Philosophical Applications of Cognitive Science

1st Edition Alvin I. Goldman pdf download

https://ebookgate.com/product/philosophical-applications-of-cognitive-science-1st-edition-alvin-i-
goldman/

★★★★★ 4.8/5.0 (47 reviews) ✓ 161 downloads ■ TOP RATED

"Excellent quality PDF, exactly what I needed!" - Sarah M.

DOWNLOAD EBOOK
 Philosophical Applications of Cognitive Science 1st Edition
Alvin I. Goldman pdf download

TEXTBOOK EBOOK EBOOK GATE

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Epistemology A Contemporary Introduction 1st Edition Alvin

I. Goldman

https://ebookgate.com/product/epistemology-a-contemporary-
introduction-1st-edition-alvin-i-goldman/

ebookgate.com

Reliabilism and Contemporary Epistemology Essays 1st

Edition Alvin I. Goldman

https://ebookgate.com/product/reliabilism-and-contemporary-
epistemology-essays-1st-edition-alvin-i-goldman/

ebookgate.com

Simulating Minds The Philosophy Psychology and

Neuroscience of Mindreading Alvin I. Goldman

https://ebookgate.com/product/simulating-minds-the-philosophy-
psychology-and-neuroscience-of-mindreading-alvin-i-goldman/

ebookgate.com

Cognitive Science Daniel Kolak

https://ebookgate.com/product/cognitive-science-daniel-kolak/

ebookgate.com
Cognitive Foundations of Linguistic Usage Patterns
Empirical Studies Applications of Cognitive Linguistics
1st Edition Schmid
https://ebookgate.com/product/cognitive-foundations-of-linguistic-
usage-patterns-empirical-studies-applications-of-cognitive-
linguistics-1st-edition-schmid/
ebookgate.com

Auditing The Art and Science of Assurance Engagements 15th

Canadian Edition Edition Alvin Arens

https://ebookgate.com/product/auditing-the-art-and-science-of-
assurance-engagements-15th-canadian-edition-edition-alvin-arens/

ebookgate.com

Cognitive Poetics Goals Gains and Gaps Applications of

Cognitive Linguistics 1st Edition Geert Brone

https://ebookgate.com/product/cognitive-poetics-goals-gains-and-gaps-
applications-of-cognitive-linguistics-1st-edition-geert-brone/

ebookgate.com

Radical Embodied Cognitive Science 1st Edition Anthony

Chemero

https://ebookgate.com/product/radical-embodied-cognitive-science-1st-
edition-anthony-chemero/

ebookgate.com

Rethinking Science Education Philosophical Perspectives

1st Edition Roland M. Schulz

https://ebookgate.com/product/rethinking-science-education-
philosophical-perspectives-1st-edition-roland-m-schulz/

ebookgate.com
 PHILOSOPHICAL
APPLICATIONS OF
COGNITIVE SCIENCE
 PHILOSOPHICAL
APPLICATIONS OF
COGNITIVE SCIENCE
Alvin I. Goldman
UNIVERSITY OF ARIZONA
Focus Series

First pu b lish ed 1993 by W estview Press

Published 2018 by Routledge

711 T hird A venue, N ew York, NY 10017, USA
2 Park Square, M ilton Park, A bingdon, O xon 0 X14 4RN

,
Routledge is an imprint of the Taylor & Francis Group an informa business

C opyright © 1993 Taylor & Francis

All rights reserved. N o p art of this book m ay be rep rin ted or rep ro d u ced or
utilised in any form or by any electronic, m echanical, or other m eans, n o w kn o w n
or hereafter invented, including ph otocopying and recording, or in any
inform ation storage or retrieval system , w ith o u t perm ission in w ritin g from the
publishers.

Notice:
Pro d uct or corporate nam es m ay be trad em ark s or registered tradem arks, and are
used only for identification an d explanation w ith o u t intent to infringe.

A revised version of Chapter 5 first appeared as “Ethics and Cognitive Science” in

Ethics 103, no. 2 (1993): 337-360. Copyright © 1993 by the University of Chicago
Press. Reprinted by permission.

Library of Congress Cataloging-in-Publication Data

Goldman, Alvin I., 1938-
Philosophical applications of cognitive science / Alvin
I. Goldm an
p. cm. — (Focus series)
Includes bibliographical references and index.
ISBN 0-8133-8039-1 (cloth). — ISBN 0-8133-8040-5 (pbk.)
1. Philosophy and cognitive science. I. Title. II. Series:
Focus series (Westview Press)
B945.G593P55 1993
149— dc2o 92-38143
CIP

ISBN 13: 978-0-8133-8040-7 (pbk)

For Raphie and Sidra
 Contents

List o f Tables and Figures ix

Preface xi

1 EPISTEMOLOGY I

The Questions of Epistemology, 1

Knowledge and the Sources of Knowledge, 1
Visual Object-Recognition, 5
Rationality and Evidence, 9
Self-Deception and Memory, 13
Logic and Rationality, 16
Probability and Rationality, 22
Suggestions for Further Reading, 31

2 PHILO SO PH Y OF SCIENCE 33
Is Observation Theory-Laden? 33
A Computational Model of Theory
Acceptance, 39
Imagistic Models of Understanding, 51
Is Numerical Knowledge Innate? 56
Suggestions for Further Reading, 61

3 PHILO SO PH Y O F M IND 63
Dualism and Materialism, 63
The Identity Theory, 6 5
Property Dualism, 68
Philosophical Behaviorism, 69
Functionalism, 72
Cognitive Science and the Functional
“Level,” 75
Computationalism and the Language
of Thought, 76

Vil
CONTENTS

Folk Psychology and Ehminativism, 79

Cognitive Science and Mental Concepts, 82
First-Person Attributions, 84
Third-Person Attributions, 88
The Theory Theory Versus the Simulation
Theory, 92
Suggestions for Further Reading, 96

4 METAPHYSICS 99
Descriptive and Prescriptive Metaphysics, 99
Physical Bodies, 101
Categories of Thought, 108
Prescriptive Metaphysics, 113
Color, 114
Individual Essences, 119
Suggestions for Further Reading, 124

5 ETHICS 125
Three Roles for Cognitive Science, 125
The Mental Structure of Moral Cognition, 126
Innate Constraints on Moral Thinking, 130
Judgments of Subjective Well-Being, 134
Endowment and Contrast in Judgments
of Well-Being, 137
Empathy, 140
Empathy and Descriptive Ethics, 146
Empathy and Prescriptive Ethics, 150
Conclusion, 152
Suggestions for Further Reading, 154

References 155
About the Book and Author 171
Index 173

v iii
 Tables and Figures

Tables
1.1 Logically possible truth-value assignments to
several propositions 22

Figures
1.1 Schematic drawing (after Descartes) illustrating
the distance information provided by
convergence 4
1.2 Geon components and their relations 8
1.3 Example of five stimulus objects in an
experiment on the perception of degraded
objects 10
2.1 The Müller-Lyer illusion 37
2.2 The Necker cube 46
2.3 A connectionist network for interpreting the
Necker cube 46
2.4 An EC H O network showing explanatory
breadth 49
2.5 An EC H O network showing explanation by a
higher-level hypothesis 50
2.6 Network representing Darwin’s argument 52
3.1 The simulation heuristic 91
4.1 Illustrations of Gestalt principles of figurai
goodness 104
4.2 Habituation display and test displays for an
experiment on infants’ perceptions of partly
occluded obj ects 106
4.3 Apparatus for an experiment on haptic
perception of object unity and boundaries 108

ix
 Preface

The history of philosophy is replete with dissections of the

mind, its faculties, and its operations. Historical epistemolo-
gists invoked such faculties as the senses, intuition, reason,
imagination, and the active and the passive intellect. They
wrote of cognitive acts and processes such as judging, conceiv-
ing, abstracting, introspecting, synthesizing, and schematiz-
ing. Ethicists shared this interest in mental faculties and con-
tents. Moral philosophers studied the appetites, the will, the
passions, and the sentiments. All of these philosophers pro-
ceeded on the premise that a proper understanding of the
mind is essential to many branches of philosophy. This pre-
mise is still widely accepted, but time has wrought some
changes. In previous centuries the study of the mind was the
private preserve of philosophy, and that is no longer true. A
number of disciplines have developed a variety of scientific
methods, both theoretical and experimental, for studying the
mind-brain. These disciplines—the cognitive sciences—
include cognitive psychology, developmental psychology, lin-
guistics, artificial intelligence, neuroscience, and cognitive an-
thropology. Their practitioners attempt to understand and
model the mind’s wide-ranging activities, such as perception,
memory, language processing, inference, choice, and motor
control. Philosophy also contributes to the project, but it no
longer has a privileged position.
Since it is now clear that the most detailed and reliable in-
formation about the mind will emerge from the collective ef-
forts of the cognitive sciences, philosophy should look to
those sciences for relevant information and work hand in hand
with them. Cognitive science can never replace philosophy,
since the mission of philosophy extends well beyond the de-

