Hi there,
Kindly check the instructions.
Subject to argue : Social media impact on young adults
·
IMPORTANT NOTE: this assignment is intended to be performed over the course of modules ‘The Building Blocks of Arguments’ and ‘Extended Argumentative Reconstruction’. I recommend writing the essay first and then doing the analytic steps as you get that material under your belt. ‘The Building Blocks of Arguments’ will give you what you need for the
annotations and ‘Extended Argumentative Reconstruction’ will prepare you for
arranging the argument in Standard Form and Diagram.
· Writing Assignment 1:
· Start by just
composing a 500 word essay arguing for some position (any position!) you want to (on any issue!).
· Advice/Request: make it something you honestly care about please! A good idea is to just use something you’ve been thinking about recently, or perhaps had a conversation about. Also, you can write SUPER casually. This isn’t a formal academic paper you’re putting together.
· After you are done (don’t do this along the way!),
annotate your own argument by looking for the following elements: Reason Markers (RM), Conclusion Markers (CM), Assuring (A), Discounting (D), Guarding (G), Positive Evaluative Terms (E+), Negative Evaluative Terms (E-).
· I want to see every use of these terms identified, not just a couple. Really pour over your writing and see what you naturally got!
· And finally
arrange your argument in Standard Form and
diagram the argument.
· Do NOT use the copy/paste method! Listen for the main appeals your argument is making – what are the IDEAS you are presenting?
· Build the structure as you go – start with the conclusion and work your way backwards through the lines of support.
· I will grade for
effort. To judge this I will look at 1) length and 2) richness (ex: Is there a lot of different points, or just one or two dragged out to 500 words?) This is also where I’ll be checking to make sure your essay is actually making arguments!
· 2 total points
· I will grade for
skill. I will be looking at your analysis and seeing how exhaustively you were able to analyze your own argument.
· 1.5 total points for the analysis
· 1.5 total points for the arranging in Standard Form/Diagramming
·
Understanding
Arguments
This page intentionally left blank
Walter Sinnott-Armstrong
Dartmouth College
Robert J. Fogelin
Dartmouth College
Understanding
Arguments
An Introduction to Informal Logic
EIGHTH EDITION
Australia • Brazil • Japan • Korea • Mexico • Singapore • Spain • United Kingdom • United States
Understanding Arguments:
An Introduction to Informal Logic,
Eighth Edition
Walter Sinnott-Armstrong and
Robert J. Fogelin
Publisher/Executive Editor: Clark Baxter
Senior Sponsoring Editor: Joann Kozyrev
Associate Media Editor: Diane Akerman
Assistant Editor: Nathan Gamache
Editorial Assistant: Michaela Henry
Marketing Manager: Mark Haynes
Marketing Coordinator: Josh Hendrick
Marketing Communications
Manager: Kim Soltero
Project Manager, Editorial
Production: Abigail Greshik
Creative Director: Rob Hugel
Art Director: Faith Brosnan
Print Buyer: Betsy Donaghey
Permissions Editors: Scott Bragg/
Karyn L. Morrison
Production Service: Pre-PressPMG
Photo Researcher: John Hill
Copy Editor: Pre-PressPMG
Cover Designer: RHDG
Cover Image: Getty Images/Stella Salazar
Compositor: Pre-PressPMG
For product information and technology assistance,
contact us at Cengage Learning
Academic Resource Center, 1-800-423-0563.
For permission to use material from this text or
product, submit all requests online at
www.cengage.com/permissions.
Further permissions questions can be emailed to
permissionrequest@cengage.com.
© 2010, 2005 Wadsworth, Cengage Learning
ALL RIGHTS RESERVED. No part of this work covered by
the copyright herein may be reproduced, transmitted,
stored, or used in any form or by any means graphic,
electronic, or mechanical, including but not limited to
photocopying, recording, scanning, digitizing, taping, Web
distribution, information networks, or information storage
and retrieval systems, except as permitted under Section
107 or 108 of the 1976 United States Copyright Act,
without the prior written permission of the publisher.
Library of Congress Control Number: 2008939692
ISBN-13: 978-0-495-60395-5
ISBN-10: 0-495-60395-3
Wadsworth Cengage Learning
10 Davis Drive
Belmont, CA 94002-3098
USA
Cengage Learning products are represented in Canada by
Nelson Education, Ltd.
For your course and learning solutions, visit
academic.cengage.com.
Purchase any of our products at your local college store or
at our preferred online store www.ichapters.com.
Printed in the United States of America
1 2 3 4 5 6 7 8 12 11 10 09
www.cengage.com/permissions
www.ichapters.com
To Eric, John, Lars, Miranda, and Nicholas
and the colleges of their choice
This page intentionally left blank
PREFACE xv
PART I HOW TO ANALYZE ARGUMENTS 1
Chapter 1
USES OF ARGUMENTS 3
What Arguments Are 3
Justifications 4
Explanations 7
Combinations: An Example 10
CHAPTER 2
THE WEB OF LANGUAGE 17
Language and Convention 17
Linguistic Acts 19
Speech Acts 22
Performatives 23
Kinds of Speech Acts 26
Speech Act Rules 28
Conversational Acts 32
Conversational Rules 34
Conversational Implication 37
Violating Conversational Rules 40
Rhetorical Devices 42
Deception 45
Bronston v. United States 46
Summary 48
CHAPTER 3
THE LANGUAGE OF ARGUMENT 51
Argument Markers 51
If . . . , then . . . 53
CONTENTS
vi i
v i i i
Contents
Arguments in Standard Form 55
Some Standards for Evaluating Arguments 57
Validity 57
Truth 59
Soundness 60
A Tricky Case 60
A Problem and Some Solutions 62
Assuring 63
Guarding 65
Discounting 66
Evaluative Language 69
Spin Doctoring 72
CHAPTER 4
THE ART OF CLOSE ANALYSIS 77
An Extended Example 77
Clerk Hire Allowance, House of Representatives 77
CHAPTER 5
DEEP ANALYSIS 105
Getting Down to Basics 105
Clarifying Crucial Terms 109
Dissecting the Argument 109
Arranging Subarguments 111
Suppressed Premises 116
Contingent Facts 117
Linguistic Principles 119
Evaluative Suppressed Premises 120
Uses and Abuses of Suppressed Premises 121
The Method of Reconstruction 122
Digging Deeper 125
An Example of Deep Analysis: Capital Punishment 127
PART II HOW TO EVALUATE ARGUMENTS:
DEDUCTIVE STANDARDS 139
CHAPTER 6
PROPOSITIONAL LOGIC 141
The Formal Analysis of Arguments 141
i x
Contents
Basic Propositional Connectives 142
Conjunction 142
Disjunction 150
Negation 150
Process of Elimination 153
How Truth-Functional Connectives Work 154
Testing for Validity 156
Some Further Connectives 160
Conditionals 162
Truth Tables for Conditionals 163
Logical Language and Everyday Language 169
Other Conditionals in Ordinary Language 172
CHAPTER 7
CATEGORICAL LOGIC 179
Beyond Propositional Logic 179
Categorical Propositions 180
The Four Basic Categorical Forms 182
Translation into the Basic Categorical Forms 184
Contradictories 187
Existential Commitment 189
Validity for Categorical Arguments 190
Categorical Immediate Inferences 192
The Theory of the Syllogism 194
Appendix: The Classical Theory 203
The Classical Square of Opposition 205
The Classical Theory of Immediate Inference 209
The Classical Theory of Syllogisms 210
PART III HOW TO EVALUATE ARGUMENTS:
INDUCTIVE STANDARDS 213
CHAPTER 8
ARGUMENTS TO AND FROM GENERALIZATIONS 215
Induction versus Deduction 215
Statistical Generalizations 219
Should We Accept the Premises? 220
Is the Sample Large Enough? 220
Is the Sample Biased? 222
Is the Result Biased in Some Other Way? 223
Statistical Applications 225
x
Contents
CHAPTER 9
CAUSAL REASONING 231
Reasoning About Causes 231
Sufficient Conditions and Necessary Conditions 233
The Sufficient Condition Test 236
The Necessary Condition Test 237
The Joint Test 238
Rigorous Testing 240
Reaching Positive Conclusions 242
Applying These Methods to Find Causes 243
Normality 243
Background Assumptions 244
A Detailed Example 245
Calling Things Causes 249
Concomitant Variation 250
CHAPTER 10
INFERENCE TO THE BEST EXPLANATION AND
FROM ANALOGY 257
Inferences to the Best Explanation 257
Arguments from Analogy 267
CHAPTER 11
CHANCES 277
Some Fallacies of Probability 277
The Gambler’s Fallacy 277
Strange Things Happen 278
Heuristics 279
The Language of Probability 282
A Priori Probability 283
Some Rules of Probability 285
Bayes’s Theorem 291
CHAPTER 12
CHOICES 303
Expected Monetary Value 303
Expected Overall Value 306
Decisions Under Ignorance 308
x i
Contents
PART IV FALLACIES 315
CHAPTER 13
FALLACIES OF VAGUENESS 317
Uses of Unclarity 317
Vagueness 318
Heaps 320
Slippery Slopes 322
Conceptual Slippery-Slope Arguments 322
Fairness Slippery-Slope Arguments 325
Causal Slippery-Slope Arguments 327
CHAPTER 14
FALLACIES OF AMBIGUITY 333
Ambiguity 333
Equivocation 337
Definitions 343
CHAPTER 15
FALLACIES OF RELEVANCE 353
Relevance 353
Ad Hominem Arguments 354
Appeals to Authority 360
More Fallacies of Relevance 364
CHAPTER 16
FALLACIES OF VACUITY 369
Circularity 369
Begging the Question 370
Self-Sealers 375
CHAPTER 17
REFUTATION 381
What Is Refutation? 381
Counterexamples 382
Reductio Ad Absurdum 386
Straw Men and False Dichotomies 390
Refutation by Parallel Reasoning 392
x i i
Contents
PART V AREAS OF ARGUMENTATION 401
CHAPTER 18
LEGAL REASONING 403
Components of Legal Reasoning 404
Questions of Fact 404
Questions of Law 405
The Law of Discrimination 411
The Equal Protection Clause 411
Applying the Equal Protection Clause 412
The Strict Scrutiny Test 413
The Bakke Case 414
Regents of the University of California v. Bakke 416
Legal Developments Since Bakke 418
Grutter v. Bollinger 419
Gratz v. Bollinger 425
Burden of Proof 430
CHAPTER 19
MORAL REASONING 433
Moral Disagreements 433
The Problem of Abortion 434
The “Pro-Life” Argument 435
“Pro-Choice” Responses 437
Analogical Reasoning in Ethics 442
Weighing Factors 444
“A Defense of Abortion,” by Judith Jarvis Thomson 446
“An Argument that Abortion Is Wrong,”
by Don Marquis 459
CHAPTER 20
SCIENTIFIC REASONING 477
Standard Science 477
Scientific Revolutions 479
“Molecular Machines: Experimental Support for the
Design Inference,” by Michael J. Behe 481
“Living with Darwin,” by Philip Kitcher 494
x i i i
Contents
CHAPTER 21
RELIGIOUS REASONING 505
“Five Reasons to Believe in God,” by William Lane Craig 506
“Seven Deadly Objections to Belief in the Christian God,” by
Edwin Curley 512
CHAPTER 22
PHILOSOPHICAL REASONING 523
“Computing Machinery and Intelligence,” by A. M. Turing 524
“The Myth of the Computer,” by John R. Searle 536
Credits 543
Index 545
This page intentionally left blank
Traditionally, logic has been considered the most general science dealing
with arguments. The task of logic is to discover the fundamental principles
for distinguishing good arguments from bad ones.
For certain purposes, arguments are best studied as abstract patterns of
reasoning. Logic can then focus on these general forms rather than on par-
ticular arguments, such as your attempt to prove to the bank that they, not
you, made a mistake. The study of those general principles that make cer-
tain patterns of argument valid and other patterns of argument invalid is
called formal logic. Two chapters of this work are dedicated to formal logic.
A different but complementary way of viewing an argument is to treat it
as a particular use of language: Presenting arguments is one of the impor-
tant things we do with words. This approach stresses that arguing is a lin-
guistic activity. Instead of studying arguments as abstract patterns, it
examines them as they occur in concrete settings. It raises questions of the
following kind:
What is the place of argument within language as a whole?
What words or phrases are characteristic of arguments?
How do these words function?
What task or tasks are arguments supposed to perform?
When an approach to argument has this emphasis, the study is called infor-
mal logic. Though it contains a substantial treatment of formal logic, Under-
standing Arguments, as its subtitle indicates, is primarily a textbook in
informal logic.
The eighth edition of Understanding Arguments differs from the seventh
edition in a number of significant ways. The uses of arguments have been
brought right up front for emphasis. The chapters have been split up and re-
organized for clarity. Some of the more difficult and confusing topics have
been dropped to simplify and streamline the text. This edition also contains
new readings on scientific reasoning in Chapter 20 and on religious reason-
ing in Chapter 21. These new readings make the text more relevant to con-
temporary debates. Finally, this edition includes a large-scale updating of
examples, exercises, and discussion questions throughout the text, includ-
ing a liberal sprinkling of quotations from Jon Stewart and Stephen Colbert
PREFACE
xv
xv i
Preface
as well as Colin Powell’s argument that Iraq was seeking weapons of mass
destruction.
This new edition has been influenced by our teaching of this material
with various colleagues, including visitors, at Dartmouth College. In this re-
gard, we would like to thank three student assistants—Ben Rump, Jane
Tucker, and especially David Lamb—in addition to the many others who
helped us on previous editions. We are also indebted to the following re-
viewers: André Ariew, University of Missouri, Columbia; John Zillmer,
Michigan State University; K. D. Borcoman, Coastline College; Barbara A.
Brown, Community College of Allegheny County; Marina F. Bykova, North
Carolina State University; Eric Parkinson, Syracuse University; and Alison
Reiheld, Michigan State University. At Wadsworth and Cengage Learning,
we received expert advice and assistance from Worth Hawes, former Acqui-
sitions Editor for Wadsworth; Joann Kozyrev, Sponsoring Editor for Philoso-
phy and Religion; Sarah Perkins, Assistant Editor; Deborah Bader, copy
editor; and Abigail Greshik, Pre-PressPMG Project Manager. Finally, we owe
a great debt to Bill Fontaine (librarian) as well as Sarah Kopper and Kier
Olsen DeVries (research assistants). Without all of these people, this book
would contain many more mistakes than it undoubtedly still does.
Walter Sinnott-Armstrong
Robert J. Fogelin
How to Analyze
Arguments
Arguments are all around us. They bombard us constantly in advertisements; in
courtrooms; in political, moral, and religious debates; in academic courses on
mathematics, science, history, literature, and philosophy; and in our personal lives
when we make decisions about our careers, finances, and families. These crucial
aspects of our lives cannot be understood fully without understanding arguments.
The goal of this book, then, is to help us understand arguments and, thereby, to
understand our lives.
We will view arguments as tools. To understand a tool, we need to know the pur-
poses for which it is used, the material out of which it is made, and the forms that it
takes. For example, hammers are normally used to drive nails or to pound malleable
substances. Hammers are usually made out of a metal head and a handle of wood,
plastic, or metal. A typical hammer’s handle is long and thin, and its head is per-
pendicular to its handle. Similarly, in order to understand arguments, we need to
investigate their purposes, materials, and forms.
Chapter 1 discusses the main purposes or uses of arguments. The material from
which arguments are made is language, so Chapters 2–3 explore language in gen-
eral and then the language of argument in particular. Chapters 4–5 use the lessons
learned by then to analyze concrete examples of arguments in detail. The following
chapters turn to the forms of arguments, including deductive forms in Part II
(Chapters 6–7) and inductive forms in Part III (Chapters 8–12). Each form of argu-
ment comes with its own standards of adequacy. Part IV (Chapters 13–17) will then
consider the main ways in which arguments can go astray, including fallacies of
clarity, relevance, and vacuity. Finally, Part V (Chapters 18–22) will explore exam-
ples of arguments in different fields—law, morality, religion, science, and philosophy—
in order to see both how such arguments differ and how they share common features
of arguments in general. By the end of this journey, we should understand argu-
ments much better.
I
1
This page intentionally left blank
Uses of Arguments
What are arguments? In our view, arguments are tools, so the first step toward
understanding arguments is to ask what they are used for—what people are trying
to accomplish when they give arguments. This brief chapter will propose a defini-
tion of arguments and then explore two main purposes of arguments: justification
and explanation. Both justifications and explanations try to provide reasons, but
reasons of different kinds. Justifications are supposed to give reasons to believe their
conclusions, whereas explanations are supposed to give reasons why their conclu-
sions are true. Each of these purposes is more complicated and fascinating than is
usually assumed.
WHAT ARGUMENTS ARE
The word “argument“ may suggest quarrels or squabbles. That is what a
child means when she reports that her parents are having an argument.
Arguments of that sort often include abuse, name-calling, and yelling. That
is not what this book is about. The goal here is not to teach you to yell
louder, to be more abusive, or to beat your opponents into submission.
Our topic is the kind of argument defined by Monty Python in their justly
famous “Argument Clinic.” In this skit, a client enters a clinic and pays for
an argument. In the first room, however, all he gets is abuse, which is not
argument. When he finally finds the right room to get an argument, the per-
son who is supposed to give him an argument simply denies whatever the
client says, so the client complains that mere denial is different from argu-
ment, because “an argument is a connected series of statements to establish
a definite proposition.” This definition is almost correct. As we will see, the
purpose of an argument need not always be to “establish” its conclusion,
both because some conclusions were established in advance and because
many reasons are inconclusive. Nonetheless, Monty Python’s definition
needs to be modified only a little in order to arrive at an adequate definition:
An argument is a connected series of sentences, statements, or propositions
(called “premises”) that are intended to give reasons of some kind for a sentence,
statement, or proposition (called the “conclusion”).
1
3
This definition does not pretend to be precise, but it does tell us what
arguments are made of (sentences, statements, or propositions) and what
their purpose is (to give reasons).
Another virtue of this definition is that it is flexible enough to cover the
wide variety of arguments that people actually give. Different arguments
are intended to give reasons of very different sorts. These reasons might be
justificatory reasons to believe or to disbelieve some claim. They might, in-
stead, be explanatory reasons why something happened. They might even
be practical reasons to do some act. Because reasons come in so many kinds,
arguments are useful in a great variety of situations in daily life. Trying to
determine why your computer crashed, why your friend acted the way she
did, and whether it will rain tomorrow as well as trying to decide which
political candidate to vote for, which play to use at a crucial point in a foot-
ball game, where to go to college, and whether to support or oppose capital
punishment—all involve weighing and evaluating reasons.
