasons for rejecting infinitism, then
sider an account of justification that has largely these typical defenses of foundationalism and of
been ignored in epistemology. When it has been coherentism fail.
considered, it has usually been dismissed as so I will not rehearse the many arguments
obviously wrong that arguments against it are not against foundationalism or coherentism in any
necessary. The view that I ask you to consider can detail here. But very briefly, foundationalism is
be called "Infinitism."l Its central thesis is that the unacceptable because it advocates accepting an
structure of justificatory reasons is infinite and arbitrary reason at the base, that is, a reason for
non-repeating. My primary reason for recom- which there are no further reasons making it
mending infinitism is that it can provide an even slightly better to accept than any of its con-
acceptable account of rational beliefs, i.e., beliefs traries. Traditional coherentism is unacceptable
held on the basis of adequate reasons, while because it advocates a not too thinly disguised
the two alternative views, foundationalism and form of begging the question; and seemingly
coherentism, cannot provide such an account. more plausible forms of coherentism are just
Typically, just the opposite viewpoint is foundationalism in disguise.
expressed. Infinitism is usually mentioned as one Thus, if having rational beliefs is a necessary
of the logically possible forms that our reasoning condition of some type of knowledge, both
can take; but it is dismissed without careful con- foundationalism and coherentism lead directly
sideration because it appears initially to be so to the consequence that this type of knowledge
implausible. 2 Foundationalists often begin by is not possible because each view precludes the
somewhat cavalierly rejecting infinitism. Then possibility of having beliefs based upon ade-
they proceed by eliminating coherentism through quate reasons. On the other hand, infinitism
a series of complex and carefully developed argu- makes such knowledge at least possible because
ments. Coherentists often follow a similar general it advocates a structure of justificatory reasons
strategy by first rejecting infinitism without any that satisfies the requirements of rational belief
careful examination of the view and then they possession.
This paper has two main sections. In the first
section I sketch infinitism in broad outline and
Originally published in Philosophical Perspectives 13, argue that it is the only account of the structure of
Epistemology (1999), pp. 297-325. reasons that can satisfy two intuitively plausible

constraints on good reasoning. In the second Infinitism is like foundationalism in holding

section I defend infinitism against the best that there are features of the world, perhaps non-
objections to it. normative features, that make a belief a reason.
Not just any old belief is a reason. Infinitism is
unlike foundationalism because infinitism holds
I. A Sketch of Infinitism that there are no ultimate, foundational reasons.
Every reason stands in need of another reason.
Let me begin by pointing out some important This can be stated in a principle - the Principle of
similarities and dissimilarities between infinitism Avoiding Arbitrariness (PAA).
and the two alternative accounts of justification.
Infinitism is like most forms of traditional coher- PAA: For all x, if a person, S, has a justifica-
entism in holding that only reasons can justify a tion for x, then there is some reason, r l ,
belief.! Infinitism is unlike traditional coherent- available to S for x; and there is some
ism because infinitism does not endorse question reason, r 2 , available to S for r l ; etc.
begging reasoning. 4 Indeed, this can be captured
in what can be called the "Principle of Avoiding Note that there are two features of this principle.
Circularity" (PAC). The first is that it is reasons (as opposed to some-
thing else like appropriate causal conditions
PAC: For all x, if a person, S, has a justification responsible for a belief) that are required when-
for x, then for all y, if Yis in the eviden- ever there is a justification for a belief. The second
tial ancestry of x for S, then x is not in is that the chain of reasons cannot end with an
the evidential ancestry of y for S. arbitrary reason - one for which there is no fur-
ther reason. I conjoin these features in one princi-
By "evidential ancestry" I am referring to the links ple because both are needed to capture the
in the chains of reasons, sometimes branching, well-founded intuition that arbitrary beliefs,
that support beliefs.5 For example, if r is a reason beliefs for which no reason is available, should be
for p, and q is a reason for r, then r is in the evi- avoided. I will consider some objections to both
dential ancestry of p, and q is in the evidential aspects of PAA shortly.
ancestry of both p and r.6 I will not defend PAC in Some foundationalists could accept PAA by
this paper because it strikes me as an obvious pre- claiming that the available reason, r, could just be
supposition of good reasoning. It is intended x, itself. They could assert that some propositions
merely to make explicit the intuition behind the are "self-justified." That is not ruled out by PAA;
prohibition of circular reasoning. but coupled with PAC, that possibility is ruled
Not all so-called "coherentists" would deny out. Indeed, the combination of PAC and PAA
PAC. These "coherentists" are really closet foun- entails that the evidential ancestry of a justified
dationalists because it is not the propositions belief be infinite and non-repeating. Thus, some-
within a set of coherent propositions that serve as one wishing to avoid infinitism must reject either
reasons for other beliefs in the set; rather the reason PAC or PAA (or both).8 It is the straight-forward
for every belief in the set is simply that it is a intuitive appeal of these principles that is the best
member of such a see Thus, these non-traditional reason for thinking that if any beliefs are justified,
coherentists avoid question begging reasoning by the structure of reasons must be infinite and non-
a two stage procedure. First, they define what it repeating.
means for a set of propositions to be coherent PAA requires that the reason for a belief must
(perhaps mutual probability enhancements plus be available to S. "Availability" is a key notion in
some other conditions) and, then, they claim that my account of infinitism for, among other things,
the reason for accepting each proposition in the it has the potential for anchoring justification, as
set is that it is a member of such a set of beliefs. understood by the infinitist, in non-normative
That is consistent with endorsing PAC. But as properties. 9 So, it would be well for us to dwell a
we will see, this type of coherentism,