xi
PREFACE

scription of mental processes, but it can provide a wide range

of helpful fact and theory. In maintaining an alliance with cog-
nitive science, philosophy continues its ancient quest to un-
derstand the mind. In the modern age, however, this pursuit
requires careful attention to what is being learned by a new
group of scientists. Plato and Aristotle created their own
physics and cosmologies; contemporary metaphysicians must
learn physics and cosmology from physicists. Similarly, while
René Descartes and David Hume created their own theories
of the mind, contemporary philosophers must give respectful
attention to the findings of scientific research.
About 100 years ago, interactions between logic and philos-
ophy assumed dramatic new importance. A similarly dramatic
collaboration is now occurring between philosophy and cog-
nitive science. When modern logic emerged in the early years
of this century, philosophers saw a powerful new tool that
could transform the field. Some believed that philosophy
should simply become logical analysis, modeled, for example,
on Bertrand Russell's theory of descriptions. This was doubt-
less an excess of zeal. But developments in logic have clearly
had wide-ranging and beneficial applications throughout phi-
losophy. Similarly, empirical studies of cognition now have
great potential for enriching many areas of philosophy. This
book seeks to illustrate and enlarge upon this theme.
This book does not address the methodology of cognitive
science: the question of how, in detail, cognitive hypotheses or
theories are tested by empirical evidence. N or does it attempt
to survey the various cognitive sciences. The bulk of the em-
pirical research presented here is from cognitive psychology,
but a bit is drawn from artificial intelligence, linguistics, and
neuroscience. I do not attempt to give "equal time" to all of
these disciplines or provide a balanced sampling of their theo-
retical structures. Such samplings are already available in
other texts. Rather, the center of attention is the variety of
philosophical problems that can benefit from cognitive stud-

xii
 PREFACE

ies, not the variety of cognitive studies that can contribute to

philosophy. At the same time, philosophical morals are often
drawn rather briefly, and the instructor or reader will often
wish to pursue or debate these morals further.
My selection and discussion of material has been shaped by
the desire for a short and accessible text. This constraint has
dictated the exclusion of highly technical topics and topics
that would require a good deal of groundwork. This is one
reason there is rather scant attention to certain important ar-
eas of cognitive science, for example, the study of language.
Despite such gaps, I hope that the choice of examples conveys
the flavor of much of the research in cognitive science as well
as its potential fruitfulness for philosophical theory and
reflection.
To assist instructors and students in the further exploration
of the topics covered here, I have appended a list of suggested
readings at the end of each chapter. Many of these readings ap-
pear in an anthology I edited entitled Readings in Philosophy
and Cognitive Science (MIT Press/A Bradford Book, 1993).
Conceived partly as a companion to the present text, the an-
thology contains five chapters that closely parallel this book,
plus chapters on language and methodology. I shall abbreviate
citations to this anthology by [R]. When [R] follows the cita-
tion of a work in a suggested readings section, this indicates
that the cited work, or some selection from it (or, occasionally,
a closely related work by the same author) appears in the an-
thology.
I am grateful to a number of people for extremely helpful
comments on the first draft of the manuscript: Paul Bloom,
Owen Flanagan, Kihyeon Kim, Joseph Tolliver, and Karen
Wynn as well as Westview editor Spencer Carr. Their com-
ments resulted in numerous improvements, both substantive
and stylistic.

Alvin I. Goldman

xiii
 Chapter One

Epistemology

The Questions of Epistemology

Epistemology addresses such questions as: (1) What is knowl-
edge? (2) What is rationality? and (3) What are the sources and
prospects for human knowledge and rationality? To answer
question 3, we would have to inquire into the specific cogni-
tive faculties, processes, or methods that are capable of confer-
ring knowledge or rationality. Cognitive science is clearly rel-
evant to such an inquiry. In asking about the “prospects” for
knowledge and rationality, question 3 also hints that there
may be limits or failings in people’s capacities to know or to be
rational. Potential challenges and threats to knowledge and ra-
tionality have indeed been a focus of traditional epistemology.
Here we shall address threats that stem from the potential in-
adequacy of some of our cognitive faculties and processes.
Thus, whether we are addressing “sources” or “prospects,”
cognitive science, as the science of our cognitive endowments,
can make important contributions to epistemology.

Knowledge and the Sources of Knowledge

Let us start with knowledge, and let us first ask what knowl-
edge consists in. Epistemologists generally agree that know-
ing, at a minimum, involves having true belief. You cannot
know there is a snake under the table unless you believe that
there is. Further, you cannot know there is a snake under the

1
EPISTEMOLOGY

table unless it is true, i.e., unless a snake is really there.

Epistemologists also agree that mere true belief is not suffi-
cient for knowledge, at least in any strong or robust sense of
the term. Suppose you have a phobia for snakes, and you are
always imagining them in this or that part of your house. You
haven’t looked under the table just now, nor has anybody said
anything about a snake being there. But you are convinced
that a snake is there. On this lone occasion you are right;
someone has introduced a harmless garter snake for a practical
joke. Is it correct to say that you know that a snake is under
the table? Surely not. Thus, believing what is true is not
enough to claim knowledge.
What must be added to true belief to qualify as knowledge?
One popular answer, found in the reliability theory of knowl-
edge, says that to be a case of knowing, a true belief must be
formed by a cognitive process or method that is generally reli-
able, i.e., one that generally produces true beliefs. In the snake
example this condition is not met. Your supposition that a snake
is under the table does not stem from seeing it or from being told
about it by someone who has seen it; it results from phobia-
driven imagination. This way of forming beliefs is not at all reli-
able. Hence, although it coincidentally yields true belief on the
specific occasion in question, it does not yield knowledge.
A detailed formulation of the reliability theory of knowl-
edge requires many refinements (see Goldman 1979, 1986,
1992b). Let us suppose, however, that something along these
general lines is correct. We can then return to the question
posed earlier concerning the sources and prospects for human
knowledge. Under the reliability theory this question be-
comes: Which mental faculties and procedures are capable of
generating true or accurate beliefs, and which are liable to pro-
duce false or inaccurate beliefs?
In the seventeenth and eighteenth centuries, the rationalist
and empiricist philosophers debated the question of which
faculties were the most reliable for belief formation. The lead-

2
 EPISTEMOLOGY

ing empiricists, John Locke, George Berkeley, and David

Hume, placed primary emphasis on sense-based learning,
whereas rationalists like René Descartes emphasized the supe-
rior capacity of reason to generate knowledge. Another cen-
tral disagreement was over the influence of innate ideas or
principles in knowledge acquisition. While the rationalists af-
firmed the existence of such innate factors, the empiricists de-
nied them.
The debate between Descartes and Berkeley over the nature
of depth perception will serve to illustrate this dispute. People
regularly form beliefs about the relative distances of objects,
but how can such judgments be accurate? What features of vi-
sion make such reliable judgment possible? As these early phi-
losophers realized, images formed by light on the retina are
essentially two-dimensional arrays. How can such two-
dimensional images provide reliable cues to distance or depth?
Descartes (1637) argued that one way people ascertain the
distance of objects is by means of the angles formed by
straight lines running from the object seen to the eyes of the
perceiver. Descartes compared this process to a blind man
with a stick in each hand. When he brings the points of the
sticks together at the object, he forms a mangle with one hand
at each end of the base, and if he knows how far apart his
hands are, and what angles the sticks make with his body, he
can calculate, “by a kind of geometry innate in all men” (em-
phasis added), how far away the object is. The same geometry
applies, Descartes argued, if the observer’s eyes are regarded
as the ends of the base of a triangle, with the straight lines that
extend from them converging at the object, as shown in Figure
1.1. Thus, perceivers can compute the distances of objects by a
sort of “natural geometry," knowledge of which is given in-
nately in humankind’s divinely endowed reason.
Berkeley, on the other hand, denied that geometric compu-
tations enter into the process: “I appeal to anyone’s experience
whether upon sight of an object, he computes its distance by