It is inaccurate, therefore, to think of arguments as serving only one
single simple purpose. People often assume that you always use every
argument to make other people believe what you believe and what they did
not believe before hearing or reading the argument. Actually, however,
some arguments are used for that purpose, but others are not. To fully un-
derstand arguments in all their glory, then, we need to distinguish different
uses of argument. In particular, we will focus on two exemplary purposes:
justification and explanation.
JUSTIFICATIONS
One of the most prominent uses of arguments is to justify a disputed claim. For
example, if I claim that September 11, 2001, was a Tuesday, and you deny this
or simply express some doubt, then we might look for a calendar. But suppose
we don’t have a calendar for 2001. Luckily, we do find a calendar for 2002.
Now I can justify my claim to you by presenting this argument: The calendar
shows that September 11 was on Wednesday in 2002; 2002 was not a leap year,
since 2002 is not divisible by four; nonleap years have 365 days, which is 1
more day than 52 weeks; so September 11 must have been on Tuesday in 2001.
You should now be convinced.
What have I done? My utterance of this argument has the effect of chang-
ing your mind by getting you to believe a conclusion that you did not
believe before. Of course, I might also be able to change your mind by hyp-
notizing you. But normally I do not want to use hypnosis. I also do not want
to change your mind by manufacturing a fake calendar for 2002 with the
wrong dates. Such tricks would not satisfy my goals fully. This shows that
changing your mind is not all that I am trying to accomplish. What else do I
want? My additional aim is to show you that you should change your mind,
4
CHAPTER 1 ■ Uses of Arguments
5
Just if icat ions
and why. I want to give you a good reason to change your mind. I want my
argument not only to make you believe my conclusion but also to make you
justified in believing my conclusion.
The above example is typical of one kind of justification, but there are
other patterns. Suppose that I share your doubts about which day of the
week it was on September 11, 2001. Then I might use the same argument to
justify my belief as well as yours. Indeed, you don’t even need to be present.
If I am all alone, and I just want to figure out which day of the week it was
on September 11, 2001, then I might think in terms of this same argument.
Here the goal is not to convince anybody else, but the argument is still used
to find a good reason to believe the conclusion.
In cases like these, we can say that the argument is used for impersonal
normative justification. The justification is normative because the goal is to
find a reason that is a good reason. It is impersonal because what is sought
is a reason that is or should be accepted as a good reason by everyone ca-
pable of grasping this argument, regardless of who they are. The purpose
is to show that there is a reason to believe the conclusion, regardless of
who has a reason to believe it. Other arguments, in contrast, are aimed at
specific people, and the goal is to show that those particular people are
committed to the conclusion or have a reason to believe the conclusion.
Such individualized uses of arguments seek what can be called personal
justification.
There should be nothing surprising about different people having differ-
ent reasons. I might climb a mountain to appreciate the view at the top,
whereas you climb it to get exercise, and your friend climbs it to be able to
talk to you while you climb it. Different people can have different reasons
for the same action. Similarly, different people can have different reasons to
believe the same conclusion. Suppose that someone is murdered in the ball-
room with a revolver. I might have good reason to believe that Miss Peacock
did not commit the murder, because I saw her in the library at the time the
murder was committed. You might not trust me when I tell you that I saw
her, but you still might have good reason to believe that she is innocent, be-
cause you believe that Colonel Mustard did it alone. Even if I doubt that
Colonel Mustard did it, we still each have our own reasons to agree that
Miss Peacock is innocent.
When different people with different beliefs are involved, we need to
ask who is supposed to accept the reason that is given in an argument.
A speaker might give an argument to show a listener that the speaker has a
reason to believe something, even though the speaker knows that the audi-
ence does not and need not accept that reason. Suppose that you are an athe-
ist, but I am an evangelical Christian, and you ask me why I believe that
Jesus rose from the dead. I might respond that the Bible says that Jesus rose
from the dead, and what the Bible says must be true, so Jesus rose from the
dead. This argument tells you what my reasons are for believing what I
believe, even if you do not accept those reasons. My argument can be used
to show you that I have reasons and what my reasons are, regardless of
whether you believe that my reasons are good ones and also regardless of
whether my reasons really are good ones.
The reverse can also happen. A speaker might give an argument to show
a listener that the listener has a reason to believe something, even though
the speaker does not accept that reason. Suppose that you often throw loud
parties late into the night close to my bedroom. I want to convince you to
stop or at least quiet down. Fortunately, you think that every citizen ought
to obey the law. I disagree, for I am an anarchist bent on undermining all
governments and laws. Still, I want to get a good night’s sleep before the
protest tomorrow, so I might argue that it is illegal to make that much noise
so late, and you ought to obey the law, so you ought to stop throwing such
loud parties. This argument can show you that you are committed to its
conclusion, even if I believe that its premises are false.
Of course, whether I succeed in showing my audience that they have a
reason to believe my conclusion depends on who my audience is. My
argument won’t work against loud neighbors who don’t care about the law.
Consequently, we need to know who the audience is and what they believe
in order to be able to show them what reason they have to believe a
conclusion.
In all of these cases, arguments are used to show that someone has a
reason to believe the conclusion of the argument. That is why all of these
uses can be seen as providing different kinds of justification. The differences
become crucial when we try to evaluate such arguments. If my goal is to
show you that you have a reason to believe something, then I can be
criticized for using a premise that you reject. Your beliefs are no basis for
criticism, however, if all I want is to show my own reasons for believing the
conclusion. Thus, to evaluate an argument properly, we often need to deter-
mine not only whether the argument is being used to justify a belief but also
which kind of justification is sought and who the audience is.
6
CHAPTER 1 ■ Uses of Arguments
Write the best brief argument you can to justify each of the following claims to
someone who does not believe them.
1. Nine is not a prime number.
2. Seven is a prime number.
3. A molecule of water has three atoms in it.
4. Water is not made up of carbon.
5. The U.S. president lives in Washington.
6. The Earth is not flat.
7. Humans have walked on the moon.
8. Most bicycles have two wheels.
Exercise I
EXPLANATIONS
A different but equally important use of arguments is to provide explanations.
Explanations answer questions about how or why something happened. We
explain how a mongoose got out of his cage by pointing to a hole he dug
under the fence. We explain why Smith was acquitted by saying that he got
off on a technicality. The purpose of explanations is not to prove that some-
thing happened, but to make sense of things.
An example will bring out the difference between justification and expla-
nation. One person claims that a school’s flagpole is thirty-five feet tall, and
someone else asks her to justify this claim. In response, she might produce a
receipt from the Allegiance Flagpole Company acknowledging payment for a
flagpole thirty-five feet in height. Alternatively, she may put a stick straight
up into the ground, measure the stick’s length and its shadow’s length, then
measure the length of the flagpole’s shadow, and calculate the length of the
flagpole. Neither of these justifications, however, will answer a different ques-
tion: Why is the flagpole thirty-five feet tall? This new question could be
answered in all sorts of ways, depending on context: The school could not
afford a taller one. It struck the committee as about the right height for the
location. That was the only size flagpole in stock. There is a state law limiting
flagpoles to thirty-five feet. And so on. These answers help us understand
why the flagpole is thirty-five feet tall. They explain its height.
Sometimes simply filling in the details of a story provides an explanation.
For example, we can explain how a two-year-old girl foiled a bank robbery
by saying that the robber tripped over her while fleeing from the bank. Here
we have made sense out of an unusual event by putting it in the context of a
plausible narrative. It is unusual for a two-year-old girl to foil a bank robbery,
but there is nothing unusual about a person tripping over a child when run-
ning recklessly at full speed in a crowded area.
Although the narrative is probably the most common form of explanation
in everyday life, we also often use arguments to give explanations. We can
explain a certain event by deriving it from established principles and ac-
cepted facts. This argument then has the following form:
(1) General principles or laws
(2) A statement of initial conditions
(3) A statement of the phenomenon to be explained
The symbol “ ” is pronounced “therefore” and indicates that the premises
above the line are supposed to give a reason for the conclusion below the
7
Explanat ions
When, if ever, is it legitimate to try to convince someone else to believe something
on the basis of a premise that you yourself reject? Consider a variety of cases.
Discussion Question
line. By “initial conditions” we mean those facts in the context that, together
with appropriate general principles and laws, allow us to derive the result
that the event to be explained occurs.
This sounds quite abstract, but an example should clarify the basic idea.
Suppose we put an ice cube into a glass and then fill the glass with water to
the brim. The ice will stick out above the surface of the water. What will
happen when the ice cube melts? Will the water overflow? Will it remain at
the same level? Will it go down? Here we are asking for a prediction, and it
will, of course, make sense to ask a person to justify whatever prediction he
or she makes. Stumped by this question, we let the ice cube melt to see what
happens. We observe that the water level remains unchanged. After a few
experiments, we convince ourselves that this result always occurs. We now
have a new question: Why does this occur? Now we want an explanation of
this phenomenon. The explanation turns upon the law of buoyancy, which
says that an object in water is buoyed up by a force equal to the weight of
the water it displaces. This law implies that, if we put an object in water, it
will continue to sink until it displaces a volume of water whose weight is
equal to its own weight (or else the object hits the bottom of the container).
With this in mind, go back to the original problem. An ice cube is itself
simply water in a solid state. Thus, when it melts, it will exactly fill in the
volume of water it displaced, so the water level will remain unchanged.
We can now see how this explanation conforms to the argumentative
pattern mentioned above:
(1) General principles or laws (Primarily the law of buoyancy)
(2) Initial conditions (An ice cube in a glass of water filled to the brim)
(3) Phenomenon explained (The level of the water remaining
unchanged after the ice cube melts)
This explanation is fairly good. People with only a slight understanding of
science can follow it and see why the water level remains unchanged. We
should also notice that it is not a complete explanation, because certain things
are simply taken for granted—for example, that things do not change
weight when they pass from a solid to a liquid state. To put the explanation
into perfect argumentative form, this assumption and many others would
have to be stated explicitly. This is never done in everyday life and is only
rarely done in the most exact sciences.
Is this explanation any good? Explanations are satisfactory if they remove
bewilderment or surprise by telling us how or why something happened in a
way that is relevant to the concerns of a particular context. Our example
does seem to accomplish that much. However, it might seem that even the
best explanations are not very useful because they take so much for granted.
In explaining why the water level remains the same when the ice cube melts,
we cited the law of buoyancy. Now, why should that law be true? What ex-
plains it? To explain the law of buoyancy, we would have to derive it from
other laws that are more general and, perhaps, more intelligible. In fact, this
has been done. Archimedes simultaneously proved and explained the law
8
CHAPTER 1 ■ Uses of Arguments
of buoyancy by deriving it from the laws of the lever. How about the laws of
the lever? Can they be proved and explained by deriving them from still
higher and more comprehensive laws? Perhaps. Yet reasons give out, and
sooner or later explanation (like justification) comes to an end. It is the task
of science and all rational inquiry to move that boundary further and further
back. But even when there is more to explain, that does not show that a par-
tial explanation is totally useless. As we have seen, explanations can be use-
ful even when they are incomplete, and even though they are not used to
justify any disputed claim. Explanation is, thus, a separate use of arguments.
9
Explanat ions
Houses in Indonesia sometimes have their electrical outlets in the middle of
the wall rather than at floor level. Why? A beginning of an explanation is that
flooding is a danger in the Netherlands. Citing this fact does not help much,
however, unless one remembers that Indonesia was formerly a Dutch colony.
We can understand why the Dutch might put their electrical outlets above
floor level in the Netherlands. It is safer in a country where flooding is a dan-
ger. Is flooding, then, a similar danger in Indonesia? Apparently not; so why
did the Dutch continue this practice in Indonesia? The answer is that colonial
settlers tend to preserve their home customs, practices, and styles. The Dutch
continued to build Dutch-style houses with the electrical outlets where (for
them) they are normally placed—that is, in the middle of the wall rather than
at floor level. Restate this explanation in the form of an argument (that is, spec-
ify its premises and conclusion).
Exercise II
Write a brief argument to explain each of the following. Indicate what facts
and what general principles are employed in your explanations. (Do not for-
get those principles that may seem too obvious to mention.)
1. Why a lighter-than-air balloon rises.
2. Why there is an infield fly rule in baseball.
3. Why there is an international date line.
4. Why there are more psychoanalysts in New York City than in any other
city or, for that matter, in most countries in the world.
5. Why average temperatures tend to be higher closer to the equator.
6. Why there are usually more college freshmen who plan to go to medical
school than there are seniors who still plan to go to medical school.
7. Why almost no textbooks are more than eighteen inches high.
8. Why most cars have four tires (instead of more or fewer).
9. Why paintings by Van Gogh cost so much.
10. Why wages go up when unemployment goes down.
Exercise III
COMBINATIONS: AN EXAMPLE
Although justification and explanation are distinct uses of arguments, we
often want to know both what happened and also why it happened. Then we
need to combine justifications and explanations. We can see how this works
by considering a fictional example.
Imagine that Madison was arrested for murdering her husband, Victor.
Now she is on trial, and you are on the jury. Presumably, the police and the
prosecuting attorneys would not have arrested and prosecuted her if they
did not believe that Madison committed the murder, but are their beliefs
justified? Should she be convicted and sent to prison? That’s up to you and
the other jurors to decide.
You do not want to convict her arbitrarily, of course, so you need argu-
ments to justify you in believing that Madison is guilty. The goal of prose-
cuting attorneys is to provide such justification. Their means of reaching this
goal is to present evidence and arguments during the trial. Although their
ultimate conclusion is that you should find Madison guilty of murder, the
prosecutors need to justify lots of little claims along the way.
It might seem too obvious to mention, but the prosecution first needs an
argument to show that the victim died. After all, if nobody died, nobody
was killed. This first argument can be pretty simple: This person was walk-
ing and talking before he was shot in the head; now his heart has stopped
beating for a long time; so he must be dead. There can be complications,
since some gunshot victims can be revived, but let’s assume that an argu-
ment like this justifies the claim that the victim is dead.
We also want to know who the victim was. The body was identified by
several of Victor’s friends, we assume, so all the prosecution needs to argue
is that identifications like this are usually correct, so it was Victor who died.
This second argument also provides a justification, but it differs from the
first argument in several ways. The first argument referred directly to the
facts about Victor that show he died, whereas this second argument does not
say which features of the victim show that it was Victor. Instead, this argu-
ment relies on trusting other people—Victor’s friends—without knowing
what it was about the victim’s face that made them think it was Victor. Such
appeals to authority will be discussed in more detail in Chapters 3 and 15.
The third issue is the cause of death. Here it is common to appeal to a
medical authority. In our case, the coroner or medical examiner makes
observations or runs scientific tests that provide premises for another
10
CHAPTER 1 ■ Uses of Arguments
It is sometimes said that science tells us how things happen but does not tell us
why they happen. In what ways is this contention right, and in what ways is it
wrong?
Discussion Question
argument that is supposed to justify the conclusion that Victor’s death was
caused by a bullet to the head. This argument is also an appeal to an author-
ity, but here the authority is a scientific expert rather than a friend.
Yet another argument, possibly based on firing marks on the bullet, can
then justify you in believing that the bullet came from a certain gun. More
arguments, possibly based on eyewitnesses, then justify the claims that
Madison was the person who fired that gun at Victor. And so on.
All of these arguments depend on background assumptions. When you
see the marks on the bullet that killed Victor line up with the marks on
another bullet that was fired from the alleged murder weapon, you as-
sume that guns leave distinctive marks on bullets and that nobody
switched the bullets. A good prosecutor will provide arguments for these
assumptions, but nobody can prove everything. Arguments always start
from assumptions. This problem will occupy us at several points later, in-
cluding parts of Chapters 3 and 5. The point for now is just that the prose-
cution needs to produce several arguments of various kinds in order to
justify the claim that Madison killed Victor.
It is also crucial that killing violates the law. If not, then Madison should
not be found guilty for killing Victor. So, how can the prosecutor justify the
assumption that such killing is illegal? Prosecutors usually just quote a
statute or cite a common law principle and apply it to the case, but that ar-
gument assumes a lot of background information. In the case of a statute,
there must be a duly elected legislature, it must have jurisdiction over the
place and time where and when the killing occurred, it must follow required
procedures, and the content of the law must be constitutionally permissible.
Given such a context, if the legislature says that a certain kind of killing is il-
legal, then it is illegal. It is fascinating that merely announcing that some-
thing is illegal thereby makes it illegal. We will explore such performatives
and speech acts in Chapter 2. For now we will simply assume that all of
these arguments could be provided if needed.
Even so, Madison might have had some justification for killing Victor,
such as self-defense. This justification for her act can be presented in an ar-
gument basically like this: I have a reason to protect my own life, and I need
to kill Victor first in order to protect my own life, so I have a reason to kill
Victor. This justification differs in several ways from the kind of justification
that we have been discussing so far. For one thing, this argument provides a
reason for a different person—a reason for Madison—whereas the preced-
ing arguments provided a reason for you as a juror. This argument also pro-
vides a reason with a different kind of object, since it justifies an action
(killing Victor) whereas the previous arguments justified a belief (the belief
that Madison did kill Victor). It provides a practical reason instead of an in-
tellectual reason. Despite these differences, however, if her attorneys want
to show that Madison has this new kind of justification, they need to give an
argument to show that she was justified in doing what she did.
Even if Madison had no justification, she still might have had an excuse.
Whereas a justification is supposed to show that the act was the right thing
11
Comb inat ions : An Example
to do, an excuse admits that the act was wrong but tries to show that the
agent was not fully responsible for doing it. Madison might, for example,
argue that she honestly believed that Victor was going to kill her if she did
not kill him first. If she offers this only as an excuse, she can admit that her
belief was mistaken, so she had no justification for killing Victor. Her claim
is, instead, that she was not fully responsible for his death because she was
only trying to defend herself.
Excuses like this are, in effect, explanations. By citing her mistake, Madi-
son explains why she did what she did. If she had killed Victor because she
hated him or because she wanted to take his money, then she would have
no excuse. Her act is less blameworthy, however, if she was mistaken. Of
course, you should be careful before you shoot someone, so Madison could
still be guilty of carelessness or negligence. But that is not as bad as killing
someone out of hatred or for money. Her mistake might even be reasonable.
If Victor was aiming a gun at her, then, even if it turned out not to be loaded,
any rational person in her position might have thought that Victor was on
the attack. Such reasonable mistakes might reduce or even remove responsi-
bility. Thus, by explaining her act as a mistake, Madison puts her act in a
better light than it would appear without that explanation. In general, an
excuse is just an explanation of an act that puts that act in a better light by
reducing the agent’s responsibility.