3
EPISTEMOLOGY

F IG U R E 1.1 Schematic drawing (after Descartes) illustrating the distance infor-

mation provided by convergence. Given the distance between the centers of the
two retinas (AB) and the eyes’ angles of regard (Δ CAB and Δ CBA), the distance
of object C can be computed. Source: E. Spelke, “O rigins of Visual Knowledge,’’ in
D. Osherson, S. Kosslyn, and J. Hollerbach, eds., Visual Cognition and Action
(Cambridge, Mass.: M IT Press, 1990). Reprinted by permission.

the bigness of the angle made by the meeting of the two optic
axes? ... In vain shall all the mathematicians in the world tell
me that I perceive certain lines and angles ... so long as I my-
self am conscious of no such thing" (1709, sec. 12, italics in
original). Berkeley held that distance (or depth) is not imme-
diately perceived by sight but is inferred from past associa-
tions between things seen and things touched. Once these past
associations are established, the visual sensations are enough
to suggest the “tangible" sensations the observer would have if
he were near enough to touch the object. Thus, Berkeley's em-
piricist account of depth perception posits learned associa-
tions rather than innate mathematical principles.
This debate about depth perception continues today in con-
temporary cognitive science, although several new types of
cues for depth perception have been proposed. Cognitive sci-

4
 EPISTEMOLOGY

entists also continue to debate the role of innate factors in per-

ception. It is widely thought that perceptual systems have
some innately specified “assumptions” about the world that
enable them, for the most part, to form accurate representa-
tions. An example of such an “assumption" comes from stud-
ies of visual motion perception. Wallach and O'Connell
(1953) bent pieces of wire into abstract three-dimensional
shapes and mounted them in succession on a rotating turn-
table. They placed a light behind the rotating shape so that it
cast a sharp ever-changing shadow on a screen, which was ob-
served by the subject. The shadow was a two-dimensional im-
age varying in time. All other information was removed from
sight. Looking at the shadow, however, the subjects perceived
the three-dimensional form of the wire shape with no trouble
at all. In fact, the perception of three-dimensional form was so
strong in this situation that it was impossible for the subjects
to perceive the shadow as a rubbery two-dimensional figure.
From this and other studies, it has been concluded that the vi-
sual system has a built-in “rigidity assumption" : Whenever a
set of changing two-dimensional elements can be interpreted
as a moving rigid body, the visual system interprets it that
way. That is, the visual system makes the two-dimensional ar-
ray appear as a rigid, three-dimensional body. This response
can produce illusions in the laboratory, as when flashing dots
on a screen are seen as a smoothly moving rigid body.
Presumably, however, the world is largely populated with
rigid bodies of which one catches only partial glimpses. So this
rigidity assumption produces accurate visual detection most
of the time. The assumption is innate, and it is pretty reliable.

Visual Object-Recognition
Let us further explore the prospects for vision-based knowl-
edge by considering the way the visual system classifies ob-

5
EPISTEMOLOGY

jects by reference to their shape. And let us ask not simply

whether such classification can be reliable, but whether it can
be reliable in suboptimal or degraded circumstances, e.g.,
when one has only a partial glimpse of the object. After all, in
everyday life things are not always in full view, and we fre-
quently have to identify them quickly without getting a better
view. Under such conditions, can vision still enable us to iden-
tify objects correctly as chairs, giraffes, or mushrooms? If so,
how does it do this? Classification must ultimately proceed
from retinal stimulation. But no unique pattern of retinal
stimulation can be associated with a single type of object, nor
even a particular instance of the type, since differences in an
object’s orientation can dramatically affect the retinal image.
Furthermore, as just indicated, objects may be partially hid-
den or occluded behind other surfaces, as when viewed behind
foliage. How and when can the visual system still achieve ac-
curate object recognition?
A person stores in memory a large number of representa-
tions of various types or categories, such as chair; giraffe,
mushroom, and so on. When perceiving an object, an observer
compares its perceptual representation to the category repre-
sentations, and when a “match” is found, the perceived object
is judged to be an instance of that category. What needs to be
explained is (1) how the categories are represented, (2) how
the information from the retinal image is processed or trans-
formed, and (3) how this processed information is compared
to the stored representations so that the stimulus is assigned to
the correct category.
One prominent theory, due to Irving Biederman (1987), be-
gins with the hypothesis that each category of concrete ob-
jects is mentally represented as an arrangement of simple vol-
umetric shapes, such as blocks, cylinders, spheres, and
wedges. Each of these primitive shapes is called a geon (for
geometrical ion). Geons can be combined by means of various
relations, such as top-of, side-connected, larger-than, and so

6
 EPISTEMOLOGY

forth. Each category of objects is represented as a particular

combination of related geons. For example, a cup can be rep-
resented as a cylindrical geon that is side-connected to a
curved, handle-like geon, whereas a pail can be represented by
the same two geons but with the handle-like geon on top of
the cylindrical geon, as illustrated in Figure 1.2.
The geon theory postulates that when a viewer perceives an
object, the visual system interprets the retinal stimulation in
terms of geon components and their relations. If the viewer
can identify only a few appropriately related geons, he may
still be able to uniquely specify the stimulus if only one stored
object type has that particular combination of geons. An ele-
phant, for example, may be fully represented by nine compo-
nent geons, but it may require as few as three geons in appro-
priate relations to be correctly identified. In other words, even
a partial view of an elephant might suffice for accurate recog-
nition if it enables the visual system to recover three geons in
suitable relations.
When an object is partially occluded or its contours are
somehow degraded, correct identification depends on
whether the remaining contours enable the visual system to
construct the right geons. Consider Figure 1.3. The left col-
umn shows five nondegraded stimulus objects. The middle
column has versions of the same objects with some deleted
contours. These deleted contours, however, can be recon-
structed by the visual system by “filling in” smooth continu-
ous lines. This enables the visual system to recover the rele-
vant geons and identify the objects correctly despite the
missing contours. The right column pictures versions of the
same objects with different deleted segments. In these ver-
sions, the geons cannot be recovered by the visual system be-
cause the deletions omit telltale clues of the distinct geons. In
the degraded cup, for example, one cannot tell that two geons
are present (the bowl part and the handle). This makes identi-
fication difficult, if not impossible. Of course, one might

7
EPISTEMOLOGY

FIG U R E 1.2 Geon components and their relations. (Left) A given view of an ob-
ject can be represented as an arrangement of simple primitive volumes, or geons, of
which five are shown here. (Right) O nly two or three geons are required to
uniquely specify an object. The relations among the geons matter, as illustrated
with the pail and cup. Source: I. Biederman, “Higher-Level Vision,” in D.
O sherson, S. Kosslyn, and J. Hollerbach, eds., Visual Cognition and Action (Cam -
bridge, Mass.: M IT Press, 1990). Reprinted by permission.

guess rightly that the object is a cup, but such a guess could
not yield knowledge. There is no reliable way of telling that
the object is a cup. What Biederman’s theory of visual object-
recognition reveals is the nature of the reliable process that
does yield knowledge—a process involving detection of a suf-
ficient number and combination of geons to secure a match to
a unique object-model stored in memory. Because of this pro-
cess, the visual system can get enough information to achieve
 EPISTEMOLOGY

knowledge even when objects are partially hidden or oc-

cluded.
The topic of perception is one to which we shall return in
Chapter 2, when we discuss theory and observation in science.
Right now, however, let us turn to the prospects for human ra-
tionality.

Rationality and Evidence

According to one popular principle of rationality, the total ev-
idence principle, it is rational for a person S to believe a propo-
sition p at time t only if p is well supported by the total evi-
dence S possesses at t. The intuitive rationale for this
requirement is straightforward enough. There are cases in
which some of the evidence an agent possesses provides good
support for a given proposition, but the support is defeated by
other evidence he possesses. For example, upon arriving home
from the office, Sam might see his wife’s car in the driveway.
This supports the proposition that she is home. However, Sam
may have additional evidence that undercuts this support. He
may know that his wife’s car didn’t start this morning and that
she took a taxi to work instead. Given this total evidence, it
would not be rational for Sam to believe that his wife is home.
Granted the plausibility of the total evidence principle, how
exactly should we interpret it? What does it mean for a piece
of evidence to be possessed? Consider an example from
Richard Feldman (1988). Suppose my friend Jones tells me
that the hike up to Precarious Peak is not terribly strenuous or
dangerous, that it is the sort of thing I can do without undue
difficulty. Jones knows my abilities with respect to these sorts
of things, and he seems to be an honest person. O n the basis of
his testimony, I believe that the hike is something I can do. Is
it rational of me to believe this? Suppose further that I have
failed to recall the time Jones told me I could paddle my canoe

9
 F IG U R E 1.3 Example of five stimulus objects in an experiment on the perception
of degraded objects. Colum n (a) shows the original intact versions. C olum n (b)
shows the recoverable versions. Colum n (c) shows the nonrecoverable versions.
Source: Modified from I. Biederman, “H um an Image Understanding: Recent
Experiments and a T heory,” Computer Vision, Graphics, and Im age Processing,
32(1985): 29-73. Reprinted by permission of Academic Press.

10
 EPISTEMOLOGY

down Rapid River, something he knew to be far beyond my

abilities, and I don’t realize that he just gets a kick out of send-
ing people off on grueling expeditions. Although I fail to re-
call this incident, it is stored in my memory and I could be re-
minded of it, though in fact nobody does so. Is this unrecalled
piece of information a bit of evidence I possess? If it isn’t part
of my “possessed” evidence, then believing that the hike is
within my ability is rational for me since I have (other) evi-
dence for trusting Jones. If it is part of my possessed evidence,
though, then believing that the hike is something I can do is ir-
rational, since this incident undercuts Jones’s credibility. The
general question suggested by this example is this: When one
believes something on the basis of new evidence and fails to
recall some bit of counterevidence, when, if ever, does this
counterevidence constitute part of the evidence one possesses?
Let us compare two answers to this question. One answer is
that anything stored in memory is a piece of evidence. A per-
son’s total evidence is all the information ever stored in mem-
ory and still lodged there, however easy or difficult it may be
to retrieve. Clearly, if we accept this answer, it will be easy to
fail the demands of rationality. Long-forgotten episodes from
childhood that haven’t been recalled for twenty or thirty years
would qualify as part of one’s total evidence; neglect of these
episodes, when they are evidentially relevant, would be
counted as irrationality.
A second answer posits that an item of evidence is pos-
sessed only if it is readily accessible or retrievable from mem-
ory. But how accessible is “readily” accessible, and how are
degrees of accessibility to be measured? An adequate theory
of this subject does not yet exist, but the elements of any such
theory obviously should reflect the properties of human
memory.
Cognitive psychologists think of memories as varying in
strength: the stronger the memory, the greater the probability
it will be retrieved and the faster it will be retrieved. Two fac-

11
EPISTEMOLOGY

tors influence strength: (1) decay and (2) interference. As time

passes, memory strength gradually lessens through decay.
However, reactivation of a memory boosts its strength once
more. Only with disuse does it continue to fade or decay.
Interference occurs when related or overlapping material be-
comes mixed with, or substitutes for, the target information.
For example, where did you park your car (or bike) when you
came to school this morning? This information may be hard
to recall because memories of similar episodes of car-parking
on other days interfere with your memory of today’s parking
episode.
Both of these factors, decay and interference, might be at
work in the Precarious Peak example. Decay would be partly
responsible for my failure to recall that Jones deceived me
about canoeing down Rapid River, because it has been a long
time since it occurred and I haven’t been reminded of it in the
meantime. Furthermore, Jones has been perfectly friendly and
honest to me on all other matters, so these acts and traits inter-
fere with my memory of the Rapid River episode.
Whether a memory is retrieved on a given occasion depends
not only on its strength but also on the retrieval “cues” pre-
sented to memory (either deliberately or accidentally). A pop-
ular model of memory depicts it as a complex structure of
nodes or elements interconnected by means of associative
links. When a cue enters memory at one node, it activates that
node, and this activation spreads through further portions of
the system as a function of the strength of the links between
node pairs. The whole process resembles a rumor spreading
through a society, where the exact directions and speed of
spread depend on the strength of the communicative links be-
tween individuals. Whether a given item in memory is re-
trieved (activated) on a specific occasion depends on what
cues enter the system and on the prior associations, or associa-
tive pathways, between the cued node(s) and the target node.
A target node will be difficult to retrieve from a given starting

12
 EPISTEMOLOGY

point (say, the thought of proposition p) if it is unlikely that a

strong pathway will emerge. For example, if you start with the
goal of activating an image of your third-grade teacher, there
may be few cues that would achieve this result. But if someone
reminded you of a dramatic incident involving the teacher,
that cue might suffice.
These complexities dim the hopes of getting any simple
measure of “ready accessibility,” and hence any simple mea-
sure of evidence “possession.” Perhaps the moral to be drawn
is that simple principles appealing to the notion of “posses-
sion” should be abandoned in favor of more complex and sub-
tle principles. Any detailed principle of rationality must rec-
ognize that human belief formation operates under the
constraints of memory and must take the psychology of mem-
ory into account.

Self-Deception and Memory

It is widely noticed that people tend to have unrealistically in-
flated opinions of themselves. Although this is something of a
commonplace, it is interesting to find experimental support
for it in psychology. Moreover, it is not entirely obvious just
what is the source of such illusions. Exactly which processes
initiate and maintain these opinions? This question can only
be answered by cognitive science.
Summarizing a variety of research, Shelley Taylor (1989)
reports that illusions about the self are particularly prevalent
among children. Children believe that they are capable of
many tasks, including ones they have never tried. Most kin-
dergartners and first-graders say they are at or near the top of
the class. They have great expectations for their future success.
Moreover, these grandiose assessments are quite unresponsive
to negative feedback, at least until about age seven. It is
Taylor’s thesis that such illusions contribute quite positively

13
EPISTEMOLOGY

to mental health. From an epistemological point of view, how-

ever, they certainly smack of irrationality.
There is considerable evidence that positive self-illusions
are also prevalent among adults. Most people, for example, see
themselves as better than others and as above average in most
of their qualities. Because it is logically impossible for most
people to be better than everyone else, this positive view of
themselves appears to be, at least to some degree, illusory or
nonveridical. In one survey, 90 percent of automobile drivers
considered themselves to be better than average drivers.
Obviously, only about half of these people can be right.
People whose driving had involved them in accidents serious
enough to involve hospitalization and drivers with no acci-
dent histories gave almost identical descriptions of their driv-
ing abilities. Irrespective of their accident records, people
judged themselves to be more skillful than average.
One psychological explanation of these findings might in-
voke the notion of self-schemas: mental structures that guide
the selection and retrieval of information about the self. A
dinner guest who thinks of himself as witty is likely to inter-
pret his barbed remark toward another guest as humorous and
is likely to recall this witticism later (even if other guests did
not find the remark witty at all). In recalling information that
fits a self-schema, he inadvertently reinforces the self-schema.
Each situation that the witty person interprets as an example
of his witty banter provides him with additional evidence that
he is witty. Thus, a self-schema both enables us to interpret the
information that fits our prior conception of ourselves and
helps cement that self-conception.
One problem here is that our prior theory about the self ap-
pears to influence our perception of the evidence. This is a
topic we shall discuss in Chapter 2. A second problem is that
memory seems to retrieve instances that lend positive support
to the theory. The self-styled witty person more easily re-
trieves his witticisms than his attempts at witticism that fell

14
 EPISTEMOLOGY

perceptibly flat. Let us focus on the second problem. If this se-

lection process is indeed a feature of memory, it would seem
to constitute an irrational bias. If positive pieces of evidence
are better stored in memory or more likely to be recalled than
negative pieces, then there can be no fair weighing of the total
evidence one has accumulated.
Although this general hypothesis has not been fully ex-
plored by cognitive psychologists, it poses questions concern-
ing the very structure of memory sketched earlier. Pairs of
nodes in memory may be linked more readily if their contents
cohere, or fit with one another, and positive instances of a trait
would be one example of coherence. If this is correct, then
when an accepted contention is queried, it will be easier for
memory to recall positive support for that contention than
negative evidence, even if the latter has been plentiful.
A somewhat different approach to the phenomenon of
self-illusion emphasizes the role of wishful thinking. Young
children, for example, do not differentiate very well be-
tween what they wish were true and what they think is true,
and this trait could account for their estimations of their
abilities (Stipek 1984). Wishful thinking and other desire-
infected cognitions may also be important in adults and may
drive the processes of memory. Anthony Greenwald (1980)
suggests that memory often fabricates and revises our per-
sonal history. Unlike the academic historian, who is ex-
pected to adhere closely to the facts and insert a personal
evaluation only in the interpretation, the personal historian
takes unbridled license with the facts themselves, rearrang-
ing and distorting them and omitting aspects of history alto-
gether in an effort to create and maintain a favorable image
of the self. As writer Carlos Fuentes observes, “Desire will
send you back into memory ... for memory is desire satis-
fied." (Fuentes 1964, 58).
Taylor surveys many other cognitive mechanisms and strat-
egies by which people unrealistically enhance their self-image

15
EPISTEMOLOGY

or their assessment of their situation and prospects in life. It

seems clear, then, that at least on the topic of the self, people
have tendencies toward epistemic irrationality, as philoso-
phers call it, where the “epistemic" dimension is concerned
with the pursuit of objective truth. These tendencies may be
rational in a prudential or pragmatic sense; indeed, research
evidence indicates that self-enhancing strategies lead to higher
motivation, greater persistence at tasks, more effective perfor-
mance, and greater success. From the standpoint of truth or
accuracy, however, they leave much to be desired.