To offer an excuse, then, Madison’s defense attorneys will need to give
arguments whose purpose is not justification but explanation. This excuse
will then determine what she is guilty of. Whether Madison is guilty of first-
degree murder or some lesser charge, such as second-degree murder or
manslaughter, or even no crime at all, depends on the explanation for her act
of killing Victor.
Several of the earlier arguments also provided explanations. The medical
examiner cited the head wound to explain why Victor stopped breathing.
The victim’s identity explained why his friends said he was Victor. The fact
that the bullet came out of a particular gun explained why it had certain
markings. The legislature’s vote explained why the killing was illegal. And
so on.
In this way, what appears appears at first to be a simple case actually
depends on a complex chain of arguments that mixes justifications with
explanations. All of these justifications and explanations can be understood
by presenting them explicitly in the form of arguments.
One final point is crucial. Suppose that Madison has no justification or
excuse for killing Victor. It is still not enough for the prosecutor to give any
old argument that Madison killed Victor. The prosecution must prove guilt
beyond a reasonable doubt. This burden of proof makes the strength of the
argument crucial. You as a juror should not convict, even if you think Madi-
son is guilty, unless the prosecution’s argument meets this high standard. In
this case, as in many others, it is not enough just to be able to identify the ar-
gument and to understand its purpose. You also need to determine how
strong it is.
12
CHAPTER 1 ■ Uses of Arguments
For such reasons, we all need to understand arguments and to be able to
evaluate them. This need arises not only in law but also in life, such as when
we decide which candidate to vote for, what course to take, whether to be-
lieve that your spouse is cheating on you, and so on. The goal of this book is
to teach the skills needed for understanding and assessing arguments about
important issues like these.
13
Comb inat ions : An Example
The following arguments mix justification with explanation. For each part of
the argument, determine whether it is a justification or an explanation. How
does each sub-argument work? How strong is it? How would you respond if
you disagreed? How would you defend that part against criticisms?
It will, of course, be difficult to answer these questions before studying the
rest of this book. However, it is worthwhile to reflect on how much you al-
ready understand at the start. It is also useful to have some concrete examples
to keep in mind as you study arguments in more depth.
1. Dinosaurs are fascinating, and we cannot help but wonder what killed
these magnificent creatures. The following argument tries to show that
they were killed when a giant meteor struck the earth, but first the authors
need to argue against alternatives to this impact hypothesis.
AN EXTRATERRESTRIAL IMPACT
by Walter Alvarez and Frank Asaro
from Scientific American (October 1990), pp. 78–92
About 65 million years ago something killed half of all the life on the earth.
This sensational crime wiped out the dinosaurs, until then undisputed mas-
ters of the animal kingdom, and left the humble mammals to inherit their
estate. Human beings, descended from those survivors, cannot avoid asking
who or what committed the mass murder and what permitted our distant
ancestors to survive. . . .
Murder suspects typically must have means, motive and opportunity. An
impact [of a giant meteor, probably in the Yucatan peninsula] certainly had the
means to cause the Cretaceous extinction, and the evidence that an impact
occurred at exactly the right time points to opportunity.
The impact hypothesis provides, if not motive, then at least a mechanism
behind the crime. How do other suspects in the killing of dinosaurs fare?
Some have an air-tight alibi: they could not have killed all the different
organisms that died at the KT boundary. The venerable notion that mammals
ate the dinosaurs’ eggs, for example, does not explain the simultaneous extinc-
tion of marine foraminifera and ammonites.
Discussion Questions
(continued)
Stefan Gartner of Texas A&M University once suggested that marine life
was killed by a sudden huge flood of fresh water from the Arctic Ocean, which
apparently was isolated from other oceans during the late Cretaceous and
filled with fresh water. Yet this ingenious mechanism cannot account for the
extinction of the dinosaurs or the loss of many species of land plants.
Other suspects might have had the ability to kill, but they have alibis
based on timing. Some scientific detectives have tried to pin the blame for
mass extinction on changes in climate or sea level, for example. Such changes,
however, take much longer to occur than did the extinction; moreover, they do
not seem to have coincided with the extinction, and they have occurred repeat-
edly throughout the earth’s history without accompanying extinctions.
Others consider volcanism a prime suspect. The strongest evidence impli-
cating volcanoes is the Deccan Traps, an enormous outpouring of basaltic lava
in India that occurred approximately 65 million years ago. Recent paleomag-
netic work by Vincent E. Courtillot and his colleagues in Paris confirms previ-
ous studies. They show that most of the Deccan Traps erupted during a single
period of reversed geomagnetic polarity, with slight overlaps into the preced-
ing and succeeding periods of normal polarity. The Paris team has found that
the interval in question is probably 29R, during which the KT extinction
occurred, although it might be the reversed-polarity interval immediately be-
fore or after 29R as well.
Because the outpouring of the Deccan Traps began in one normal interval
and ended in the next, the eruptions that gave rise to them must have taken
place over at least 0.5 Myr. Most workers interested in mass extinction there-
fore have not considered volcanism a serious suspect in a killing that evidently
took place over 0.001 Myr or less. . . .
Moreover, basaltic spherules in the KT boundary argue against explosive
volcanism in any case; spherules might be generated by quieter forms of vol-
canism, but then they could not be transported worldwide. The apparent global
distribution of the iridium anomaly, shocked quartz and basaltic spherules is
strong evidence exonerating volcanism and pointing to impact. Eruptions take
place at the bottom of the atmosphere; they send material into the high strato-
sphere at best. Spherules and quartz grains, if they came from an eruption,
would quickly be slowed by atmospheric drag and fall to the ground.
Nevertheless, the enormous eruptions that created the Deccan Traps did
occur during a period spanning the KT extinction. Further, they represent the
greatest outpouring of lava on land in the past quarter of a billion years (al-
though greater volumes flow continually out of mid-ocean ridges). No investi-
gator can afford to ignore that kind of coincidence.
It seems possible that impact triggered the Deccan Traps volcanism. A few
minutes after a large body hit the earth the initial crater would be 40 kilome-
ters deep, and the release of pressure might cause the hot rock of the underly-
ing mantle to melt. Authorities on the origin of volcanic provinces, however,
find it very difficult to explain in detail how an impact could trigger large-
scale basaltic volcanism.
In the past few years the debate between supporters of each scenario has
become polarized: impact proponents have tended to ignore the Deccan Traps
14
CHAPTER 1 ■ Uses of Arguments
as irrelevant, while volcano backers have tried to explain away evidence for
impact by suggesting that it is also compatible with volcanism.
Our sense is that the argument is a Hegelian one, with an impact thesis and
a volcanic antithesis in search of a synthesis whose outlines are as yet unclear. . . .
2. In his famous testimony to the United Nations Security Council on Febru-
ary 5, 2003, which was 42 days before U.S. troops entered Iraq, Secretary of
Defense Colin Powell gave several arguments for his main conclusion that
Saddam Hussein was at that time still trying to obtain fissile material for a
nuclear weapons program.
. . . Let me turn now to nuclear weapons. We have no indication that Saddam
Hussein has ever abandoned his nuclear weapons program. On the contrary,
we have more than a decade of proof that he remains determined to acquire
nuclear weapons.
To fully appreciate the challenge that we face today, remember that in 1991
the inspectors searched Iraq’s primary nuclear weapons facilities for the first
time, and they found nothing to conclude that Iraq had a nuclear weapons pro-
gram. But, based on defector information, in May of 1991, Saddam Hussein’s
lie was exposed. In truth, Saddam Hussein had a massive clandestine nuclear
weapons program that covered several different techniques to enrich uranium,
including electromagnetic isotope separation, gas centrifuge and gas diffusion.
We estimate that this illicit program cost the Iraqis several billion dollars.
Nonetheless, Iraq continued to tell the IAEA that it had no nuclear weapons
program. If Saddam had not been stopped, Iraq could have produced a nuclear
bomb by 1993, years earlier than most worst case assessments that had been
made before the war.
In 1995, as a result of another defector, we find out that, after his invasion of
Kuwait, Saddam Hussein had initiated a crash program to build a crude nuclear
weapon, in violation of Iraq’s UN obligations. Saddam Hussein already pos-
sesses two out of the three key components needed to build a nuclear bomb. He
has a cadre of nuclear scientists with the expertise, and he has a bomb design.
Since 1998, his efforts to reconstitute his nuclear program have been fo-
cused on acquiring the third and last component: sufficient fissile material to
produce a nuclear explosion. To make the fissile material, he needs to develop
an ability to enrich uranium. Saddam Hussein is determined to get his hands
on a nuclear bomb.
He is so determined that he has made repeated covert attempts to acquire
high-specification aluminum tubes from 11 different countries, even after in-
spections resumed. These tubes are controlled by the Nuclear Suppliers Group
precisely because they can be used as centrifuges for enriching uranium.
By now, just about everyone has heard of these tubes and we all know
that there are differences of opinion. There is controversy about what these
tubes are for. Most U.S. experts think they are intended to serve as rotors in
centrifuges used to enrich uranium. Other experts, and the Iraqis themselves,
argue that they are really to produce the rocket bodies for a conventional
weapon, a multiple rocket launcher.
15
Comb inat ions : An Example
(continued)
16
CHAPTER 1 ■ Uses of Arguments
Let me tell you what is not controversial about these tubes. First, all the
experts who have analyzed the tubes in our possession agree that they can be
adapted for centrifuge use.
Second, Iraq had no business buying them for any purpose. They are
banned for Iraq.
I am no expert on centrifuge tubes, but this is an old army trooper. I can
tell you a couple things.
First, it strikes me as quite odd that these tubes are manufactured to a tol-
erance that far exceeds U.S. requirements for comparable rockets. Maybe
Iraqis just manufacture their conventional weapons to a higher standard than
we do, but I don’t think so.
Second, we actually have examined tubes from several different batches
that were seized clandestinely before they reached Baghdad. What we notice in
these different batches is a progression to higher and higher levels of specifica-
tion, including in the latest batch an anodized coating on extremely smooth
inner and outer surfaces.
Why would they continue refining the specifications? Why would they go
to all that trouble for something that, if it was a rocket, would soon be blown
into shrapnel when it went off?
The high-tolerance aluminum tubes are only part of the story. We also
have intelligence from multiple sources that Iraq is attempting to acquire mag-
nets and high-speed balancing machines. Both items can be used in a gas cen-
trifuge program to enrich uranium.
In 1999 and 2000, Iraqi officials negotiated with firms in Romania, India,
Russia and Slovenia for the purchase of a magnet production plant. Iraq
wanted the plant to produce magnets weighing 20 to 30 grams. That’s the
same weight as the magnets used in Iraq’s gas centrifuge program before the
Gulf War.
This incident, linked with the tubes, is another indicator of Iraq’s attempt
to reconstitute its nuclear weapons program.
Intercepted communications from mid-2000 through last summer showed
that Iraqi front companies sought to buy machines that can be used to balance
gas centrifuge rotors. One of these companies also had been involved in a
failed effort in 2001 to smuggle aluminum tubes into Iraq.
People will continue to debate this issue, but there is no doubt in my
mind. These illicit procurement efforts show that Saddam Hussein is very
much focused on putting in place the key missing piece from his nuclear
weapons program, the ability to produce fissile material. . . .
The Web of Language
Arguments are made up of language, so we cannot understand arguments without
first understanding language. This chapter will examine some of the basic features
of language, stressing three main ideas. First, language is conventional. Words
acquire meaning within a rich system of linguistic conventions and rules. Second,
the uses of language are diverse. We use language to communicate information, but
we also use it to ask questions, issue orders, write poetry, keep score, formulate
arguments, and perform an almost endless number of other tasks. Third, meaning is
often conveyed indirectly. To understand the significance of many utterances, we
must go beyond what is literally said to examine what is conversationally implied
by saying it.
LANGUAGE AND CONVENTION
The preceding chapter stressed that arguing is a practical activity. More
specifically, it is a linguistic activity. Arguing is one of the many things that
we can do with words. In fact, unlike things that we can accomplish both
with words and without words (like making people happy, angry, and so
forth), arguing is something we can only do with words or other meaningful
symbols. That is why nonhuman animals never give arguments. To under-
stand how arguments work, then, it is crucial to understand how language
works.
Unfortunately, our understanding of human language is far from
complete, and linguistics is a young science in which disagreement exists on
many important issues. Still, certain facts about language are beyond
dispute, and recognizing them will provide a background for understand-
ing how arguments work.
As anyone who has bothered to think about it knows, language is con-
ventional. There is no reason why we, as English speakers, use the word
“dog” to refer to a dog rather than to a cat, a tree, or the number of planets
in our solar system. It seems that any word might have been used to stand
for anything. Beyond this, there seems to be no reason why we put words
together the way we do. In English, we put adjectives before the nouns they
modify. We thus speak of a “green salad.” In French, adjectives usually
2
17
follow the noun, and so, instead of saying “verte salade,” the French say
“salade verte.” The conventions of our own language are so much with us
that it strikes us as odd when we discover that other languages have differ-
ent conventions. A French diplomat once praised his own language because,
as he said, it followed the natural order of thought. This strikes English
speakers as silly, but in seeing why it is silly, we see that the word order in
our own language is conventional as well.
Although it is important to realize that language is conventional, it is also
important not to misunderstand this fact. From the idea that language is
conventional, it is easy to conclude that language is totally arbitrary. If
language is totally arbitrary, then it might seem that it really does not matter
which words we use or how we put them together. It takes only a little
thought to see that this view, however daring it might seem, misrepresents
the role of conventions in language. If we wish to communicate with others,
we must follow the system of conventions that others use. Grapefruits are
more like big lemons than like grapes, so you might want to call them
“mega-lemons.” Still, if you order a glass of mega-lemon juice in a restau-
rant, you will get stares and smirks but no grapefruit juice. The same point
lies behind this famous passage in Through the Looking Glass, by Lewis
Carroll:
“There’s glory for you!”
“I don’t know what you mean by ‘glory’,” Alice said.
Humpty Dumpty smiled contemptuously.
“Of course you don’t—till I tell you. I meant ‘there’s a nice knock-down
argument for you!’”
“But ‘glory’ doesn’t mean ‘a nice knock-down argument’,” Alice objected.
“When I use a word,” Humpty Dumpty said, in a rather scornful tone, “it means
just what I choose it to mean—neither more nor less.”
“The question is,” said Alice, “whether you can make words mean so many dif-
ferent things.”
The point, of course, is that Humpty Dumpty cannot make a word mean
whatever he wants it to mean, and he cannot communicate if he uses words
in his own peculiar way without regard to what those words themselves
mean. Communication can take place only within a shared system of
conventions. Conventions do not destroy meaning by making it arbitrary;
conventions bring meaning into existence.
A misunderstanding of the conventional nature of language can lead to
pointless disputes. Sometimes, in the middle of a discussion, someone will
declare that “the whole thing is just a matter of definition” or “what you say
is true by your definition, false by mine.” There are times when definitions
are important and the truth of what is said turns on them, but usually this is
not the case. Suppose someone has fallen off a cliff and is heading toward
certain death on the rocks below. Of course, it is a matter of convention that
we use the word “death” to describe the result of the sudden, sharp stop
at the end of the fall. We might have used some other word—perhaps
18
CHAPTER 2 ■ The Web of Language
“birth“—instead. But it certainly will not help a person who is falling to his
certain death to shout out, “By ‘birth’ I mean death.” It will not help even if
everyone agrees to use these words in this new way. If we all decided to
adopt this new convention, we would then say, “He is falling from the cliff
to his certain birth” instead of “He is falling from the cliff to his certain
death.” But speaking in this way will not change the facts. It will not save
him from perishing. It will not make those who care for him feel better.
The upshot of this simple example is that the truth of what we say is
rarely just a matter of definition. Whether what we have said is true or not
will depend, for the most part, on how things stand in the world. Abraham
Lincoln, during his days as a trial lawyer, is reported to have cross-examined
a witness like this:
“How many legs does a horse have?”
“Four,” said the witness.
“Now, if we call a tail a leg, how many legs does a horse have?”
“Five,” answered the witness.
“Nope,” said Abe, “calling a tail a leg don’t make it a leg.”
In general, then, though the meaning of what we say is dependent on conven-
tion, the truth of what we say is not.
In the preceding sentence we used the qualifying phrase, “in general.” To
say that a claim holds in general indicates that there may be exceptions. This
qualification is needed because sometimes the truth of what we say is
simply a matter of definition. Take a simple example: The claim that a trian-
gle has three sides is true by definition, because a triangle is defined as “a
closed figure having three sides.” Again, if someone says that sin is wrong,
he or she has said something that is true by definition, for a sin is defined as,
among other things, “something that is wrong.” In unusual cases like these,
things are true merely as a matter of convention. Still, in general, the truth of
what we say is settled not by appealing to definitions but, instead, by
looking at the facts. In this way, language is not arbitrary, even though it is
conventional.
LINGUISTIC ACTS
In the previous section we saw that a language is a system of shared conven-
tions that allows us to communicate with one another. If we examine lan-
guage, we will see that it contains many different kinds of conventions.
These conventions govern what we will call linguistic acts, speech acts, and
conversational acts. We will discuss linguistic acts first.
We have seen that words have meanings conventionally attached to
them. The word “dog” is used conventionally to talk about dogs. Given
what our words mean, it would be incorrect to call dogs “airplanes.” Proper
names are also conventionally assigned, for Harry Jones could have been
19
Lingu ist ic Acts
named Wilbur Jones. Still, given that his name is not Wilbur, it would be
improper to call him Wilbur. Rules like these, which govern meaning and
reference, can be called semantic rules.
Other conventions concern the ways words can be put together to form
sentences. These are often called syntactic or grammatical rules. Using the
three words “John,“ “hit,” and “Harry,” we can formulate sentences with
very different meanings, such as “John hit Harry” and “Harry hit John.” We
recognize that these sentences have different meanings, because we under-
stand the grammar of our language. This grammatical understanding also
allows us to see that the sentence “Hit John Harry” has no determinate
meaning, even though the individual words do. (Notice that “Hit John,
Harry!” does mean something: It is a way of telling Harry to hit John.) Gram-
matical rules are important, for they play a part in giving a meaning to com-
binations of words, such as sentences.
Some of our grammatical rules play only a small role in this important
task of giving meaning to combinations of words. It is bad grammar to say,
“If I was you, I wouldn’t do that,” but it is still clear what information the
person is trying to convey. What might be called stylistic rules of grammar
are of relatively little importance for logic, but grammatical rules that affect
the meaning or content of what is said are essential to logical analysis.
Grammatical rules of this kind can determine whether we have said one
thing rather than another, or perhaps failed to say anything at all and have
merely spoken nonsense.
It is sometimes hard to tell what is nonsense. Consider “The horse raced
past the barn fell.” This sentence usually strikes people as nonsense when
they hear it for the first time. To show them that it actually makes sense, all
we need to do is insert two words: “The horse that was raced past the barn
fell.” Since English allows us to drop “that was,” the original sentence
means the same as the slightly expanded version. Sentences like these are
called “garden path sentences,” because the first few words “lead you down
the garden path” by suggesting that some word plays a grammatical role
that it really does not play. In this example, “The horse raced . . .” suggests
at first that the main verb is “raced.” That makes it hard to see that the main
verb really is “fell.”
Another famous example is “Buffalo buffalo buffalo.” Again, this
seems like nonsense at first, but then someone points out that “buffalo”
can be a verb meaning “to confuse.” The sentence “Buffalo buffalo buf-
falo” then means “North American bison confuse North American bison.”
Indeed, we can even make sense out of “Buffalo buffalo Buffalo buffalo
buffalo buffalo Buffalo buffalo Buffalo buffalo buffalo.” This means
“North American bison from Buffalo, New York, that North American
bison from Buffalo, New York, confuse also confuse North American
bison from Buffalo, New York, that North American bison from Buffalo,
New York, confuse.”
Examples like these show that sentences can have linguistic meaning
when they seem meaningless. To be meaningful, sentences need to follow
20
CHAPTER 2 ■ The Web of Language
both semantic conventions that govern meanings of individual words and
also syntactic or grammatical conventions that lay down rules for combining
words into meaningful wholes. When a sentence satisfies essential semantic
and syntactic conventions, we will say that the person who uttered that
sentence performed a linguistic act: The speaker said something meaningful
in a language.1 The ability to perform linguistic acts shows a command of a
language. What the speaker says may be false, irrelevant, boring, and so on;
but, if in saying it linguistic rules are not seriously violated, then that person
can be credited with performing a linguistic act.
Later, in Chapters 13–14, we will look more closely at semantic and
syntactic conventions, for they are common sources of fallacies and other
confusions. In particular, we shall see how these conventions can generate
fallacies of ambiguity and fallacies of vagueness. Before examining the
defects of our language, however, we should first appreciate that language
is a powerful and subtle tool that allows us to perform a wide variety of jobs
important for living in the world.
21
Lingu ist ic Acts
Read each of the following sentences aloud. Did you perform a linguistic act?
If so, explain what the sentence means and why it might not seem meaningful.
1. The old man the ship.
2. Colorless green ideas sleep furiously.
3. Time flies like an arrow. Fruit flies like bananas.
4. The cotton clothing is made of grows in Mississippi.
5. The square root of pine is tree.
6. The man who whistles tunes pianos.
7. To force heaven, Mars shall have a new angel. (from Monk)
8. “’Twas brillig, and the slithy toves did gyre and gimble in the wabe.”
(from Lewis Carroll)
And now some weird examples from Dan Wegner’s Hidden Brain Damage
Scale. If these make sense to you, it might be a sign of hidden brain damage.
If they don’t make sense, explain why:
9. People tell me one thing one day and out the other.
10. I feel as much like I did yesterday as I do today.
11. My throat is closer than it seems.
12. Likes and dislikes are among my favorites.
13. I’ve lost all sensation in my shirt.
14. There’s only one thing for me.
15. I don’t like any of my loved ones.
Exercise I
SPEECH ACTS
When we are asked about the function of language, it is natural to reply that we
use language to communicate ideas. This is, however, only one of the purposes
for which we use language. Other purposes become obvious as soon as we
look at the ways in which our language actually works. Adding up a column of
figures is a linguistic activity—though it is rarely looked at in this way—but it
does not communicate any ideas to others. When I add the figures, I am not
even communicating anything to myself: I am trying to figure something out.
A look at our everyday conversations produces a host of other examples of
22
CHAPTER 2 ■ The Web of Language
1. When an actor on a stage says lines such as “To be or not to be, that is the
question,” does the actor perform a linguistic act?
2. When someone hums (but does not sing) the “Star-Spangled Banner,” does
she perform a linguistic act? Why or why not?
3. Can a speaker mispronounce a word in a sentence without performing any
linguistic act? Why or why not?
Discussion Questions
Image not available due to copyright restrictions
language being used for different purposes. Grammarians, for example, have
divided sentences into various moods, among which are:
Indicative: Barry Bonds hit a home run.
Imperative: Get in there and hit a home run, Barry!
Interrogative: Did Barry Bonds hit a home run?
Expressive: Hurray for Barry Bonds!
The first sentence states a fact. We can use it to communicate information
about something that Barry Bonds did. If we use it in this way, then what we
say will be either true or false. Notice that none of the other sentences can be
called either true or false even though they are all meaningful.
PERFORMATIVES
The different types of sentences recognized by traditional grammarians
indicate that we use language to do more than convey information. But this
traditional classification of sentences gives only a small idea of the wide
variety of things that we can accomplish using language. Sometimes, for
example, in using language we actually do things in the sense of bringing
something about. In one familiar setting, if one person says, “I do,” and
another person says, “I do,” and finally a third person says, “I now pro-
nounce you husband and wife,” the relationship between the first two peo-
ple changes in a fundamental way: They are thereby married. With luck,
they begin a life of wedded bliss, but they also alter their legal relationship.
For example, they may now file joint income tax returns and may not legally
marry other people without first getting divorced.
In uttering sentences of this kind, the speaker thereby does something more
than merely stating something. The philosopher J. L. Austin labeled such utter-
ances performatives in order to contrast performing an action with simply
describing something.2 For example, if an umpire shouts, “You’re out!” then
the batter is out. The umpire is not merely describing the situation but declaring
the batter out. By way of contrast, if someone in the stands shouts, “He’s out!”
the batter is not thereby out, although the person who shouts this may be en-
couraging the umpire to call the batter out or complaining because he didn’t.
Performatives come in a wide variety of forms. They are often in the first
person (like “I do”), but not always. “You’re all invited to my house after the
game” is in the second person, but uttering it performs the act of inviting. In
some circumstances, one person can speak for another person, a group, or
an institution. At political conventions, heads of delegations say things like
this: “The delegates from Kentucky, the Bluegrass State and the home of the
Kentucky Derby, cast their votes for the next President of the United States,
Joe W. Blodgett.” In saying this, the speaker performs the act of casting
Kentucky’s votes in favor of Blodgett. Even silence can amount to a perfor-
mative act in special situations. When the chairperson of a meeting asks if
there are any objections to a ruling and none is voiced, then the voters,
through their silence, have accepted the ruling.
23
Speech Acts
Because of this diversity of forms that performatives can take, it is not easy
to formulate a definition that covers them all. To avoid this difficulty, we will
not even try to define performatives here. Instead, we will concentrate on one
particularly clear subclass of performatives, what J. L. Austin called explicit
performatives. All explicit performatives are utterances in the first-person
singular indicative noncontinuous3 present. But not all utterances of that
form are explicit performatives. There is one more requirement:
An utterance of that form is an explicit performative if and only if it yields a true
statement when plugged into the following pattern:
In saying “I _____” in appropriate circumstances, I thereby _____.
For example, “I congratulate you” expresses an explicit performative, be-
cause, in saying “I congratulate you,” I thereby congratulate you. Here a
quoted expression occurs on the left side of the word “thereby” but not on
the right side. This reflects the fact that the formula takes us from the words
(which are quoted) to the world (the actual act that is performed). The say-
ing, which is referred to on the left side of the pattern, amounts to the doing
referred to on the right side of the word “thereby.” We will call this the
thereby test for explicit performatives. It provides a convenient way of identi-
fying explicit performatives.
The thereby test includes an important qualification: The context of the ut-
terance must be appropriate. You have not congratulated anyone if you say, “I
congratulate you,” when no one is around, unless you are congratulating
yourself. Congratulations said by an actor in a play are not real congratula-
tions, and so on. Later in this chapter, we will try to clarify what makes a
context appropriate.
Assuming an appropriate context, all of the following sentences meet the
thereby test:
I promise to meet you tomorrow.
I bid sixty-six dollars. (said at an auction)
I bid one club. (said in a bridge game)
I resign from this club.
I apologize for being late.
Notice that it doesn’t make sense to deny any of these performatives. If someone
says, “I bid sixty-six dollars,” it is not appropriate for someone to reply “No,
you don’t” or “That’s false.” It could, however, be appropriate for someone to
reply, “You can’t bid sixty-six dollars, because the bidding is already up to sev-
enty dollars.” In this case, the person tried to make a bid, but failed to do so.
Several explicit performatives play important roles in constructing argu-
ments. These include sentences of the following kind:
I conclude that this bill should be voted down.
I base my conclusion on the assumption that we do not want to hurt
the poor.
24
CHAPTER 2 ■ The Web of Language
I stipulate that anyone who earns less than $10,000 is poor.
I assure you that this bill will hurt the poor.
I concede that I am not absolutely certain.
I admit that there is much to be said on both sides of this issue.
I give my support to the alternative measure.
I deny that this alternative will hurt the economy.
I grant for the sake of argument that some poor people are lazy.
I reply that most poor people contribute to the economy.
I reserve comment on other issues raised by this bill.
We will call this kind of performative an argumentative performative.
Studying such argumentative performatives can help us to understand what
is going on in arguments (which is one main reason why we are studying
performatives here).
In contrast to the above utterances, which pass the thereby test, none of
the following utterances does:
I agree with you. (This describes one’s thoughts or beliefs, so, unlike a
performative, it can be false.)
I am sorry for being late. (This describes one’s feelings and could be false.)
Yesterday I bid sixty dollars. (This is a statement about a past act and
might be false.)
I’ll meet you tomorrow. (This utterance may only be a prediction that can
turn out to be false.)
Questions, imperatives, and exclamations are not explicit performatives,
because they cannot sensibly be plugged into the thereby test at all. They do
not have the right form, since they are not in the first-person singular indica-
tive noncontinuous present.
25
Speech Acts
Using the thereby test as described above, indicate which of the following sen-
tences express explicit performatives (EP) and which do not express explicit
performatives (N) in appropriate circumstances:
1. I pledge allegiance to the flag.
2. We pledge allegiance to the flag.
3. I pledged allegiance to the flag.
4. I always pledge allegiance at the start of a game.
5. You pledge allegiance to the flag.
Exercise II
(continued)
KINDS OF SPEECH ACTS
Recognizing explicit performatives helps break the spell of the idea that lan-
guage functions only to transmit information. It also introduces us to a kind
of act distinct from linguistic acts. We will call them speech acts.4 They in-
clude such acts as stating, promising, swearing, and refusing. A speech act is
the conventional move that a remark makes in a language exchange. It is
what is done in saying something.
It is difficult to give a precise definition of a speech act, but we can begin
by contrasting speech acts with linguistic acts. A linguistic act, we said, is the
act of saying something meaningful in a language. It is important to see that
the same linguistic act can play different roles as it occurs in different con-
texts. This is shown by the following brief conversations.
A: Is there any pizza left?
B: Yes.
A: Do you promise to pay me back by Friday?
B: Yes.
A: Do you swear to tell the truth?
B: Yes.
A: Do you refuse to leave?
B: Yes.
26
CHAPTER 2 ■ The Web of Language
6. He pledges allegiance to the flag.
7. He doesn’t pledge allegiance to the flag.
8. Pledge allegiance to the flag!
9. Why don’t you pledge allegiance to the flag?
10. Pierre is the capital of South Dakota.
11. I state that Pierre is the capital of South Dakota.
12. I order you to leave.
13. Get out of here!
14. I didn’t take it.
15. I swear that I didn’t take it.
16. I won’t talk to you.
17. I refuse to talk to you.
18. I’m out of gas.
19. I feel devastated.
20. Bummer!
21. I claim this land for England.
22. I bring you greetings from home.
Here the same linguistic act, uttering the word “yes,” is used to do four dif-
ferent things: to state something, to make a promise, to take an oath, and to
refuse to do something.
We can make this idea of a speech act clearer by using the notion of an ex-
plicit performative. Explicit performatives provide a systematic way of identify-
ing different kinds of speech acts. The basic idea is that different speech acts are
named by the different verbs that occur in explicit performatives. We can thus
use the thereby test to search for different kinds of speech acts. For example:
If I say, “I promise,” I thereby promise. So “I promise” is a performative,
and promising is a kind of speech act.
If I say, “I resign,” I thereby resign. So “I resign” is a performative, and
resigning is a kind of speech act.
If I say, “I apologize,” I thereby apologize. So “I apologize” is a performative,
and apologizing is a kind of speech act.
If I say, “I question his honesty,” I thereby question his honesty. So “I
question his honesty” is a performative, and questioning is a kind of
speech act.
If I say, “I conclude that she is guilty,” I thereby conclude that she is
guilty. So “I conclude that she is guilty” is a performative, and conclud-
ing is a kind of speech act.
The main verbs that appear in such explicit performatives can be called
performative verbs. Performative verbs name kinds of speech acts.5
Still, the same speech act can also be performed without any performa-
tive verb. I can deny my opponent’s claim by saying either “I deny that” or
simply “No way!” Both utterances perform the speech act of denying, even
though only the former is a performative. The latter is not a performative
and does not contain any performative verb, but it still performs a speech
act. Similarly, I can assure you by saying either “I assure you that I am right”
or “There’s no doubt about it.” Both utterances perform the speech act of as-
suring, even though only the former is a performative.
Thus far, we have emphasized that we do a great deal more with lan-
guage than make statements, assert facts, and describe things—that is, we
do more with language than put forward claims that are either true or false.
But we also use language to do these things, so stating, asserting, and
describing are themselves kinds of speech acts. This can be shown by using
the thereby test:
If I say, “I state that I am a U.S. citizen,” I thereby state that I am a U.S.
citizen.
If I say, “I assert that the defendant was in Detroit at the time of the
crime,” I thereby assert that the defendant was in Detroit at that time.
If I say, “I describe him as being dark haired and just over six feet tall,” I
thereby describe him as being dark haired and just over six feet tall.
27
Speech Acts
We now have a more accurate conception of the way in which language
functions than the common conception that the function of language is to
convey ideas. Making claims that are either true or false is one important
kind of speech act, but we perform a great many other kinds of speech acts
that are also important.
28
CHAPTER 2 ■ The Web of Language
Which of the following verbs names a speech act?
1. capture the suspect 6. punish the defendant
2. assert that the suspect is guilty 7. revoke the defendant’s driver’s license
3. stare accusingly at the suspect 8. welcome the prisoner to prison
4. find the defendant guilty 9. order the prisoner to be silent
5. take the defendant away 10. lock the cell door
Exercise III
Using a dictionary, find ten verbs that can be used to construct explicit perfor-
matives that have not yet been mentioned in this chapter.
Exercise IV
SPEECH ACT RULES
The distinctive feature of a performative utterance is that, in a sense we have
tried to make clear, the saying constitutes a doing of something. In saying,
“I pronounce you husband and wife,” a minister is not simply describing a
marriage ceremony, she is performing it. Here, however, an objection might
arise. Suppose someone who is a supporter of family values goes about the
streets pronouncing random couples husband and wife. Unless this person
is a member of the clergy, a justice of the peace, a ship’s captain, or the like,
that person will have no right to make such pronouncements. Furthermore,
even if this person is, say, a crazed member of the clergy, the pronouncement
will still not come off—that is, the utterance will not succeed in making any-
one husband and wife. The parties addressed have to say, “I do,” they must
have a proper license, and so on. This example shows that a speech act will
fail to come off or will be void unless certain rules or conventions are satisfied.
These rules or conventions that must be satisfied for a speech act to come off
and not be void will be called speech act rules.
The main types of speech act rules can be discovered by considering the
following questions:
1. Must the speaker use any special words or formulas to perform the speech act?
Sometimes a speech act will come off only if certain words or formulas are
used. In baseball the umpire must say, “Strike two,” or something very close
to this, in order to call a second strike. In a pickup game it might be all right to
say instead, “Hey, that’s two bad ones on you, baby!” but that way of calling
strikes is not permitted in serious play. Similarly, certain legal documents are
not valid if they are not properly signed, endorsed, notarized, and so forth.
2. Is any response or uptake by the audience needed in order to complete the
speech act?
Sometimes a speech act will come off only if there is an uptake by another
person. A person can offer a bet by saying, “I bet you ten dollars that the An-
gels will win today,” but this person will have made a bet only if the other
person says, “Done” or “You’re on,” shakes hands, or in some other way ac-
cepts the bet. A marriage ceremony is completely void if one of the parties
does not say, “I do,” but instead says, “Well, maybe I should think about this
for a while.”
3. Must the (a) speaker or (b) audience hold any special position or role in order
for the speaker to perform the speech act?
Sometimes a speech act will come off only if it is performed by someone with
an official position. We have already seen that, for someone to make two peo-
ple husband and wife by pronouncing them husband and wife, that person
must hold a certain official position. Similarly, even if a body is plainly dead
when it arrives at the hospital, a janitor cannot pronounce it dead on arrival.
That is the job of a doctor or a coroner. In the same way, although a shortstop
can perform the linguistic act of shouting, “You’re out,” a shortstop cannot
perform the speech act of calling someone out. Only an umpire can do that.
Moreover, even an umpire cannot call out the catcher or a spectator, so some-
times the audience of the speech also needs to have some special position.
4. Are any other special circumstances required for the speech act?
Most speech acts also involve assumptions or presuppositions that certain
facts obtain. A father cannot bequeath an antique car to his son if he does not
own such a car. You cannot resign from the American Civil Liberties Union
or the Veterans of Foreign Wars if you are not a member. These special cir-
cumstances might sometimes include the audience’s desires. In promising
someone to do something, for example, we usually do so in the belief that the
person wants us to do it. For example, I will promise to drive someone to the
airport only if I believe that person wants to go the airport and would like me
to drive her there. Sometimes, however, we do promise to do things a person
does not want done. I can promise to throw someone out if he doesn’t behave
himself. Here, however, I am making a threat, not a promise. Different an-
swers to this question, thus, reveal differences among speech acts.
5. What feelings, desires, or beliefs is the speaker expected to have?
If we apologize for something, we are expected to feel sorry for what we
have done. If we congratulate someone, we are usually supposed to be
pleased with that person’s success. If we state something, we are expected
29
Speech Acts
to believe what we say. In all these cases—in apologizing, congratulating,
and stating—if the speaker lacks the expected feelings, desires, or beliefs, the
speaker still does succeed in performing the speech act, but that speaker and
speech act are subject to criticism. In this respect, this rule differs from the
preceding rules. Those preceding rules reflected conventions that must be
satisfied for the speech act to come off (for it not to be void). In contrast, the
person who says, “I apologize,” has apologized even if he or she does not
feel sorry. The speech act does come off and is not void, even though the
apology can be criticized as insincere.