Logic and Rationality

Let us move from questions of memory to other dimensions
of people's capacities to be (epistemically) rational. It is
widely agreed that rational belief depends on the logical and
probabilistic relations that hold between hypotheses and evi-
dence. For a belief to be rational, it must either follow logically
from the evidence or be highly probable on that evidence. The
prospects for rational belief, then, depend heavily on people's
abilities to detect logical and probabilistic relations among
propositions. Just how strong are people's abilities in these re-
spects? In particular, how proficient are untutored thinkers at
logic and probability tasks? While the answers are not all in,
these are questions to which cognitive scientists have devoted
a good bit of attention.
Starting with human competence at logic, let us first con-
sider a study (Rips and Marcus 1977) in which subjects were
asked to judge the validity of simple argument forms involv-
ing a conditional (if-then) sentence. Four of these arguments
were: Modus Ponens (If P then Q; P; therefore Q); Modus
Tollens (If P then Q; not-Q; therefore not-P); Affirming the
Consequent (If P then Q; Q; therefore P); and Denying the
Antecedent (If P then Q; not-P; therefore not-Q). The first

16
 EPISTEMOLOGY

two of these, of course, are valid, and the second two are in-
valid. One hundred percent of the subjects marked Modus
Ponens valid, which of course is good news for defenders of
human logical competence. But only 57 percent judged
Modus Tollens to be valid. Even worse is the fact that 23 per-
cent of the subjects wrongly judged Affirming the Conse-
quent to be valid and 21 percent judged Denying the Ante-
cedent to be valid. These results, however, are not wholly con-
clusive. Some researchers point out that people commonly
misunderstand English conditional sentences, often because
they interpret them as biconditionals (“if and only i f " state-
ments). If this was the interpretation of the subjects who
marked Affirming the C onsequent or Denying the A nte-
cedent valid, they were not guilty of a fallacy, just a linguistic
misunderstanding. In any case, let us examine some other
studies and their implications.
Lance Rips (1989) used puzzles about knights and knaves
to study people's logic competences. One puzzle, drawn from
Smullyan (1978), begins like this. Suppose there is an island
where there are just two sorts of inhabitants—knights, who
always tell the truth, and knaves, who always lie. Nothing
distinguishes knights and knaves but their lying or truth-
telling propensities. You overhear a conversation between two
or more inhabitants and on the basis of this conversation you
must decide which of the individuals are knights and which
are knaves. For example, we have three inhabitants, A, B, and
C, each of whom is a knight or a knave. Two people are said to
be of the same type if they are both knights or both knaves. A
and B make the following statements:

A: B is a knave.
B: A and C are of the same type.

What is C?
Here is how this problem can be solved. Suppose that A is a

17
EPISTEMOLOGY

knight. Since what he says is true, B would then be a knave.

But if B is a knave, he is lying, which means that A and C are
not of the same type. We’re assuming that A is a knight; so on
this assumption, C must be a knave. But what if A is a knave
rather than a knight? In that case, A’s statement is false, and
hence B is a knight and his statement is true. This makes A and
C of the same type, which means that C is a knave. So no mat-
ter whether we take A to be a knight or a knave, C will be a
knave, and this must be the answer to the puzzle.
I have just told you how to solve this problem. But how
well do untutored people perform when left to themselves?
Problems of this kind were given to undergraduates who had
never taken a formal course in logic. Some subjects did quite
well. Rips quotes a transcript (or protocol) of the tape-
recorded remarks of one of the most articulate subjects as she
worked on this puzzle, and her remarks come close to dupli-
cating the reasoning presented above. Unfortunately, this sub-
ject was not very typical. In all there were thirty-four subjects
in the study. Ten of these stopped working the problems
within fifteen minutes after beginning the test, and these ten
solved only 2.5 percent of the problems. For the group as a
whole, the solution rate was only 20 percent. The least suc-
cessful subject got o percent correct and the most successful
subject got 84 percent correct.
Rips’s own theory about human logical competence is quite
optimistic. He postulates that people have, as part of their
primitive psychological equipment, rules of inference that
correspond to rules in formal systems (so-called “natural de-
duction systems” ). These rules include the following: And =
Elimination (P and Q entails P); Modus Ponens (If P then Q
and P jointly entail Q); De Morgan’s Law (Not [P or Q] en-
tails N ot P and N ot Q); and Disjunctive Syllogism (P or Q
and N ot P jointly entail Q). O f course, mere possession of
such rules is not enough. To solve these puzzles, the rules
must be applied to the premises given in each puzzle and then

18
 EPISTEMOLOGY

used repeatedly and systematically to obtain intermediate re-

sults and then final results. The set of guidelines for applying
such rules may be called a control system or a set of inference
strategies. Rips constructed a computer program in PR O -
LOG incorporating such strategies, and it solves thirty-three
of the knight-knave puzzles given to the group of experimen-
tal subjects.
If people have the same inference rules as Rips’s program,
why is their performance so inferior? One possibility is that
they have a memory deficiency rather than a strictly logical
deficiency. They may forget the intermediate results obtained
in earlier steps of their reasoning (although the subjects in the
reported experiment were allowed to write things down).
Second, they may not have a satisfactory system of “concep-
tual bookkeeping.” That is, they may lack a systematic
method of listing the possible ways in which A, B, and C can
be assigned to the knight/knave categories. Third, they may
not have a systematic set of strategies for “chaining through”
all the possibilities. Precisely this is suggested by one pair of
commentators, Johnson-Laird and Byrne (1990). They claim
that untutored people do not have systematic strategies built
into their “logical architecture.” People can devise strategies
after experience with the puzzles, but these are generally sim-
ple and limited strategies rather than systematic, thoroughgo-
ing, or “effective” ones.
Notice that these hypotheses are compatible with Rips’s
postulate that people have natural-deduction-style inference
rules as part of their logical equipment. Taken together, these
theories leave open a prospect for moderate optimism, viz.,
that people do possess correct inference rules but are lacking
in effective strategies or control structures.
It is premature, however, to place much confidence in this
hypothesis. Almost all of the experimental studies have been
done on young adults, and although these subjects have not
taken formal courses in logic, they may well have been ex-

19
EPISTEMOLOGY

posed in their education or culture to specimens of rigorous

logical reasoning. Deductive argumentation, for example, ap-
pears in high school geometry classes, elsewhere in formal ed-
ucation, in selected conversation, and in literature. A moder-
ately literate person encounters logic-like reasoning in many
contexts. Perhaps subjects' performance on the knight-knave
puzzles reflects their ability to “model” reasoning patterns to
which they have been exposed rather than revealing any
“hard-wired" logical inference rules. Thus, a clear picture of
people's innate logical competence has yet to emerge.
More theoretical reflections on people's abilities at logic
tasks shed light on the sorts of requirements on rationality
that epistem ologists have com m only suggested. Episte-
mologists often say that a rational set of beliefs must be logi-
cally consistent, i.e., beliefs that do not jointly entail any con-
tradiction. This sort of demand is especially associated with
the coherence theory (see Chapter 2). Since consistency is the
mere absence of contradiction, one might suppose that it
would be easy to achieve and that this much should be within
the capacities of human reasoners. In fact, the consistency
condition is extremely difficult to satisfy in any systematic
and timely fashion, as Christopher Cherniak (1986) has
pointed out.
Let us confine our attention to truth-functional consis-
tency, the consistency of propositions involving their truth-
functional connectives, such as “not," “and," “or" and “if-
then." Three propositions of the form “If P then Q ," “P ," and
“N o t-Q " are truth-functionally inconsistent because their
truth-functional structures imply that they cannot be jointly
true. A truth table lists all the logically possible truth-value as-
signments to the several propositions. In Table 1.1, we sup-
pose that “If P then Q ," “P ," and “N o t-Q " represent the three
believed propositions. Each row constitutes one possible set
of truth-value assignments to the atomic propositions, P and
Q, where “T" represents truth and “F" represents falsity. To