6. What general purpose or purposes are served by this kind of speech act?
This final question asks why a certain kind of speech act exists at all. Why,
for example, is there the speech act of promising? That is a rather compli-
cated question, but the primary reason for the institution of promising is
that it helps people coordinate their activities. People who make promises
place themselves under an obligation to do something. When promises are
contractual, this obligation is a legal obligation. Promise making, then, in-
creases the confidence we can have that someone will do what they said
they will do, and, for legal promises at least, provides remedies when they
do not. To cite another example of the purpose of a speech act, apologizing
expresses regret for harming or insulting someone. One of its purposes is to
normalize relations between the speaker and the person harmed or insulted.
Answering the six questions listed above for a particular kind of speech
act is called giving a speech act analysis. For example, here is a brief speech act
analysis of “to appoint,” as in, “I appoint you to the judiciary committee“:
1. Appointments are usually made by using the word “appoint,” but
other words can be used as well; for example, “name” and “designate”
can also be used to do this job. You cannot, however, say, “I wish you
were on the judiciary committee.”
2. Sometimes further actions by others are necessary for an appointment
to come off. Perhaps ratification is needed. Before ratification, the
word “nominate” is often used. In such cases, only after the nomination
is ratified has the appointment been made. Usually the appointment
does not come off if the person declines the appointment.
3. Normally, someone who appoints a person to something must have
the power to make such appointments. For example, Queen Elizabeth
II does not have the power to appoint the commissioner of baseball.
4. This speech act presupposes a wide variety of facts, for example, that
a position exists, that the person appointed to it is eligible for this
appointment, and so on.
5. Appointments are often made with the belief that the person appointed
will do a good job. This is not always the case, however, as appointments
are made for all sorts of different reasons—rewarding an important
supporter, for example.
30
CHAPTER 2 ■ The Web of Language
6. An important purpose of an appointment is to explicitly designate
someone to play a particular role. For example, it is often important to
know who is in charge. It can also be important that the person who
gains this role does so through regular, authorized procedures.
31
Speech Acts
1. Imagine that an actor on stage during a modern play screams, “Fire! No, I
really mean it. ” The audience realizes that he is just acting, so they laugh.
Then the actor sees a real fire behind the stage out of view of the audience.
The actor again screams, “Fire! No, I really mean it. Fire!” Which speech
act, if any, does the actor perform in uttering these words the second
time? Why? What does this show about speech acts?
2. Do the speech acts in which people get married presuppose that the peo-
ple who are getting married are of different sexes? Should these speech
acts presuppose this fact? Why or why not?
3. The importance of deciding what kind of speech act has been performed
is illustrated by a classic case from the law of contracts, Hawkins v. McGee.6
McGee performed an operation on Hawkins that proved unsuccessful,
and Hawkins sued for damages. He did not sue on the basis of malprac-
tice, however, but on the basis of breach of contract. His attorney argued
that the doctor initiated a contractual relationship in that he tried to per-
suade Hawkins to have the operation by saying things such as “I will
guarantee to make the hand a hundred percent perfect hand.” He made
statements of this kind a number of times, and Hawkins finally agreed to
undergo the operation on the basis of these remarks. Hawkins’s attorney
maintained that these exchanges, which took place in the doctor’s office
on a number of occasions, constituted an offer of a contract that Hawkins
explicitly accepted. The attorney for the surgeon replied that these words,
even if uttered, would not constitute an offer of a contract, but merely
Discussion Questions
Give a speech act analysis of the ten verbs below by writing two or three sen-
tences in response to each of the six questions above. Speech act analyses can
go on much longer, but your goal here is just to bring out the most interesting
features of the speech act named by each verb.
1. to bet 6. to deny
2. to promise 7. to vote
3. to congratulate 8. to give up (in a fight)
4. to state 9. to thank
5. to apologize 10. to invite
Exercise V
(continued)
CONVERSATIONAL ACTS7
In examining linguistic acts (saying something meaningful in a language)
and then speech acts (doing something in using words), we have largely ig-
nored some central features of language: It is usually—though not always—
a social activity that takes place among people. It is also normally a practical
activity with certain goals. We use language in order to inform people of
things, get them to do things, amuse them, calm them down, and so on. We
can capture these social and practical aspects of language by introducing the
notion of a conversational exchange, that is, a situation where various speak-
ers use speech acts in order to bring about some effects in each other. We
will call this act of using a speech act to cause a standard effect in another a
conversational act.
Suppose, for example, Amy says to Bobbi, “Someone is following us.” In
this case, Amy has performed a linguistic act; that is, she has uttered a mean-
ingful sentence in the English language. Amy has also performed a speech
act—specifically, she has stated that they are being followed. The point of
performing this speech act is to produce in Bobbi a particular belief—
namely, that they are being followed. (Amy’s utterance might also have
other purposes, such as to alert Bobbi to some danger, but it accomplishes
those other purposes by means of getting Bobbi to believe they are being fol-
lowed.) If Amy is successful in this, then Amy has successfully performed
the conversational act of producing this belief in Bobbi. Amy, of course,
might fail in her attempt to do this. Amy’s linguistic act could be successful
and her speech act successful as well, yet, for whatever reason, Bobbi might
not accept as true what Amy is telling her. Perhaps Bobbi thinks that Amy is
32
CHAPTER 2 ■ The Web of Language
expressed a strong belief, and that reasonable people should know that
doctors cannot guarantee results.
It is important to remember that contracts do not have to be written
and signed to be binding. A proper verbal offer and acceptance are usually
sufficient to constitute a contract. The case, then, turned on two questions:
(1) Did McGee utter the words attributed to him? In other words, did
McGee perform the linguistic act attributed to him? The jury decided
that he did. (2) The second, more interesting question was whether these
words, when uttered in this particular context, amounted to an offer of a
contract, as Hawkins’s attorney maintained, or merely were an expression
of strong belief, as McGee’s attorney held. In other words, the fundamen-
tal question in this case was what kind of speech act McGee performed
when trying to convince Hawkins to have the operation.
Explain how you would settle this case. (The court actually ruled in
favor of Hawkins, but you are free to disagree.)
paranoid or just trying to frighten her as some kind of joke. In that case,
Amy failed to perform her intended conversational act, even though she did
perform her intended linguistic and speech acts.
Here are some other examples of the difference between performing a
speech act and performing a conversational act:
We can warn people about something in order to put them on guard
concerning it.
Here warning is the speech act; putting them on guard is the intended
conversational act.
We can urge people to do things in order to persuade them to do these
things.
Here urging is the speech act; persuading is the intended conversational act.
We can assure people concerning something in order to instill confidence
in them.
Here assuring is the speech act; instilling confidence is the intended
conversational act.
We can apologize to people in order to make them feel better about us.
Here apologizing is the speech act; making them feel better about us is
the intended conversational act.
In each of these cases, our speech act may not succeed in having its intended
conversational effect. Our urging, warning, and assuring may, respectively,
fail to persuade, put on guard, or instill confidence. Indeed, speech acts may
bring about the opposite of what was intended. People who brag (a speech
act) in order to impress others (the intended conversational act) often actu-
ally make others think less of them (the actual effect). In many ways like
these, we can perform a speech act without performing the intended conver-
sational act.
The relationship between conversational acts and speech acts is confus-
ing, because both of them can be performed at once by the same utterance.
Suppose Carl says, “You are invited to my party.” By means of this single ut-
terance, he performs a linguistic act of uttering this meaningful sentence, a
speech act of inviting you, and perhaps also a conversational act of getting
you to come to his party. Indeed, he would not be able to perform this con-
versational act without also performing such a speech act, assuming that
you would not come to his party if you were not invited. He would also not
be able to perform this speech act without performing this linguistic act or
something like it, since he cannot invite you by means of an inarticulate
grunt or by asking, “Are you invited to my party?”
As a result, we cannot sensibly ask whether Carl’s utterance of “You are
invited to my party” is a linguistic act, a speech act, or a conversational act.
That single utterance performs all three acts at once. Nonetheless, we can
distinguish those kinds of acts that Carl performs in terms of the verbs that
describe the acts. Some verbs describe speech acts; other verbs describe
33
Conversat ional Acts
conversational acts. We can tell which verbs describe which kinds of acts by
asking whether the verb passes the thereby test (in which case the verb de-
scribes a speech act) or whether, instead, it describes a standard effect of the
utterance (in which case the verb describes a conversational act).
34
CHAPTER 2 ■ The Web of Language
Indicate whether the verbs in the following sentences name a speech act, a con-
versational act, or neither. Assume a standard context. Explain your answers.
1. She thought that he did it. 11. He praised her lavishly.
2. She asserted that he did it. 12. His praise made her happy.
3. She convinced them that he did it. 13. He threatened to reveal her secret.
4. She condemned him in front of 14. He submitted his resignation.
everyone. 15. Her news frightened him half to
5. She challenged his integrity. death.
6. She embarrassed him in front 16. He advised her to go into another
of them. line of work.
7. He denied doing it. 17. She blamed him for her troubles.
8. They believed her. 18. His lecture enlightened her.
9. They encouraged him to admit it. 19. His jokes amused her.
10. She told him to get lost. 20. His book confused her.
Exercise VI
CONVERSATIONAL RULES
Just as there are rules that govern linguistic acts and other rules that govern
speech acts, so too there are rules that govern conversational acts. This
should not be surprising, because conversations can be complicated inter-
personal activities in need of rules to make them effective in attaining their
goals. These underlying rules are implicitly understood by users of the lan-
guage, but the philosopher Paul Grice was the first person to examine them
in careful detail.
We can start by examining standard or normal conversational exchanges
where conversation is a cooperative venture—that is, where the people in-
volved in the conversation have some common goal they are trying to
achieve in talking with one another. (A prisoner being interrogated and a
shop owner being robbed are not in such cooperative situations.) According
to Grice, such exchanges are governed by what he calls the Cooperative Prin-
ciple. This principle states that the parties involved should use language in a
way that contributes toward achieving their common goal. It tells them to
cooperate.8
This general principle gains more content when we consider other forms
of cooperation. Carpenters who want to build a house need enough nails
and wood, but not too much. They need the right kinds of nails and wood.
They also need to put the nails and wood together in the relevant way—that
is, according to their plans. And, of course, they also want to perform their
tasks quickly and in the right order. Rational people who want to achieve
common goals must follow similar general restrictions in other practical ac-
tivities. Because cooperative conversations are one such practical activity,
speakers who want to cooperate with one another must follow rules analo-
gous to those for carpenters.
Grice spells out four such rules. The first he calls the rule of Quantity. It
tells us to give the right amount of information. More specifically:
1. Make your contribution as informative as is required (for the current
purposes of the exchange).
and possibly:
2. Do not make your contribution more informative than is required.
Here is an application of this rule: A person rushes up to you and asks,
“Where is a fire extinguisher?” You know that there is a fire extinguisher five
floors away in the basement, and you also know that there is a fire extin-
guisher just down the hall. Suppose you say that there is a fire extinguisher
in the basement. Here you have said something true, but you have violated
the first part of the rule of Quantity. You have failed to reveal an important
piece of information that, under the rule of Quantity, you should have pro-
duced. A violation of the second version of the rule would look like this: As
smoke billows down the hall, you say where a fire extinguisher is located on
each floor, starting with the basement. Eventually you will get around to
saying that there is a fire extinguisher just down the hall, but you bury the
point in a mass of unnecessary information.
Grice’s second rule is called the rule of Quality. In general: Try to make
your contribution one that is true. More specifically:
1. Do not say what you believe to be false.
2. Do not say that for which you lack adequate evidence.
In a cooperative activity, you are not supposed to tell lies. Beyond this, you
are expected not to talk off the top of your head either. When we make a
statement, we can be challenged by someone asking, “Do you really believe
that?” or “Why do you believe that?” That a person has the right to ask such
questions shows that statement making is governed by the rule of Quality.
In a court of law, witnesses promise to tell the whole truth and nothing
but the truth. The demand for nothing but the truth reflects the rule of Qual-
ity. The demand for the whole truth roughly reflects the rule of Quantity. Ob-
viously, nobody really tells every truth he or she knows. Here the whole truth
concerns all the known truths that are relevant in the context.
This brings us to our next rule, the rule of Relevance. Simply stated, the
rule of Relevance says:
Be relevant!
35
Conversat ional Acts
Though easy to state, the rule is not easy to explain, because relevance
itself is a difficult notion. It is, however, easy to illustrate. If someone asks
me where he can find a doctor, I might reply that there is a hospital on the
next block. Though not a direct answer to the question, it does not violate
the rule of Relevance because it provides a piece of useful information.
If, however, in response I tell the person that I like his haircut, then I have
violated the rule of Relevance. Clear-cut violations of this principle often
involve changing the subject.
Another rule concerns the manner of our conversation. We are expected
to be clear in what we say. Under the general rule of Manner come various
special rules:
1. Avoid obscurity of expression.
2. Avoid ambiguity.
3. Be brief.
4. Be orderly.
As an example of the fourth part of this rule, when describing a series of
events, it is usually important to state them in the order in which they oc-
curred. It would certainly be misleading to say that two people had a child
and got married when, in fact, they had a child after they were married.
Many other rules govern our conversations. “Be polite!” is one of them.
“Be charitable!” is another. That is, we should put the best interpretation on
what others say, and our replies should reflect this. We should avoid quib-
bling and being picky. For the most part, however, we will not worry about
these other rules.
36
CHAPTER 2 ■ The Web of Language
Indicate which, if any, of Grice’s conversational rules are violated by the itali-
cized sentence of each of the following conversations. Assume a standard con-
text. More than one rule might be violated.
1. “What did you get on the last test?” “A grade.”
2. “Did you like her singing?” “Her costume was beautiful.”
3. “The governor has the brains of a three-year-old.”
4. “The Lone Ranger rode into the sunset and jumped on his horse.”
5. “Without her help, we’d be up a creek without a paddle.”
6. “Where is Palo Alto?” “On the surface of the Earth.”
7. “It will rain tomorrow.” “How do you know?” “I just guessed.”
8. “Does the dog need to go out for a W-A-L-K [spelled out]?”
9. “Why did the chicken cross the road?” “To get to the other side.”
10. Psychiatrist: “You’re crazy.” Patient: “I want a second opinion.”
Psychiatrist: “Okay. You’re ugly, too.”
Exercise VII
CONVERSATIONAL IMPLICATION
In a normal setting where people are cooperating toward reaching a shared
goal, they often conform quite closely to Grice’s conversational rules. If, on
the whole, people did not do this, we could not have the linguistic practices
we do. If we thought, for example, that people very often lied (even about
the most trivial matters), the business of exchanging information would be
badly damaged.
Still, people do not always follow these conversational rules. They withhold
information, they elaborate needlessly, they assert what they know to be false,
they say the first thing that pops into their heads, they wander off the subject,
and they talk vaguely and obscurely. When we observe actual conversations, it
is sometimes hard to tell how any information gets communicated at all.
The explanation lies in the same conversational rules. Not only do we
usually follow these conventions, we also (1) implicitly realize that we are
following them, and (2) expect others to assume that we are following them.
This mutual understanding of the commitments involved in a conver-
sational act has the following important consequence: People are able to
convey a great deal of information without actually saying it.
A simple example will illustrate this point. Again suppose that a person,
with smoke billowing behind him, comes running up to you and asks,
37
Conversat ional Acts
Image not available due to copyright restrictions
“Where’s a fire extinguisher?” You reply, “There’s one in the lobby.” Through
a combination of conversational rules, notably relevance, quantity, and man-
ner, this commits you to the claim that this is the closest, or at least the most
accessible, fire extinguisher. Furthermore, the person you are speaking to
assumes that you are committed to this. Of course, you have not actually
said that it is the closest fire extinguisher; but you have, we might say, implied
this. When we do not actually say something but imply it by virtue of a mutu-
ally understood conversational rule, the implication is called a conversational
implication.
It is important to realize that conversational implication is a pervasive
feature of human communication. It is not something we employ only occa-
sionally for special effect. In fact, virtually every conversation relies on these
implications, and most conversations would fall apart if people refused to
go beyond literal meanings to take into account the implications of saying
things. In the following conversation, B is literal-minded in just this way:
A: Do you know what time it is?
B: Not without looking at my watch.
B has answered A’s question, but it is hard to imagine that A has received
the information she was looking for. Presumably, she wanted to know what
time it was, not merely whether B, at that very moment, knew the time.
Finding B rather obtuse, A tries again:
A: Can you tell me what time it is?
B: Oh, yes, all I have to do is look at my watch.
Undaunted, A gives it another try:
A: Will you tell me what time it is?
B: I suppose I will as soon as you ask me.
Finally:
A: What time is it?
B: Two o’clock. Why didn’t you ask me that in the first place?
Notice that in each of these exchanges B gives a direct and accurate answer
to A’s question; yet, in all but the last answer, B does not provide A with
what A wants. Like a computer in a science-fiction movie, B is taking A’s
questions too literally. More precisely, B does nothing more than take A’s
remarks literally. In a conversational exchange, we expect others to take our
remarks in the light of the obvious purpose we have in making them. We
expect them to share our commonsense understanding of why people ask
questions. At the very least, we expect people to respond to us in ways that
are relevant to our purposes. Except at the end, B seems totally oblivious to
the point of A’s questions. That is what makes B unhelpful and annoying.
Though all the conversational rules we have examined can be the basis of
conversational implication, the rule of Relevance is particularly powerful in
this respect. Normal conversations are dense with conversational implications
38
CHAPTER 2 ■ The Web of Language
that depend on the rule of Relevance. Someone says, “Dinner’s ready,” and
that is immediately taken to be a way of asking people to come to the table.
Why? Because dinner’s being ready is a transparent reason to come to the
table to eat. This is an ordinary context that most people are familiar with.
Change the context, however, and the conversational implications can be en-
tirely different. Suppose the same words, “Dinner’s ready,” are uttered
when guests have failed to arrive on time. In this context, the conversational
implication, which will probably be reflected in an annoyed tone of voice,
will be quite different.