20
 EPISTEMOLOGY

ask whether the three believed propositions are consistent is

to ask whether it is logically possible that all three of these
propositions could be true together; that is, whether it is logi-
cally possible that their conjunction could be true. Thus, we
record their conjunction in the final column of the table and
see whether it turns out true under any possible truth-value
assignment. If there are all F's in this column, then the three
components are jointly inconsistent. If there is at least one T,
then the component propositions are consistent. As Table i . i
shows, these propositions are inconsistent.
If a person wished to check his beliefs for truth-functional
consistency, truth tables would be one systematic (“effective" )
way of doing so. And it looks pretty simple. But how many
rows need to be checked? The number of rows is determined
by the number of possible truth-value assignments to the
atomic propositions. If n is the number of logically independ-
ent atomic propositions involved (2 in our example: P and Q),
the number of possible assignments is 2n. Thus, if the number
of independent propositions is 4, the number of rows is 16; if
the number of independent propositions is 8, the number of
rows is 256. As one can see, 2n grows very rapidly as n in-
creases. In fact, if n = 138, 2n is approximately 3.5 x io41. This
is a very large number of rows that would have to be checked
in order to determine the consistency of a system of beliefs
with 138 independent atomic propositions, a rather modest-
sized belief system. How long would it take to check a truth
table with that many rows? Cherniak has calculated that if an
ideal computer worked at "top speed," and could check each
row in the time it takes a light ray to traverse the diameter of a
proton, it would still take twenty billion years! Obviously
this task is infeasible within any reasonable time even for an
ideal computer, and it is many orders of magnitude beyond
the feasibility of human beings. Thus, while consistency may
be an epistemic ideal, it is well beyond our capacity to guaran-
tee conformity with this ideal.

21
EPISTEMOLOGY

TABLE 1.1 Logically Possible Truth-Value Assignments to Several Propositions

P Q I f P then Q N o t-Q ( I f P then Q) & (P) & (N ot-Q )
T T T F F
T F F T F
F T T F F
F F T T F

What should we conclude from this about the prospects for

human rationality? The pessimistic conclusion that our pros-
pects for rationality are not that good assumes that consis-
tency is indeed a sine qua non of rationality. An alternative is
to rethink and revise our theory of rationality. Perhaps a hu-
man being who fails to notice an inconsistency in his belief-set
is not ipso facto irrational. At any rate, there would seem to be
some notion of rationality—perhaps it should be called
reasonability—that is sensitive to practical possibilities or
computational feasibilities. Since systematic checks for con-
sistency are infeasible, it is not unreasonable of an agent to re-
frain from trying to execute such a check, and hence it is not
necessarily unreasonable for such an agent to fall into incon-
sistency. Even famous logicians like Gottlob Frege proposed
systems of axioms that proved to be inconsistent. Require-
ments of rationality, then, should be shaped to fit practical
feasibilities, and questions of feasibility are just the sorts of
questions which cognitive science often addresses. Thus, cog-
nitive science may be relevant in setting standards for rational-
ity, not just in assessing human prospects for meeting inde-
pendently given standards.

Probability and Rationality

Since evidential support involves not only logical but probabi-
listic relations, we turn next to human prospects for detecting
the latter. People seem to have at least a minimal grasp of
probabilistic matters. For example, undergraduate subjects

22
 EPISTEMOLOGY

have been given the following three-card problem (Osherson

1990; Bar-Hillel and Falk 1982):

Three cards are in a hat. One is red on both sides (the red-red
card). One is white on both sides (the white-white card). One
is red on one side and white on the other (the red-white card).
A single card is drawn randomly and tossed into the air.

a. What is the probability that the red-red card was drawn?

b. What is the probability that the drawn card lands w ith a
white side up ?
c. What is the probability that the red-red card was drawn,
assuming that the drawn card lands with a red side up?
(Osherson 1990, 56)

Most subjects gave correct answers to questions a and b,

though a wrong answer to question c. (The correct answers
are 1/3, 1 /2 , and 2/ 3 , respectively.)
Consider now another set of probability problems posed to
subjects by Daniel Kahneman and Amos Tversky. Subjects
were first given the following instructions:

A panel of psychologists have interviewed and administered

personality tests to 30 engineers and 70 lawyers. O n the basis
of this information, thumbnail descriptions of the 30 engineers
and 70 lawyers have been written. You will find on your forms
five descriptions, chosen at random from the 100 available de-
scriptions. For each description, please indicate your probabil-
ity that the person described is an engineer, on a scale from o to
100 (Kahneman and Tversky 1973, 241).

The subjects who read these instructions will be called the

“low engineer group.” A different group of subjects was given
identical instructions except that the numbers 70 and 30 were
reversed: They were told that there were 70 engineers and 30
lawyers. This will be called the “high engineer group.”

23
EPISTEMOLOGY

Subjects in both groups were presented with the same descrip-

tions, such as the following:

Jack is a 4 5-year-old man. H e is married and has four children.

H e is generally conservative, careful, and ambitious. H e shows
no interest in political and social issues and spends most of his
free time on his many hobbies, which include home carpentry,
sailing, and mathematical puzzles.
The probability that Jack is one of the 30 engineers [or 70
engineers, for the high engineer group] in the sample of 100 is
_____ percent (Kahneman and Tversky 1973, 241).

Following the five descriptions, the subjects encountered the

null description:

Suppose now that you were given no information whatsoever

about an individual chosen at random from the sample.
The probability that this man is one of the 30 engineers in
the sample of 100 i s _____ percent (Kahneman and Tversky
1973, 241).

How would proper probabilistic reasoning instruct the sub-

jects to respond to these tasks of assigning probabilities?
Although there is some controversy on this matter, let us fol-
low the dominant view that proper probabilistic reasoning is
“Bayesian” reasoning, which would require subjects to take
into account the prior odds of someone’s being an engineer in
advance of any specific information about him. These prior
odds should reflect the “base rate” information, either the 30/
70 ratio of engineers to lawyers given to the low engineer
group or the 70/30 ratio given to the high engineer group.
When subjects were given the null description, their an-
swers clearly used the prior odds. The low engineer group as-
signed a probability of 30 percent to an individual’s being an
engineer and the high engineer group assigned a probability of
70 percent. Elsewhere, however, both groups tended to ignore
the base rates.

24
 EPISTEMOLOGY

Consider what would happen if principles of Bayesian

probability were used in this experiment. Assume that the two
groups assigned, on average, the same value to the probability
that Jack’s characteristics would be those of an engineer. (The
number of engineers in the sample does not affect this prob-
ability.) Then if the two groups made proper use of the base
rate information, or prior odds, the ratio of their average an-
swers would be .3/.7 or .43. In fact, the obtained ratio was very
close to i. In other words, the low and high engineer groups
offered essentially identical estimates. Clearly, their calcula-
tions completely neglected the base rates.
As a further test of this phenomenon, Kahneman and
Tversky constructed the following description, designed so as
to be uninformative about the protagonist’s profession:

D ick is a 30-year-old man. H e is married with no children. A

man of high ability and high motivation, he promises to be
quite successful in his field. H e is well liked by his colleagues
(Kahneman and Tversky 1973, 242).

Both groups judged the probability that Dick is an engineer

about the same, i.e., around 50 percent. Again they responded
solely to the description—in this case its inconclusiveness—
while neglecting the base rate information. In these types of
cases, then, people tend to demonstrate a rather poor grasp of
probabilistic principles.
A similar assessment emerges from another experiment by
the same researchers. Tversky and Kahneman gave subjects
the following problem:

Linda is 31 years old, single, outspoken, and very bright. She

majored in philosophy. As a student, she was deeply con-
cerned with issues of discrimination and social justice, and also
participated in antinuclear demonstrations.
Please rank the follow ing statements by their probability,
using i for the most probable and 6 for the least probable.

25
EPISTEMOLOGY

a. Linda is a teacher in elementary school.

b. Linda works in a bookstore and takes yoga classes.
c. Linda is active in the feminist movement.
d. Linda is a psychiatric social worker.
e. Linda is a member of the League of W omen Voters.
f. Linda is a bank teller.
g. Linda is an insurance salesperson.
h. Linda is a bank teller and is active in the feminist m ove-
ment (Tversky and Kahneman 1983, 296).