To cite another example of context dependence, if someone says, “I broke
a finger,” people will naturally assume that it is the speaker’s own finger
that was broken. Why? Because when people break fingers, it is almost al-
ways their own fingers that they break. That is the standard context in which
this remark is made. If, however, we shift the context, that conversational
implication can be lost and another can take its place. Suppose the speaker
is a mobster in an extortion racket, that is, someone who physically harms
people who do not pay protection money. Among his fellow extortionists,
the conversational implication of “I broke a finger” is likely to be that it was
someone who refused to pay up who had his finger broken. (We can imag-
ine the extortionist canceling this implication by saying, “No, no, it was my
finger that got broken when I slugged the guy.”)
39
Conversat ional Acts
Assuming a natural conversational setting, what might a person intend to con-
versationally imply by making the following remarks? Briefly explain why
each of these conversational implications holds; that is, explain the relation-
ship between what the speaker literally says and what the speaker intends to
convey through conversational implication. Finally, for each example, find a
context where the standard conversational implication would fail and another
arise in its place.
1. It’s getting a little chilly in here. (Said by a visitor in your home)
2. Do you mind if I borrow your pen? (Said to a friend while studying)
3. We are out of soda. (Said by a child to her parents)
4. I got here before he did. (Said in a ticket line)
5. Don’t blame me if you get in trouble. (Said by someone who advised you
not to do it)
6. Has this seat been taken? (Said in a theater before a show)
7. These sweet potatoes are very filling. (Said when the cook asks if you
want more)
8. Don’t ask me. (Said in response to a question)
9. Does your dog bite? (Said to a man standing next to a dog)
10. I will be out of town that day. (Said in response to a party invitation)
Exercise VIII
VIOLATING CONVERSATIONAL RULES
If we look at basic conversational rules, we notice that these rules sometimes
clash, or at least push us in different directions. The rule of Quantity encour-
ages us to give as much information as possible, but this is constrained by
the rule of Quality, which restricts our claims to things we believe to be true
and can back up with good reasons. The demands of the rule of Quantity
can also conflict with the demand for brevity. In order to be brief, we must
sometimes simplify and even falsify, and this can come into conflict with the
rule of Quality, which demands that we say only what we believe to be true.
Sometimes it is not important to get things exactly right; sometimes it is. An
ongoing conversation can be a constant series of adjustments to this back-
ground system of rules.
Because conversational rules can come into conflict with one another,
speakers can sometimes seem to be violating the Cooperative Principle by
violating one of its maxims. This can happen when one conversational
rule is overridden by another. Grasping the resolution of such a conflict
can generate interesting conversational implications. This may sound
complicated, but an example from Grice should make it clear. Suppose A
tells B, “I’m planning to visit C; where does he live?” B replies, “Some-
where in the south of France.” If A is interested in visiting C, then B’s reply
really does not give her the information she needs and thus seems to vio-
late the first part of the rule of Quantity. We can explain this departure on
the assumption that B does not know exactly where C lives and would
thus violate the rule of Quality if he said anything more specific. In this
case, B’s reply conversationally implies that he does not know exactly
where C lives.
In a more extreme case, a person may even flout one of these conventions,
that is, may openly violate a conversational rule without, as in the previous
example, there being any other conversational rule that overrides it. Here is
an adaptation of one of Grice’s examples and his explanation of it:
A is writing a letter of recommendation about one of his students who is
applying to law school, and the letter reads as follows: “Dear Sir: Mr. X’s
command of English is excellent, and his attendance in class has been
regular. Yours, etc.” (Gloss: A cannot be opting out, since if he wished to be
uncooperative, why write at all? He cannot be unable, through ignorance, to
say more, since the person is his student; moreover, he knows that more
information is wanted. He must, therefore, be wishing to impart information
he is reluctant to write down. This supposition is only tenable on the assump-
tion that he thinks that Mr. X is not a good student. This, then, is what he is
implicating.)
This is a case of damning with faint praise. Faint praise can be damning because,
under the first part of the rule of Quantity, it conversationally implies that no
stronger praise is warranted.
40
CHAPTER 2 ■ The Web of Language
We can intentionally violate the rule of Relevance by pointedly changing
the subject. Here is variation on another one of Grice’s examples:
Standing outside a classroom, A says, “Professor X is a moron.” There is a
moment of shocked silence; then B says, “Nice day, isn’t it?”
A would have to be fairly dim not to realize that Professor X, whom he has
just called a moron, may be somewhere nearby. Why else would B reply in
such an irrelevant manner? So in saying, “Nice day, isn’t it?” B conversation-
ally implies that Professor X is nearby.
Winston Churchill reportedly provided a famous example of intention-
ally violating the rule of Manner. When criticized for ending a sentence with
a preposition, he is said to have replied, “That is the type of criticism up
with which I will not put.”
41
Conversat ional Acts
These sentences appeared in Exercise VII. For each, explain what the speaker is
conversationally implying and how that conversational implication is generated.
1. “What did you get on the last test?” “A grade.”
2. “Did you like her singing?” “Her costume was beautiful.”
3. “The governor has the brains of a three-year-old.”
4. “Does the dog need to go out for a W-A-L-K [spelled out]?”
Exercise IX
For each of the following paired questions and answers, what do the answers
conversationally imply in a normal context? Explain why these conversational
implications hold. (Try to rely on the content of what is said, rather than on the
tone of voice in which it is uttered. In particular, don’t think of these remarks
being uttered with heavy sarcasm.)
1. Are you going to vote for a Republican? I just might.
2. Are you going to vote for a Republican? You can bet on it.
3. Are you going to vote for a Republican? Not unless hell freezes over.
4. Are you going to vote for a Republican? Don’t be silly.
5. Are you going to vote for a Republican? I am voting for an independent.
6. Are you going to vote for a Republican? There is no other choice.
7. Did you vote for a Republican? Maybe yes, maybe no.
8. Did you vote for a Republican? I voted for the winner.
Exercise X
RHETORICAL DEVICES
Many rhetorical devices work by flouting conversational rules in order to
generate conversational implications. Consider exaggeration. When some-
one claims to be hungry enough to eat a horse, it does not dawn on us
to treat this as a literal claim about how much she can eat. To do so would
be to attribute to the speaker a blatant violation of Grice’s first rule of
Quality—namely, do not say what you believe to be false. Consequently,
her audience will naturally interpret her remark figuratively, rather than
literally. They will assume that she is exaggerating the amount she can eat
in order to conversationally imply that she is very hungry. This rhetorical
device is called overstatement or hyperbole. It is commonly employed, often
in heavy-handed ways.
We sometimes use the opposite ploy and attempt to achieve rhetorical ef-
fect by understating things. We say that something is pretty good or not too
bad when, as all can see, it is terrific. In these cases the speaker is violating
something akin to the rule of Quantity. He is not saying just how good
something really is. He expects his audience to recognize this and say
(inwardly, at least) something like this: “Oh, it is much better than that.”
Understatement is often used as a way of fishing for compliments.
Sometimes, then, we do not intend to have others take our words at face
value. Even beyond this, we sometimes expect our listeners to interpret us
as claiming just the opposite of what we assert. This occurs, for example, with
irony and sarcasm. Suppose at a crucial point in a game, the second baseman
fires the ball ten feet over the first baseman’s head, and someone shouts,
“Great throw.” Literally, it was not a great throw; it was the opposite of
a great throw, and this is just what the person who says “Great throw” is
indicating. How do the listeners know they are supposed to interpret it in
this way? Sometimes this is indicated by tone of voice. A sarcastic tone of
voice usually indicates that the person means the opposite of what he or she
is saying. Even without the tone of sarcasm, the remark “Great throw” is not
likely to be taken literally. The person who shouts this knows that it was
not a great throw, as do the people who hear it. Rather than attributing an
obviously false belief to the shouter, we assume that the person is blatantly
violating the rule of Quality to draw our attention to just how bad the throw
really was.
42
CHAPTER 2 ■ The Web of Language
Image not available due to copyright restrictions
Metaphors and similes are perhaps the most common forms of figurative
language. A simile is, roughly, an explicit figurative comparison. A word
such as “like” or “as” makes the comparison explicit, and the comparison is
figurative because it would be inappropriate if taken literally. To say that the
home team fought like tigers does not mean that they clawed the opposing
team and took large bites out of them. To call someone as dumb as a post is
not to claim that they have no brain at all.
With a metaphor, we also compare certain items, but without words such
as “like” or “as.” Metaphorical comparisons are still figurative because the
vocabulary, at a literal level, is not appropriate to the subject matter. George
Washington was not literally the father of his country. Taken literally, it
hardly makes sense to speak of someone fathering a country. But the
metaphor is so natural (or so familiar) that it does not cross our minds to
treat the remark literally, asking, perhaps, who the mother was.
Taken literally, metaphors are usually obviously false, and then they
violate Grice’s rule of Quality. Again, as with irony, when someone says
something obviously false, we have to decide what to make of that person’s
utterance. Perhaps the person is very stupid or a very bad liar, but often
neither suggestion is plausible. In such a situation, sometimes the best sup-
position is that the person is speaking metaphorically rather than literally.
Not all metaphors, however, are literally false. In John Donne’s Medita-
tion XVII, “No man is an island” is literally true. We treat this remark as
a metaphor because, taken literally, it is so obviously and boringly true
that we cannot imagine why anyone would want to say it. Taken literally,
it would make no greater contribution to the conversation than any other ir-
relevant, obvious truth—for example, that no man is a socket wrench. Taken
literally, this metaphor violates the rule of Relevance and, perhaps, the sec-
ond part of the rule of Quantity. Taken figuratively, it is an apt, if somewhat
overworked, way of indicating that no one is isolated and self-contained.
43
Conversat ional Acts
Here are some more true metaphors. Explain what they mean and how they work.
1. “Blood is thicker than water.”
2. “Cream rises to the top.”
3. “People who live in glass houses should not throw stones.”
4. Robert Frost’s poem “The Road Not Taken” begins, “Two roads diverged
in a yellow wood.”
5. China’s Chairman Mao Tse-tung is reported to have said, “A revolution is
not the same as inviting people to dinner, or writing an essay, or painting
a picture, or doing fancy needle-work.”
6. Cuba’s Fidel Castro is supposed to have said, “A revolution is not a bed of
roses. A revolution is a struggle between the future and the past.” (Is this
different from item 5?)
Exercise XI
44
CHAPTER 2 ■ The Web of Language
Metaphors do not appear only in statements. They also appear in imperatives.
For example, “Don’t rock the boat” can be employed literally in a context where
someone is moving around in a canoe in a way that could tip it. It can also be
used metaphorically to tell someone not to do something that will cause a fuss.
For each of the following metaphors, find a context where the imperative can
be used in its literal way and another context where it is used metaphorically.
1. Keep your eye on the ball.
2. Don’t put all your eggs in one basket.
3. Look before you leap.
4. Make hay while the sun shines.
5. Don’t count your chickens before they hatch.
6. Don’t change horses in midstream.
Exercise XIII
Unpack the following political metaphors by giving their literal content.
1. We can’t afford a president who needs on-the-job training.
2. It’s time for people on the welfare wagon to get off and help pull.
3. If you can’t stand the heat, get out of the kitchen.
4. We need to restore a level playing field.
5. The special interests have him in their pockets.
6. The bill was passed through typical horse trading.
7. He’s a lame duck.
Exercise XIV
Identify each of the following sentences as irony, metaphor, or simile. For each
sentence, write another expressing its literal meaning.
1. He missed the ball by a mile.
2. He acted like a bull in a china shop.
3. The exam blew me away.
4. He had to eat his words.
5. It was a real team effort. (Said by a coach after his team loses by forty points.)
6. They are throwing the baby out with the bathwater.
7. The concert was totally awesome.
8. A midair collision can ruin your whole day.
9. This is a case of the tail wagging the dog.
10. “Religion is the opiate of the masses.” (Marx)
Exercise XII
45
Conversat ional Acts
1. A classic example of rhetoric occurs in Marc Antony’s funeral oration in
William Shakespeare’s play, Julius Caesar (act III, scene ii). Brutus and
other conspirators had killed Julius Caesar. At Caesar’s funeral, Brutus
first argued that they needed to kill Caesar to prevent him from becoming
too powerful and taking away the freedoms of Roman citizens. Brutus
concludes, “As Caesar loved me, I weep for him; as he was fortunate, I re-
joice in it; as he was valiant, I honor him. But as he was ambitious, I slew
him.” On the other side, Marc Antony sees Brutus as a traitor, but Brutus
now has power, so Antony does not dare to call Brutus a traitor openly.
The central part of Antony’s speech is reprinted below. Indicate which
lines are ironic, and comment on any other rhetorical devices in this
speech. Why was it so effective (and famous) as a speech?
Come I to speak in Caesar’s funeral.
He was my friend, faithful and just to me.
But Brutus says he was ambitious,
And Brutus is an honorable man.
He hath brought many captives home to Rome,
Whose ransoms did the general coffers fill.
Did this in Caesar seem ambition?
When that the poor have cried, Caesar hath wept—
Ambition should be made of sterner stuff.
Yet Brutus says he was ambitious,
And Brutus is an honorable man.
You all did see that on the Lupercal
I thrice presented him a kingly crown,
Which he did thrice refuse. Was this ambition?
Yet Brutus says he was ambitious,
And, sure, he is an honorable man.
I speak not to disprove what Brutus spoke,
But here I am to speak what I do know.
You all did love him once, not without cause.
What cause withholds you then to mourn for him?
2. At the start of the U.S. war with Iraq in 2003, some described Iraq as another
Vietnam, while others described Saddam Hussein (Iraq’s president) as
another Hitler. Which metaphor was used by supporters of the war? Which
was used by opponents? How can you tell? How do these metaphors work?
Discussion Questions
DECEPTION
In the preceding examples, a speaker openly violates a conversational rule.
The listeners recognize that a rule is being intentionally broken, and the
speaker knows that the listeners recognize the violation. At other times,
however, speakers intentionally break conversational rules because they are
trying to mislead their listeners. A speaker may violate the first part of
Grice’s rule of Quality by uttering something she knows to be false with the
intention of producing a false belief in her listeners. That is called lying. No-
tice that lying depends on the general acceptance of the Cooperative Princi-
ple. Because audiences generally assume that speakers are telling the truth,
speakers can sometimes get away with lying.
Flat-out lying is not the only way (and often not the most effective way)
of intentionally misleading people. We can say something literally true that,
at the same time, conversationally implies something false. This is some-
times called making a false suggestion. If a son tells his parents that he “has
had some trouble with the car,” that could be true but deeply misleading if,
in fact, he had totaled it. It would be misleading because it would violate the
rule of Quantity. In saying only that he has had some trouble with the car, he
conversationally implies that nothing very serious happened. He conversa-
tionally implies this because, in this context, he is expected to come clean
and reveal all that actually happened.
A more complex example of false suggestion arose in a lawsuit that
reached the United States Supreme Court.
BRONSTON V. UNITED STATES
(409 U.S. 352, 1973)
MR. CHIEF JUSTICE BURGER delivered the opinion of the Court:
Petitioner’s perjury conviction was founded on the answers given by him as
a witness at that bankruptcy hearing, and in particular on the following col-
loquy with a lawyer for a creditor of Bronston Productions:
Q. Do you have any bank accounts in Swiss banks, Mr. Bronston?
A. No, sir.
Q. Have you ever?
A. The company had an account there for about six months, in Zurich.
Q. Have you any nominees who have bank accounts in Swiss banks?
A. No, sir.
Q. Have you ever?
A. No, sir.
It is undisputed that for a period of nearly five years, between October 1959
and June 1964, petitioner had a personal bank account at the International
Credit Bank in Geneva, Switzerland, into which he made deposits and
upon which he drew checks totalling more than $180,000. It is likewise
undisputed that petitioner’s answers were literally truthful. (i) Petitioner
did not at the time of questioning have a Swiss bank account. (ii) Bronston
46
CHAPTER 2 ■ The Web of Language
Productions, Inc., did have the account in Zurich described by petitioner.
(iii) Neither at the time of questioning nor before did petitioner have nomi-
nees who had Swiss accounts. The government’s prosecution for per-
jury went forward on the theory that in order to mislead his questioner,
petitioner answered the second question with literal truthfulness but unre-
sponsively addressed his answer to the company’s assets and not to his
own—thereby implying that he had no personal Swiss bank account at the
relevant time.
It is hard to read the witness’s response to the second question in any other
way than as a deliberate attempt to mislead the Court, for his response
plainly suggests that he did not have a personal account in a Swiss bank,
when, in fact, he did. But the issue before the Court was not whether he in-
tentionally misled the Court, but whether in doing so he committed perjury.
The relevant statute reads as follows:
Whoever, having taken an oath before a competent tribunal . . . that he will testify
. . . truly, . . . willfully and contrary to such oath states or subscribes to any mate-
rial matter which he does not believe to be true, is guilty of perjury.
(18 U.S.C. 1621)
The lower courts ruled that Bronston violated this statute and, thus, commit-
ted perjury. The Supreme Court reversed this decision, in part for the follow-
ing reasons:
It should come as no surprise that a participant in a bankruptcy proceeding may
have something to conceal and consciously tries to do so, or that a debtor may be
embarrassed at his plight and yield information reluctantly. It is the responsibility
of the lawyer to probe; testimonial interrogation, and cross-examination in partic-
ular, is a probing, prying, pressing form of inquiry. If a witness evades, it is the
lawyer’s responsibility to recognize the evasion and to bring the witness back to
the mark, to flush out the whole truth with the tools of adversary examination.
(409 U.S. 352 at 358–359 [1973])
In other words, in a courtroom, where the relationship is typically adversar-
ial rather than cooperative, not all the standard conversational rules are in
force or fully in force. In particular, it would be unrealistic to assume that the
rule of Quantity will be consistently honored in a courtroom clash; therefore,
it becomes the task of the cross-examiner to force the witness to produce all
the relevant facts.
47
Conversat ional Acts
Refer back to the dialogue quoted in Bronston v. United States. Because it is dif-
ficult to read the witness’s second response as anything but a willful attempt
to deceive, why should this case be treated differently from lying? Alterna-
tively, why not even drop the demand that witnesses tell the truth and make it
the responsibility of the lawyers to get at the truth itself (rather than just the
whole truth) through “probing, prying, pressing” inquiry?
Discussion Questions
SUMMARY
In this chapter we have developed a rather complex picture of the way our
language functions. In the process, we have distinguished three kinds or lev-
els of acts that are performed when we employ language. We have also ex-
amined the rules associated with each kind or level of act. The following
table summarizes this discussion:
Three Levels Of Language
Kinds of Acts Governing Rules
A LINGUISTIC ACT is an act of saying something Semantic rules (such as definitions) and
meaningful in a language. It is the basic act that syntactic rules (as in grammar).
is needed to make anything part of language.