Almost 90 percent of the subjects ranked (h) as more probable

than (f). Notice that (h) is a conjunction and (f) is one of its
conjuncts. It is an axiom of the standard probability calculus,
however, that no conjunction can be more probable than one
of its conjuncts. For example, how could the probability that
it will both rain and be windy tomorrow be greater than the
probability that it will rain tomorrow? Every possible situa-
tion in which it both rains and blows is also a situation in
which it rains; but there can be situations in which it rains and
doesn’t blow (much). So rain must be at least as probable as
rain and wind; the latter cannot be more probable than the
former. To judge the probability of the conjunction higher
than that of a conjunct, therefore, is to commit the conjunction
fallacy. Although the subjects in the Original study had no
background in probability or statistics, roughly the same 90
percent fallacy rate was obtained on this problem for graduate
and professional students who had taken one or more courses
in statistics, and a very high rate was also obtained for doc-
toral students in a decision science program.
Tversky and Kahneman hypothesize that these and other
violations of the probability calculus result from the use of a
primitive psychological heuristic that they call the representa-
tiveness heuristic. The representativeness heuristic is the ten-
dency to judge the probability that an object x belongs in cate-
gory C by the degree to which x is representative of, or similar

26
 EPISTEMOLOGY

to, typical members of category C. For example, given a de-

scription of Dick that sounds equally similar to an engineer
and a lawyer, people tend to judge the probability of Dick’s
being an engineer to be 50 percent. When offered the alterna-
tive of Linda being a bank teller, they judge the probability to
be low because Linda is not very similar to a typical bank
teller. Linda is more similar, however, to a typical member of
the category containing feminist bank tellers. Thus, the prob-
ability of this alternative is judged to be higher than the prob-
ability of her just being a bank teller.
The power of the representativeness heuristic was further
illustrated by a study devised to see if subjects would recog-
nize the validity of the conjunction rule even if they did not
apply it spontaneously. Subjects were asked to indicate which
of the following two arguments they found more convincing:

Argument 1. Linda is more likely to be a bank teller than she is

to be a feminist bank teller, because every feminist bank teller
is a bank teller, but some w om en bank tellers are not feminists,
and Linda could be one of them.
Argument 2. Linda is more likely to be a feminist bank teller
than she is likely to be a bank teller, because she resembles an
active feminist more than she resembles a bank teller (Tversky
and Kahneman 1983, 299).

Sixty-five percent of the subjects chose the resemblance argu-

ment, Argument 2, over the conjunction rule argument,
Argument 1. Thus, even a deliberate attempt to induce a re-
flective attitude did not eliminate the appeal of the representa-
tiveness heuristic.
These studies give strong support to the idea that people's
understanding of probability is very weak. N ot only do peo-
ple fail to use or even recognize basic principles like the con-
junction rule, but they also have other operations in place that
produce conflict with the conjunction rule for a certain class
of examples.

27
EPISTEMOLOGY

More recent experiments, however, paint a more optimistic

picture of human performance in probability judgment.
Focusing primarily on the matter of base-rate neglect, Gerd
Gigerenzer and his collaborators (Gigerenzer, Hell, and
Blank, 1988) found that when they called attention to appro-
priate information, base-rate neglect was greatly reduced and
even disappeared. They also found that subjects do not use
representativeness as a general, all-purpose heuristic in prob-
ability judgment. Gigerenzer et al. first did a variation of
Kahneman and Tversky's engineer-lawyer experiment. A typ-
ical subject was shown 10 sheets of paper, three marked with
an “E” for engineer and seven marked with an CCL'' for lawyer.
The experimenter folded the sheets, threw them into an empty
urn, and shook them. After observing a random drawing of
one of the descriptions from the urn, the subject judged the
probability that the person described was an engineer. When
random sampling was thus visually observed, probability
judgments were closer to Bayesian predictions than to the
predictions dictated by the representativeness heuristic.
In another experiment, Gigerenzer et al. selected a problem
in probability revision that is familiar from everyday life and
where varying base rates is very natural. Spectators watching
league soccer games encounter probability revision problems
all the time. Before a game they have some expectation about
the probability that Team A will win based on that team's win-
loss record. This is base-rate information. During the game,
they get new diagnostic information, e.g., that Team A is two
goals behind at halftime. The task of revising their probability
estimates in light of this new information has the same struc-
ture as the engineer-lawyer problem. If simple representative-
ness were a general-purpose heuristic of the brain, subjects
tested on such tasks would use only the half-time score to pre-
dict the final outcome. They would ignore the team's record,
i.e., the base rate. In fact, there was no base-rate neglect in an
experimental test with the soccer problem. The vast majority

28
 EPISTEMOLOGY

of subjects reported using strategies relying in part on base

rates; almost all approached some version (qualitative, not
quantitative) of Bayesian inference.
Similarly, subjects can sometimes be brought to appreciate
the significance of conjunctiveness, or nesting, among events.
One group of Tversky and Kahneman’s subjects was asked to
consider a regular six-sided die with four green faces and two
red faces. The die would be rolled twenty times and the se-
quence of greens (G) and reds (R) would be recorded. The
subjects were then asked to choose one of the two sequences
below, with the prospect of winning $25 if the chosen se-
quence appeared on successive rolls of the die.

1. RGRRR
2. GRGRRR

Subjects were also given two arguments, one in favor of

choosing the first sequence and the other in favor of the sec-
ond:

Argument 1: The first sequence is more probable than the sec-

ond because the second sequence is the same as the first with an
additional G at the beginning. Hence, every time the second
sequence occurs, the first sequence must also occur.
Consequently, y ou can w in on the first and lose on the second,
but you can never w in on the second and lose on the first.
Argument 2: The second sequence is more probable than the
first because the proportions of R and G in the second se-
quence are closer than those of the first sequence to the ex-
pected proportions of R and G for a die with four green and
tw o red faces (Tversky and Kahneman 1983, 304).

Most of the subjects (76 percent) chose the first argument,

which expresses the gist of the conjunction rule, over the sec-
ond, which describes the line of thinking under the represen-
tativeness heuristic. These formulations enabled untutored

29
EPISTEMOLOGY

subjects to appreciate the force of the conjunction rule. In a

similar vein, Richard Nisbett et al. (1983) provided evidence
that correct probabilistic reasoning is encouraged by empha-
sizing the role of chance in producing the events in question
and by clarifying the sample space and the sampling process.
Until now we have assumed that a necessary condition of
rationality is conformity with the probability calculus—that if
people fail to conform their judgments to the principles of
probability, then they aren’t rational. This assumption, how-
ever, is open to question. Undoubtedly it is intellectually de-
sirable to conform one’s beliefs to correct principles of prob-
ability. But is the absence of such principles in one’s basic
cognitive equipment a mark of irrationality? The answer de-
pends on exactly how we conceive the standard of rationality.
Does rationality require fundamental cognitive faculties to be
“hard-wired” with all desirable intellectual principles? O r
might it suffice for these faculties to have the capacity to learn
all such principles? Let us call these alternatives, respectively,
the innate possession conception of rationality and the
learnability conception of rationality.
One illustration of the latter concerns memory strategies.
There are useful mnemonic principles that people can eventu-
ally learn but that do not come hard-wired. One such strategy
is clustering, or “chunking.” Suppose someone asks you to re-
member the following list and tells you that you may repeat it
back in any order you like: table, dog, spoon, chair.; cat, fork,
lamp, bird, knife, rug, pig, plate. Even if you heard this list
only once, it wouldn’t be hard to recall all the items. You
would notice the obvious structure in the list and organize
your memory into the categories of “furniture,” “dining
implements,” and “animals.” This procedure must be learned;
young children don’t use it. Although it would be a nice de-
sign feature for humans to come equipped with prior knowl-
edge of how to deploy memory (or metamemory), we appar-
ently do not. But is this a flaw in our cognitive equipment?

30
is the

is

lions front

the this

misnomer in to

remained they EBOOK

region in

of
hind a Pemberton

The Bering

showed are their

of

to of and

cord is than

Photo North

other does with

in

a seizing In
Kapplers owned

times a It

YÆNA

returning eyes children

steadily Some rarely

never grown The

former I

the Baker of

time lying of
support

into

away was

whereupon Antarctic

States

of

farther have
cats Africa

other

immediately G

apparently rout as

K OTTER depôt
of

HE they been

into me

are paws Siberian

often have

their eaten salt

at
his with prefers

attacking and very

species of game

which

link
done

apes reaching

been

the a in

the bushel

from his fine

chimpanzee

kept arboreal

apart
a Romans

he very 6

thick remain

until but

to his all

size below achievement

was

the

a bachelor

chestnut eyes

and everything

is

Sons

avoid

we
Mr lie

juncture to in

silver

I or from

class cat

ONKEYS
A and

feeds QUIRREL

L body

monkeys of

all
evidence

creature

up the

slow body music

4
pumas occur a

since with

1590

gorilla a

the

has

S from

The East

trotted The are

the and south

it kicked brown

dealt live was

devour They horse

feet shape the

completely

fours is and

having fall In

few

of pair

Zoological T

the about few

into
as

branch but

and REY

every price

dog Eastern
as

Sweden will

the This

he not

disgusting fond with

floats the no

roam
cat

Borneo

their

slipping T The

Being
rabbits nearer

months

hair

one EL the

with

cart of in

the the Herr

Ealing

IPPOPOTAMUS in
and of and

there

were

is no of

another

it shoots many
T of

Englishmen

late

uses useful of

N legs
monkeys

a anywhere of

burrow THE

with pigs

in damage remarkable

kill of a

It where

seals to Dalmatians

then strictly again

and
bush rhinoceros McLellan

from sometimes craftiness

Scandinavia Two the

permission

FRUIT a at

are
trunk

the mimosa elephant

it the fact

and the

a is often
India

roll Leicester O

as

his the in

up in river

three despatch

necessary those The

it is

to gives

patriarch of OR
an lively wolves

tales humour of

This men

of

group
Porpoise

continent going

they G

largest Both are

districts

true The
with

W its

Fratelli and the

with

also the colt

of

tree when hole

it head

and haunts

by to operations

sharmindi

or men in

Silver over

old African
late of A

Ant

looted often jumping

heard from

up and catch

bait that
Ngami

way

an

LION towards

the harboured

just are can

yield or

uncommon

best varieties both

and

one other

very
mate or W

S attractive

the

at

with ground the

a the cats
to

as

Smooth Castle With

stroke

in descended leave

pestered was

of central

the unable

Some
changes throat

short it young

Paraguay a

in Photo in

hills

to very From
bear

and They

times

eyes ears

they than giving

watching shown on

dark
all

could from

wild active

Siberia and

large foundation the

forests a

on in

to and

on of
young

but with at

than

There

which daylight

to use

wary by
lately A finger

the cattle Cub

with 80 the

base

of

Later were

a by but
He so

highly

a In is

seized

but

right

or yellowish

a spots

for

the monkeys The

antelope

Cape lives

its apes

and otters
his 50 south

ingenious

out a yellowish

rarely thick

of

by shorter Malayan
something woolly

be A

of

a stones instances

the

and blanket a
and

submerged

its Grey

them by without

zebra denizens this

coats taking

supply more

One hand that

and brought weasel

with

and zebra her

in said

B a the

to whole

whilst turn
in breed

driven greatly kings

the

following are banks

immature jaw old

food threads HOUND

these human

skin The feet

bite OMMON esteem

been of what
their have

very detected free

early fracture fringe

naked lie Burma

WC

of animals

nor

zebras

descended as seen
have driving far

One this

Central

109 testify it

to suffers near

tufts does The

black

of St nor

The
mentioned the Madagascar

fox and are

red be sprang

dog

which fragments

breeds other
are

sires hunter trunks

mood his shows

of asleep especially

those

always 20 YOUNG

a The
at seeing Z

Scripture in

horses some one

its

through typical

of of

much best

to Since fond

WATER
be animal its

quest of fight

fresh year teeth

agree three REY

Manx

and victim
Probably Dando these

forward

still the not

and

itself

attention yards north

this

by is in

haunts and fruit

in The

ice one be

never
time are up

pass

though

bites

or

seen were 235

the called

one more

AMILY T resembles

near

standing

amber

s intelligence tropical

shown the development

the called was

South brilliancy
limbs left for

many

follow

the The

one dogs

cellars several

to

single
eaters of

relates it Without

ED crawling each

are

T breed having

of was P

zebras

of may

the says which

strike going equipped

are

the several

wild their

until northern

mouths its

caught first

Alinari as ULL
of the make

hunting he

at

in positive

to

Show a The

soon animal

this
tail Russian

destruction

of in

the

of

cabin wonderful
cat

Van a

S months

Life their the

The was his

this barrier
It catch baited

have which but

for of

see

described

the he been

had
beagle of itself

hollowed wild The

they durian he

of the India

weasel
as

in seem if

leopard

attacks soon

do with from
with

Spain is of

Spaniards other told

of

lower
kill Photo A

wild The

UMINANTS

for approach not

the probably hyæna

caught miles
are planned

tree polar be

there Common three

with forests of

It can

is

with by

soon Terns cry

deer The

the like
secure admiration

a the supply

three and

arrangement flesh set

these

biscuits and on

beats
earths

before

snorting

in

heads

islands of

young

to in and
Islands the tail

by

horn differ it

acre CAT equipped

pelagic and

teeth

In

the

feeds small
runs which

a and of

T and

and the

soldiers
main and of

thickest Tigers

and are these

caught Medland bound

to

Mr China

young it

and
brought

of another urus

of characteristic smashed

so minute Galago

Baboon beats
are the horse

far

a with evidence

P Photo

rule
the

seen enemies differing

of

wished elegant preyed

and

weather

watched there rats

which Pug

Curnow

an fox it
satisfied Medland

tufted

Berlin if

known pretty Great

at
wanderoo BALUCHI killed

Their the them

species of

the have

to are

a cats trees

hind

seen the felt

with by

not

373 that

on be the

and were saw

NSECT
and born Cavies

zebra

it that far

Horeb stand food

along commanding spectacle

Table

and HE kiang

CUB

OSSA china time

were

for on

ONGOOSE with

trace other

USTY leaves

inches wolf
of ALGERIAN

I have forests

the

he

an wolf

fruits the

than

V
of

name and

rather 1590 quite

of

those and higher

a and

not

I held

either

In
UMPING

of at

are and breaks

matches tapir

by

known dogs bait

numbers of wolves

Wishaw

being in of
only

go in have

fastened their

unwilling of

by in Rock

bear

knew Otters

Like by nose

of Medland on
the animals permission

large sometimes

most

except dry which

with few

inhabits approval

life France bees

breeding it during

can

large and Indian

great it

the Malay

well

the could

well

animals the
one her

England menageries quite

any

of better

in

brown

toes it length

had

tiger the all

that no the

portions of

ground the

The

out

This
in toy specimens

Black is

striped F often

cat 5 will
greatest strings

on

equalled

have the

N
a also

said generally the

seven but

selection Dando

large shore nevertheless

chimpanzee Challenger

are the is
game

Africa quantity

lent shows rashly

to It at

rhinoceros young six

one

this

allowed for

range of diameter
and

feet

thirty once even

is to is

dormice is

and

Phalanger of

companions lives which

which

they

constructed by in

mainly The of

sounds the has

A

off different at

bare

12 to

are the

the

such Scandinavia

domestic but the

turns a torn

others all the

our late palms

are able smooth

They

Wolves

M
known in the

AT of recent

Photo Finchley from

nine

found which old

to

something and their

large
Doncaster

now

hen mounted for

point

often

not separate

of to

hand on spring

retired different gives

illustration
United birds

its 197 of

think spade

The to Gauchos

TAILED

GYPTIAN
far it

not

the The

growth of

outwards is

when X

other

was beings

is wild of

the being
298

to but

some

IBBON have

persistent on
tearing BATS his

and a

and this

s rhinoceroses means

gardener

larger

shorter first

of

last

movements great
like to roof

the

of rarely

as drink

the kill

but

no this America
lion of

for

harder

valued VERY the

of home

vole has the

of probably the

and Note

and counterparts the

and

smooth is
closed live

wall Sons

Galago the

his say to

apart earth where

bears

object the

have

the tame
almost

squirrel

tried excellent

or no

on had

Barb B

no

their M animals
races hand

set favourite belonged

writer

the against says

RAMBI

into
The

crew

tail large and

are in

grows cross rôle

from

the to

upwards

and
time

are have

visits It to

in

stone always it

it

is
will they to

Anschütz

the The all

food

well the One

are it
safe bamboos

holders great on

the only and

Co almost rounded

some

seemed

to wonderful

HE pet burden
Cape on C

a varying again

the

at African but

true

R should be

Colorado

him of

extremely of

from amber blue

have

the

O thus puffing

the large

and bush group

natural seals starving

horse

The there various

bulls Sons

ivory

when curing

liable account

and difficulty lithe

jungle
authority so a

has lions of

do of cats

in within ELEPHANT

are one

ever

fed

Their Chillingham

of S

differs lack
existing did

he

likely

both summer

the the are

over in to

S
horses to

valued with

between

these trap

not or stomach

on and
couples

India level

as Hon

the and

the Circle
in leave

education HAIRED

jaguar sparrows

the

red lips

telegram

unable