A SPEECH ACT concerns the move a person makes Speech act rules about special agents,
in saying something. Different kinds of speech acts formulas, circumstances, responses,
are indicated by the various verbs found in explicit and feelings appropriate to different
performatives. kinds of speech acts, discovered by
speech act analysis.
A CONVERSATIONAL ACT is a speaker’s act of Conversational rules (the Cooperative
causing a standard kind of effect in the listener; Principle; Quantity, Quality, Relevance,
it is what I do by saying something—for example, and Manner).
I persuade someone to do something.
48
CHAPTER 2 ■ The Web of Language
1. It is late, and A is very hungry. A asks B, “When will dinner be ready?”
Describe the linguistic act, the speech act, and some of the conversational
acts this person may be performing in this context.
2. Someone is trying to solve the following puzzle: One of thirteen balls is
heavier than the others, which are of equal weight. In no more than three
weighings on a balance scale, determine which ball is the heavier one. The
person is stumped, so someone says to her: “Begin by putting four balls in
each pan of the scale.” Describe the linguistic act, the speech act, and the
conversational act of the person who makes this suggestion.
Exercise XV
NOTES
1 J. L. Austin used the phrase “locutionary act” to refer to a level of language closely related to
what we refer to as a “linguistic act.” See J. L. Austin, How to Do Things with Words, 2nd ed.
(Cambridge, MA: Harvard University Press, 1975), 94–109.
2 See, for example, J. L. Austin’s How to Do Things with Words.
3 An example of the continuous present is “I bet ten dollars every week in the lottery.” Since this
sentence is not used to make a bet, this sentence and others with the continuous present do not
pass the thereby test or express explicit performatives.
4 Austin calls speech acts “illocutionary acts.” See How to Do Things with Words, 98–132.
5Although performative verbs name kinds of speech acts, not every kind of speech act has a cor-
responding performative verb. For example, insulting seems to be a kind of speech act, but “in-
sult” is not a performative verb, because you cannot insult someone simply by saying, “I insult
you.” We might have had a convention that enabled us to insult people just by saying, “I insult
you.” In English, however, we do not.
6Supreme Court of New Hampshire, 1929, 84 N.H. 114, A. 641.
7This discussion of conversational rules and implications is based on Paul Grice’s important es-
say, “Logic and Conversation,” which appears as the second chapter of his Studies in the Way of
Words (Cambridge, MA: Harvard University Press, 1989). To avoid British references that an
American reader might find perplexing, we have sometimes altered Grice’s wording.
8Grice states the Cooperative Principle in these words: “Make your conversational contribution
such as is required, at the stage at which it occurs, by the accepted purpose or direction of the
talk exchange in which you are engaged.”
49
Summary
This page intentionally left blank
THE LANGUAGE OF ARGUMENT
Using the techniques developed in Chapter 2, this chapter will examine the use of lan-
guage to formulate arguments and will provide methods to analyze genuine arguments
in their richness and complexity. The first stage in analyzing an argument is the dis-
covery of its basic structure. To do this, we will examine the words, phrases, and special
constructions that indicate the premises and conclusions of an argument. The second
stage is to explore the standards that arguments are supposed to meet. Here we will
focus on validity, truth, and soundness. The third stage is the study of techniques used
to protect an argument. These include guarding premises so that they are less subject
to criticism, offering assurances concerning debatable claims, and discounting possi-
ble criticisms in advance.
ARGUMENT MARKERS
In Chapter 2, we saw that language is used for a great many different pur-
poses. One important thing that we do with language is construct arguments.
Arguments are constructed out of statements, but arguments are not just lists
of statements. Here is a simple list of statements:
Socrates is a man.
All men are mortal.
Socrates is mortal.
This list is not an argument, because none of these statements is presented as
a reason for any other statement. It is, however, simple to turn this list into an
argument. All we have to do is to add the single word “therefore“:
Socrates is a man.
All men are mortal.
Therefore, Socrates is mortal.
Now we have an argument. The word “therefore” converts these sentences
into an argument by signaling that the statement following it is a conclusion,
and the statement or statements that come before it are offered as reasons
on behalf of this conclusion. The argument we have produced in this way
3
51
52
CHAPTER 3 ■ The Language of Argument
is a good one, because the conclusion follows from the reasons stated on
its behalf.
There are other ways of linking these sentences to form an argument.
Here is one:
Since Socrates is a man,
and all men are mortal,
Socrates is mortal.
Notice that the word “since” works in roughly the opposite way that
“therefore” does. The word “therefore” is a conclusion marker, because it
indicates that the statement that follows it is a conclusion. In contrast, the
word “since” is a reason marker, because it indicates that the following state-
ment or statements are reasons. In our example, the conclusion comes at the
end, but there is a variation on this. Sometimes the conclusion is given at
the start:
Socrates is mortal, since all men are mortal and Socrates is a man.
“Since” flags reasons; the remaining connected statement is then taken to
be the conclusion, whether it appears at the beginning or at the end of the
sentence.
Many other terms are used to introduce an argumentative structure into
language by marking either reasons or conclusions. Here is a partial list:
REASON MARKERS CONCLUSION MARKERS
since therefore
because hence
for thus
as then
We shall call such terms “argument markers,” because each presents one
or more statements as part of an argument or backing for some other
statement.
It is important to realize that these words are not always used as argument
markers. The words “since” and “then” are often used as indicators of time, as
in, “He’s been an American citizen since 1973” and “He ate a hot dog, then a
hamburger.” The word “for” is often used as a preposition, as in “John works
for IBM.” Because some of these terms have a variety of meanings, it is not pos-
sible to identify argument markers in a mechanical way just by looking at
words. It is necessary to examine the function of words in the context in which
they occur. One test of whether a word is functioning as an argument marker
in a particular sentence is whether you can substitute another argument
marker without changing the meaning of the sentence. In the last example, it
makes no sense to say, “John works since IBM.”
Many phrases are also available to signal that an argument is being given.
Here is just a small sample:
53
Argument Markers
from which it follows that . . .
from which we may conclude that . . .
from which we see that . . .
which goes to show that . . .
which establishes that . . .
We can also indicate conclusions and reasons by using argumentative perfor-
matives, which we examined briefly in Chapter 2. If someone says, “I conclude
that . . . ,” the words that follow are given the status of a conclusion. More pre-
tentiously, if someone says, “Here I base my argument on the claim that . . . ,”
what comes next has the status of a reason.
Examination of actual arguments will show that we have a great many
ways of introducing an argumentative structure into our language by using
the two forms of argument markers: reason markers and conclusion mark-
ers. The first, and in many ways the most important, step in analyzing an
argument is to identify the conclusion and the reasons given on its behalf.
We do this by paying close attention to these argument markers.
IF . . . , THEN . . .
If-then sentences, which are also called conditionals, often occur in arguments,
but they do not present arguments by themselves. To see this, consider the
following conditional:
If the Dodgers improve their hitting, then they will win the
Western Division.
The sentence between the “if” and the “then” is called the antecedent of the
conditional. The sentence after the “then” is called its consequent. In utter-
ing such a conditional, we are not asserting the truth of its antecedent,
and we are not asserting the truth of its consequent either. Thus, the per-
son who makes the above remark is not claiming that the Dodgers will
win the Western Division. All she is saying is that if they improve their
hitting, then they will win. Furthermore, she is not saying that they will
improve their hitting. Because the speaker is not committing herself to
either of these claims, she is not presenting an argument. This becomes
clear when we contrast this conditional with a statement that does formu-
late an argument:
Conditional: If the Dodgers improve their hitting, then they will win the
Western Division.
Argument: Since the Dodgers will improve their hitting, they will win the
Western Division.
The sentence that follows the word “since” is asserted. That is why “since”
is an argument marker, whereas the connective “if . . . then . . .” is not an
argument marker.
54
CHAPTER 3 ■ The Language of Argument
Even though conditionals by themselves do not mark arguments, there
is a close relationship between conditionals and arguments: Indicative
conditionals provide patterns that can be converted into an argument
whenever the antecedent is said to be true. (We also get an argument when
the consequent is said to be false, but we will focus here on the simpler
case of asserting the antecedent.) Thus, we often hear people argue in the
following way:
If inflation continues to grow, there will be an economic crisis. But
inflation will certainly continue to grow, so an economic crisis is on
the way.
The first sentence is an indicative conditional. It makes no claims one way or
the other about whether inflation will grow or whether an economic crisis
will occur. The next sentence asserts the antecedent of this conditional and
then draws a conclusion signaled by the argument marker “so.” We might
say that when the antecedent of an indicative conditional is found to be true,
the conditional can be cashed in for an argument.
Often the antecedent of a conditional is not asserted explicitly but is con-
versationally implied. When asked which player should be recruited for a
team, the coach might just say, “If Deon is as good as our scouts say he is,
then we ought to go for Deon.” This conditional does not actually assert that
Deon is as good as the scouts report. Nonetheless, it would be irrelevant and
pointless for the coach to utter this conditional alone if he thought that the
scouts were way off the mark. The coach might immediately add that he
disagrees with the scouting reports. But unless the coach cancels the conver-
sational implication in some way, it is natural to interpret him as giving an
argument that we ought to pick Deon. In such circumstances, then, an
indicative conditional can conversationally imply an argument, even though
it does not state the argument explicitly.
This makes it easy to see why indicative conditionals are a useful fea-
ture of our language. By providing patterns for arguments, they prepare
us to draw conclusions when the circumstances are right. Much of our
knowledge of the world around us is contained in such conditionals. Here is
an example: If your computer does not start, the plug might be loose. This is
a useful piece of practical information, for when your computer does not
start, you can immediately infer that the plug might be loose, so you know
to check it out.
Other words function in similar ways. When your computer fails to
start, a friend might say, “Either the plug is loose or you are in deep
trouble.” Now, if you also assert, “The plug is not loose,” you can conclude
that you are in deep trouble. “Either . . . or . . .” sentences thus provide
patterns for arguments, just as conditionals do. However, neither if-then
sentences nor either-or sentences by themselves explicitly assert enough to
present a complete argument, so “if . . ., then . . .” and “either . . . or . . .”
should not be labeled as argument markers.
55
Arguments in Standard Form
Indicate which of the following italicized words or phrases is a reason marker,
a conclusion marker, or neither.
1. He apologized, so you should forgive him.
2. He apologized. Accordingly, you should forgive him.
3. Since he apologized, you should forgive him.
4. Provided that he apologized, you should forgive him.
5. In view of the fact that he apologized, you should forgive him.
6. He apologized. Ergo, you should forgive him.
7. Given that he apologized, you should forgive him.
8. He apologized, and because of that you should forgive him.
9. After he apologizes, you should forgive him.
10. He apologized. As a result, you should forgive him.
11. Seeing as he apologized, you should forgive him.
12. He apologized. For that reason alone, you should forgive him.
Exercise I
Indicate whether each of the following sentences is an argument.
1. Charles went bald, and most men go bald.
2. Charles went bald because most men go bald.
3. My roommate likes to ski, so I do, too.
4. My roommate likes to ski, and so do I.
5. I have been busy since Tuesday.
6. I am busy, since my teacher assigned lots of homework.
Exercise II
ARGUMENTS IN STANDARD FORM
Because arguments come in all shapes and forms, it will help to have a
standard way of presenting arguments. For centuries, logicians have used
a format of the following kind:
(1) All men are mortal.
(2) Socrates is a man.
(3) Socrates is mortal. (from 1–2)
56
CHAPTER 3 ■ The Language of Argument
The reasons (or premises) are listed and numbered. Then a line is drawn be-
low the premises. Next, the conclusion is numbered and written below the
line. The symbol “ “, which is read “therefore,” is then added to the left of
the conclusion in order to indicate the relation between the premises and the
conclusion. Finally, the premises from which the conclusion is supposed to
be derived are indicated in parentheses. Arguments presented in this way
are said to be in standard form.
The notion of a standard form is useful because it helps us see that the
same argument can be expressed in different ways. For example, the follow-
ing three sentences formulate the argument that was given in standard form
above.
Socrates is mortal, since all men are mortal, and Socrates is a man.
All men are mortal, so Socrates is mortal, because he is a man.
All men are mortal, and Socrates is a man, which goes to show that
Socrates is mortal.
More important, by putting arguments into standard form, we perform
the most obvious, and in some ways most important, step in the analysis of
an argument: the identification of its premises and conclusion.
Identify which of the following sentences expresses an argument. For each that
does, (1) circle the argument marker (or markers), (2) indicate whether it is a
reason marker or a conclusion marker, and (3) restate the argument in stan-
dard form.
1. Since Chicago is north of Boston, and Boston is north of Charleston,
Chicago is north of Charleston.
2. Toward evening, clouds formed and the sky grew darker; then the storm
broke.
3. Texas has a greater area than Topeka, and Topeka has a greater area than
the Bronx Zoo, so Texas has a greater area than the Bronx Zoo.
4. Both houses of Congress may pass a bill, but the president may still
veto it.
5. Other airlines will carry more passengers, because United Airlines is on
strike.
6. Since Jesse James left town, taking his gang with him, things have been a
lot quieter.
7. Things are a lot quieter, because Jesse James left town, taking his gang
with him.
8. Witches float because witches are made of wood, and wood floats.
9. The hour is up, so you must hand in your exams.
10. Joe quit, because his boss was giving him so much grief.
Exercise III
57
Some Standards for Evaluat ing Arguments
SOME STANDARDS FOR EVALUATING ARGUMENTS
Not all arguments are good arguments; so, having identified an argument,
the next task is to evaluate the argument. Evaluating arguments is a complex
business. In fact, this entire book is aimed primarily at developing proce-
dures for doing so. There are, however, certain basic terms used in evaluat-
ing arguments that should be introduced from the start. They are validity,
truth, and soundness. Here they will be introduced informally. Later (in
Chapters 6–7) they will be examined with more rigor.
VALIDITY
In some good arguments, the conclusion is said to follow from the premises.
However, this commonsense notion of following from is hard to pin down
precisely. The conclusion follows from the premises only when the content
of the conclusion is related appropriately to the content of the premises, but
which relations count as appropriate?
To avoid this difficult question, most logicians instead discuss whether an
argument is valid. Calling something “valid” can mean a variety of things,
but in this context validity is a technical notion. Here “valid” does not mean
“good,” and “invalid” does not mean “bad.” This will be our definition of
validity:
An argument is valid if and only if it is not possible that all of its premises
are true and its conclusion false.
Alternatively, one could say that its conclusion must be true if its premises
are all true (or, again, that at least one of its premises must be false if its con-
clusion is false). The point is that a certain combination—true premises and
a false conclusion—is ruled out as impossible.
The following argument passes this test for validity:
(1) All senators are paid.
(2) Sam is a senator.
(3) Sam is paid. (from 1–2)
Clearly, if the two premises are both true, there is no way for the conclu-
sion to fail to be true. To see this, just try to tell a coherent story in which
every single senator is paid and Sam is a senator, but Sam is not paid. You
can’t do it.
Contrast this example with a different argument:
(1) All senators are paid.
(2) Sam is paid.
(3) Sam is a senator. (from 1–2)
Here the premises and the conclusion are all in fact true, let’s assume,
but that is still not enough to make the argument valid, because validity
58
CHAPTER 3 ■ The Language of Argument
concerns what is possible or impossible, not what happens to be true. This
conclusion could be false even when the premises are true, for Sam could
leave the Senate but still be paid for some other job, such as lobbyist. That
possibility shows that this argument is invalid.
Another very common form of argument is called modus ponens:
(1) If it is snowing, then the roads are slippery.
(2) It is snowing.
(3) The roads are slippery. (from 1–2)
This argument is valid, because it is not possible for its premises to be true
when its conclusion is false. We can show that by assuming that the con-
clusion is false and then reasoning backwards. Imagine that the roads are
not slippery. Then there are two possibilities. Either it is snowing or it
is not snowing. If it is not snowing, then the second premise is false. If it
is snowing, then the first premise must be false, since we are supposing
that it is snowing and that the roads are not slippery. Thus, at least one
premise has to be false when the conclusion is false. Hence, this argument
is valid.
This argument might seem similar to another:
(1) If it is snowing, then the roads are slippery.
(2) It is not snowing.
(3) The roads are not slippery. (from 1–2)
This argument is clearly invalid, because there are several ways for its prem-
ises to be true when its conclusion is false. It might have just stopped snow-
ing or ice might make the roads slippery. Then the roads are slippery, so the
conclusion is false, even if both premises are true.
Yet another form of argument is often called process of elimination:
(1) Either Joe or Jack or Jim or Jerry committed the murder.
(2) Joe didn’t do it.
(3) Jack didn’t do it.
(4) Jim didn’t do it.
(5) Jerry committed the murder. (from 1–4)
The first premise asserts that at least one of these four suspects is guilty. That
couldn’t be true if all of the other premises were true and the conclusion were
false, because that combination would exclude all four of these suspects.
So this argument is valid.
Now compare this argument:
(1) Either Joe or Jack or Jim or Jerry committed the murder.
(2) Joe did it.
(3) Jerry did not commit the murder. (from 1–2)
59
Some Standards for Evaluat ing Arguments
To show that this argument is invalid, all we have to do is explain how the
premises could be true and the conclusion false. Here’s how: Joe and Jerry
did it together. In that case, Jerry did it, so the conclusion is false; Joe also
did it, so the second premise is true; and the first premise is true, because it
says that at least one of these four suspects did it, and that is true when more
than one of the suspects did it. That possibility of complicity, thus, makes
this argument invalid.
We will explore many more forms of argument in Chapters 6–7. The
goal for now is just to get a feel for how to determine validity. In all of these
examples, an argument is said to be valid if and only if there is no possible
situation in which its premises are true and its conclusion is false. You need
to figure out whether there could be any situation like this in order to deter-
mine whether an argument is valid. If so, the argument is invalid. If not, it
is valid.
This definition shows why validity is a valuable feature for an argument
to possess: There can be no valid argument that leads one from true prem-
ises to a false conclusion. This should square with your commonsense ideas
about reasoning. If you reason well, you should not be led from truth into
error.
What are known as deductive arguments are put forward as meeting this
standard of validity, so validity is one criterion for a good deductive argu-
ment. Other arguments—so-called inductive arguments—are not presented
as meeting this standard. Roughly, an inductive argument is presented as
providing strong support for its conclusion. The standards for evaluating in-
ductive arguments will be examined in Chapter 8. For now we will concen-
trate on deductive arguments.
TRUTH
Although a deductive argument must be valid in order to be a good
argument, validity is not enough. One reason is that an argument can be
valid even when some (or all) of the statements it contains are false. For
example:
(1) No fathers are female.
(2) Sam is a father.
(3) Sam is not female. (from 1–2)
Suppose that Sam has no children or that Sam is female, so premise 2 is false.
That would be a serious defect in this argument. Nonetheless, this argument
satisfies our definition of validity: If the premises were true, then the conclu-
sion could not be false. There is no way that Sam could be female if Sam is a
father and no fathers are female. This example makes it obvious that validity
is not the same as truth. It also makes it obvious that another requirement of
a good argument is that all of its premises must be true.
60
CHAPTER 3 ■ The Language of Argument
SOUNDNESS
We thus make at least two demands of a deductive argument:
1. The argument must be valid.
2. The premises must be true.
When an argument meets both of these standards, it is said to be sound. If it
fails to meet either one or the other, then it is unsound. Thus, an argument is
unsound if it is invalid, and it is also unsound if at least one of its premises
is false.
ALL PREMISES TRUE AT LEAST ONE FALSE PREMISE
Valid Sound Unsound
Invalid Unsound Unsound
Soundness has one great benefit: A sound argument must have a true con-
clusion. We know this because its premises are true and, since it is valid,
it is not possible that its premises are true and its conclusion is false. This
is why people who seek truth want sound arguments, not merely valid
arguments.
A TRICKY CASE
Our definition of validity yields a surprising result. Consider the following
argument:
(1) Frogs are green.
(2) Frogs are not green.
(3) I am president. (from 1–2)
It is obviously not possible for both premises to be true, so it is also not pos-
sible that both premises are true when the conclusion is false. Consequently,
this argument fits our definition of validity. So does any other argument
whose premises cannot be true. Such arguments cannot ever take us from
truths to falsehoods, because they never start with truths.
This weird example illustrates some of the ways in which the technical
notion of validity differs from the commonsense notion of following from.
The content of its premises has no relation to the content of the conclusion.
Frogs have nothing to do with who is president. Hence, the conclusion does
not follow from the premises. But that does not prevent the argument from
being valid.
This example also shows the importance of distinguishing validity from
soundness. Any argument whose premises cannot all be true is valid, no
matter how ridiculous its conclusion. However, an argument cannot be
sound if its premises can’t be true, and a valid argument that is unsound can-
not show that its conclusion is true. Consequently, this strange case won’t
cause any trouble.
61
Some Standards for Evaluat ing Arguments
Indicate whether each of the following arguments is valid and whether it is
sound. Explain your answers where necessary.
1. Most professors agree that they are paid too little, so they are.
2. David Letterman is over four feet tall, so he is over two feet tall.
3. Lee can’t run a company right, because he can’t do anything right.
4. Barack Obama is smart and good-looking, so he is smart.
5. Barack Obama is either a Democrat or a Republican, so he is a Democrat.
6. Since Jimmy Carter was president, he must have won an election.
7. Since Gerald Ford was president, he must have won an election.
8. Pat is either a mother or a father. If Pat is a mother, then she is a parent. If
Pat is a father, he is a parent. So, either way, Pat is a parent. (Assume that
this conclusion is true.)
9. People who live in the Carolinas live in either North Carolina or
South Carolina. Hillary Clinton does not live in North Carolina or
South Carolina. Hence, she does not live in the Carolinas.
10. If all of Illinois were in Canada, then Chicago would be in Canada. But
Chicago is not in Canada. Therefore, not all of Illinois is in Canada.
11. If George lives in Crawford, then George lives in Texas. If George lives in
Texas, then George lives in the United States. Hence, if George lives in
Crawford, he lives in the United States.
12. There can’t be a largest six-digit number, because six-digit numbers are
numbers, and there is no largest number.
Exercise IV
Assume that the following sentences are either true (T) or false (F) as indicated.
All my children are teenagers. (T)
All teenagers are students. (T)
All teenagers are my children. (F)
All my children are students. (T)
Using these assigned values, label each of the following arguments as (a) ei-
ther valid or invalid, and (b) either sound or unsound.
1. All my children are teenagers.
All teenagers are students.
All my children are students.
2. All my children are students.
All teenagers are students.
All my children are teenagers.
Exercise V
(continued)
62
CHAPTER 3 ■ The Language of Argument
A PROBLEM AND SOME SOLUTIONS
Although soundness guarantees a true conclusion, we usually expect even
more from an argument than soundness. In the first place, an argument can
be sound but trivially uninteresting:
(1) Nigeria is in Africa.
(2) Nigeria is in Africa. (from 1)
Compare these arguments:
(1) Al Gore was president. (1*) George W. Bush was president.
(2) Frogs are frogs. (2*) Frogs are frogs.
Are these arguments valid? Why or why not? Are they sound? Why or why
not? Is anything wrong with the argument on the right side? If so, what?
Discussion Question
3. All teenagers are my children.
All my children are students.
All teenagers are students.
4. All teenagers are students.
All my children are students.
All my children are students.
Indicate whether each of the following sentences is true. For those that are
true, explain why they are true. For those that are false, show why they are
false by giving a counterexample.
1. Every argument with a false conclusion is invalid.
2. Every argument with a false premise is invalid.
3. Every argument with a false premise and a false conclusion is invalid.
4. Every argument with a false premise and a true conclusion is invalid.
5. Every argument with true premises and a false conclusion is invalid.
6. Every argument with a true conclusion is sound.
7. Every argument with a false conclusion is unsound.
Exercise VI
63
A Problem and Some Solut ions
Here the premise is true. The argument is also valid, because the premise
cannot be true without the conclusion (which repeats it) being true as well. Yet
the argument is completely worthless as a proof that Nigeria is in Africa. The
reason is that this argument is circular. We will examine circular arguments
in detail in Chapter 16, but it should already be clear why such arguments
are useless. If A is trying to justify something to B that B has doubts about,
then citing the very matter in question will not do any good. Explanations of
a phenomenon that cite that very phenomenon itself also fail to increase our
understanding. In general, for A to argue successfully, A must marshal facts
that B accepts and then show that they justify or explain the conclusion. In
circular arguments, the worries about the conclusion immediately turn into
worries about the premise as well.
Now, however, A seems to run into a problem. A cannot cite a proposition
as a reason for itself, for that would be circular reasoning. If, however, A cites
some other propositions as premises leading to the conclusion, then the
question naturally arises why these premises should be accepted. Does A
not have to present arguments for them as well? Yet if A does that, then A
will introduce further premises that are also in need of proof, and so on in-
definitely. It now looks as if every argument, to be successful, will have to
be infinitely long.
This potential regress causes deep problems in theoretical philosophy. In
everyday life, however, we try to avoid these problems by relying on shared
beliefs—beliefs that will not be challenged. Beyond this, we expect people to
believe us when we cite information that only we possess. But there are lim-
its to this expectation, for we all know that people sometimes believe things
that are false and sometimes lie about what they know to be true. This pres-
ents a practical problem: How can we present our reasons in a way that does
not produce just another demand for an argument—a demand for more rea-
sons? Here we use three main strategies:
1. Assuring: Indicating that there are backup reasons even though we are
not giving them fully right now.
2. Guarding: Weakening our claims so that they are less subject to attack.
3. Discounting: Anticipating criticisms and dismissing them.
In these three ways we build a defensive perimeter around our premises.
Each of these defenses is useful, but each can also be abused.
ASSURING
When will we want to give assurances about some statement we have made?
If we state something that we know everyone believes, assurances are not nec-
essary. For that matter, if everyone believes something, we may not even state
it at all; we let others fill in this step in the argument. We offer assurances
when we think that someone might doubt or challenge what we say.
64
CHAPTER 3 ■ The Language of Argument
There are many ways to give assurances. Sometimes we cite authorities:
Doctors agree . . .
Recent studies have shown . . .
An unimpeachable source close to the White House says . . .
It has been established that . . .
Here we indicate that authorities have these reasons without specifying
what their reasons are. We merely indicate that good reasons exist, even if
we ourselves cannot—or choose not to—spell them out. When the authority
cited can be trusted, this is often sufficient, but authorities often can and
should be questioned. This topic will be discussed more fully in Chapter 15.
Another way to give assurances is to comment on the strength of our own
belief:
I’m certain that . . .
I’m sure that . . .
I can assure you that . . .
I’m not kidding. . . .
Over the years, I have become more and more convinced that . . .
Again, when we use these expressions, we do not explicitly present
reasons, but we conversationally imply that there are reasons that back
our assertions.
A third kind of assurance abuses the audience:
Everyone with any sense agrees that . . .
Of course, no one will deny that . . .
It is just common sense that . . .
There is no question that . . .
Nobody but a fool would deny that . . .
These assurances not only do not give any reason; they also suggest that
there is something wrong with you if you ask for a reason. We call this the
trick of abusive assurances.
Just as we can give assurances that something is true, we can also give as-
surances that something is false. For example,
It is no longer held that . . .
It is wholly implausible to suppose that . . .
No intelligent person seriously maintains that . . .
You would have to be pretty dumb to think that . . .
The last three examples clearly involve abusive assurances.
Although many assurances are legitimate, we as critics should always
view assurances with some suspicion. People tend to give assurances only
65
A Problem and Some Solut ions
when they have good reasons to do so. Yet assuring remarks often mark the
weakest parts of the argument, not the strongest. If someone says “I hardly
need argue that . . . ,” it is often useful to ask why she has gone to the trou-
ble of saying this. When we distrust an argument—as we sometimes do—
this is precisely the place to look for weakness. If assurances are used, they
are used for some reason. Sometimes the reason is a good one. Sometimes,
however, it is a bad one. In honest argumentation, assurances save time and
simplify discussion. In a dishonest argument, they are used to paper over
cracks.
GUARDING
Guarding represents a different strategy for protecting premises from attack.
We reduce our claim to something less strong. Thus, instead of saying “all,”
we say “many.” Instead of saying something straight out, we use a qualify-
ing phrase, such as “it is likely that . . .” or “it is very possible that. . . .” Law
school professors like the phrase “it is arguable that. . . .” This is wonderfully
noncommittal, for it does not indicate how strong the argument is, yet it
does get the statement into the discussion.
Broadly speaking, there are three main ways of guarding what we say:
1. Weakening the extent of what has been said: retreating from “all” to
“most” to “a few” to “some,” and so on.
2. Introducing probability phrases such as “It is virtually certain that . . . ,”
“It is likely that . . . ,” “It might happen that . . . ,” and so on.
3. Reducing our level of commitment: moving from “I know that . . .” to “I
believe that . . .” to “I suspect that . . . ,” and so on.
Such terms guard premises when they are used in place of stronger alter-
natives. “Madison probably quit the volleyball team” is weaker than “She
definitely quit” but stronger than “She could have quit.” Thus, if the con-
text makes one expect a strong claim, such as “I know she quit,” then it is
guarding to say, “She probably quit.” In contrast, if the context is one of
speculating about who might have quit the team, then it is not guarding to
say, “She probably quit.” That is a relatively strong claim when others are
just guessing. Thus, you need to pay careful attention to the context in
order to determine whether a term has the function of guarding. When a
term is used for guarding, you should be able to specify a stronger claim
that the guarding term replaces and why that stronger term would be
expected in the context.
Guarding terms and phrases are often legitimate and useful. If you want
to argue that a friend needs fire insurance for her house, you do not need to
claim that her house will burn down. All you need to claim is that there is a
significant chance that her house will burn down. Your argument is better if
you start with this weaker premise, because it is easier to defend and it is
enough to support your conclusion.
66
CHAPTER 3 ■ The Language of Argument
If we weaken a claim sufficiently, we can make it completely immune to
criticism. What can be said against a remark of the following kind: “There
is some small chance that perhaps a few politicians are honest on at least
some occasions“? You would have to have a very low opinion of politicians
to deny this statement. On the other hand, if we weaken a premise too
much, we pay a price. The premise no longer gives strong support to the
conclusion.
The goal in using guarding terms is to find a middle way: We should
weaken our premises sufficiently to avoid criticism, but not weaken them so
much that they no longer provide strong enough evidence for the conclu-
sion. Balancing these factors is one of the most important strategies in mak-
ing and criticizing arguments.
Just as it was useful to zero in on assuring terms, so it is also useful to
keep track of guarding terms. One reason is that, like assuring terms, guard-
ing terms are easily corrupted. A common trick is to use guarding terms to
insinuate things that cannot be stated explicitly in a conversation. Consider
the effect of the following remark: “Perhaps the secretary of state has not
been candid with the Congress.” This does not actually say that the secre-
tary of state has been less than candid with the Congress, but, by the rule of
Relevance, clearly suggests it. Furthermore, it suggests it in a way that is
hard to combat.
A more subtle device for corrupting guarding terms is to introduce a
statement in a guarded form and then go on to speak as if it were not
guarded at all.
Perhaps the secretary of state has not been candid with the Congress. Of
course, he has a right to his own views, but this is a democracy where
officials are accountable to Congress. It is time for him to level with us.
The force of the guarding term “perhaps” that begins this passage disappears
at the end, where it is taken for granted that the secretary of state has not
been candid. This can be called the trick of the disappearing guard.
What is commonly called hedging is a sly device that operates in the op-
posite direction from our last example. With hedging, one shifts ground
from a strong commitment to something weaker. Things, as they say, get
“watered down” or “taken back.” Strong statements made at one stage of an
argument are later weakened without any acknowledgment that the posi-
tion has thereby been changed in a significant way. A promise to pass a piece
of legislation is later whittled down to a promise to bring it to a vote.
DISCOUNTING
The general pattern of discounting is to cite a possible criticism in order to
reject it or counter it. Notice how different the following statements sound:
The ring is beautiful, but expensive.
The ring is expensive, but beautiful.
67
A Problem and Some Solut ions
Both statements assert the same facts—that the ring is beautiful and that the
ring is expensive. Both statements also suggest that there is some opposition
between these facts. Yet these statements operate in different ways. We might
use the first as a reason for not buying the ring; we can use the second as
a reason for buying it. The first sentence acknowledges that the ring is
beautiful, but overrides this by pointing out that it is expensive. In reverse
fashion, the second statement acknowledges that the ring is expensive, but
overrides this by pointing out that it is beautiful. Such assertions of the form
“A but B” thus have four components:
1. The assertion of A
2. The assertion of B
3. The suggestion of some opposition between A and B
4. The indication that the truth of B is more important than the truth of A
The word “but” thus discounts the statement that comes before it in favor of
the statement that follows it.
“Although” is also a discounting connective, but it operates in reverse
fashion from the word “but.” We can see this, using the same example:
Although the ring is beautiful, it is expensive.
Although the ring is expensive, it is beautiful.
Here the statement following the word “although” is discounted in favor of
the connected statement.
A partial list of terms that typically function as discounting connectives
includes the following conjunctions:
although even if but nevertheless
though while however nonetheless
even though whereas yet still
These terms are not always used to discount. The word “still,” for example, is
used for discounting in (a) “He is sick; still, he is happy” but not in (b) “He is
still happy” (or “Sit still”). We can tell whether a term is being used for dis-
counting by asking whether the sentence makes sense when we substitute
another discounting term: It makes sense to say, “He is sick, but he is happy.”
It makes no sense to say, “He is but happy.” It is also illuminating to try to spec-
ify the objection that is being discounted. If you cannot say which objection is
discounted, then the term is probably not being used for discounting.
The clearest cases of discounting occur when we are dealing with facts
that point in different directions. We discount the facts that go against the
position we wish to take. But discounting is often more subtle than this. We
sometimes use discounting to block certain conversational implications of
what we have said. This comes out in examples of the following kind:
Jones is an aggressive player, but he is not dirty.
The situation is difficult, but not hopeless.
68
CHAPTER 3 ■ The Language of Argument
The Republicans have the upper hand in Congress, but only for the
time being.
A truce has been declared, but who knows for how long?
Take the first example. There is no opposition between Jones being aggres-
sive and his not being dirty. Both would be reasons to pick Jones for our
team. However, the assertion that Jones is aggressive might suggest that he
is dirty. The “but” clause discounts this suggestion without, of course, deny-
ing that Jones is aggressive.
The nuances of discounting terms can be subtle, and a correct analysis is
not always easy. All the same, the role of discounting terms is often impor-
tant. It can be effective in an argument to beat your opponents to the punch
by anticipating and discounting criticisms before your opponents can raise
them. The proper use of discounting can also help you avoid side issues and
tangents.
Still, discounting terms, like the other argumentative terms we have
examined, can be abused. People often spend time discounting weak objec-
tions to their views in order to avoid other objections that they know are
harder to counter. Another common trick is to discount objections no one
would raise. This is called attacking straw men. Consider the following re-
mark: “A new building would be great, but it won’t be free.” This does not
actually say that the speaker’s opponents think we can build a new building
for free, but it does conversationally imply that they think this, because oth-
erwise it would be irrelevant to discount that objection. The speaker is thus
trying to make the opponents look bad by putting words in their mouths
that they would never say themselves. To counter tricks like this, we need to
ask whether a discounted criticism is one that really would be raised, and
whether there are stronger criticisms that should be raised.
For each of the numbered words or expressions in the following sentences, in-
dicate whether it is an argument marker, an assuring term, a guarding term, a
discounting term, or none of these. For each argument marker, specify what
the conclusion and the reasons are, and for each discounting term, specify
what criticism is being discounted and what the response to this criticism is.
1. Although [1] no mechanism has been discovered, most [2] researchers in the
field agree [3] that smoking greatly increases the chances [4] of heart disease.
2. Since [5] historically [6] public debt leads to inflation, I maintain [7] that,
despite [8] recent trends, inflation will return.
3. Take it from me [9], there hasn’t been a decent center fielder since [10] Joe
DiMaggio.
4. Whatever anyone tells you [11], there is little [12] to the rumor that Queen
Elizabeth II will step down for [13] her son, Prince Charles.
Exercise VII
69
E V A L U A T I V E L A N G U A G E
EVALUATIVE LANGUAGE
Arguments are often filled with evaluation. The clearest cases of evaluative
language occur when we say that something is good or bad, that some course
of action is right or wrong, or that something should or should not (or ought to
or ought not to) be done. The meaning of such evaluative terms is very con-
troversial, but we can begin to understand evaluative language by asking
which acts—linguistic, speech, and conversational—it is used to perform.
Evaluative terms often come into play when one is faced with a choice or
decision. If you are deciding which shirt to buy, and a friend tells you, “That
one’s nice,” your friend would normally be taken to be prescribing that
you buy it. A passenger who says, “That’s the wrong turn,” is telling the
driver not to turn that way. In such contexts, evaluative statements are
action guiding—that is, they are used to direct someone to do or refrain
from doing some action. Evaluative terms do not, however, have such direct
prescriptive force when applied to things in remote times or places. Saying
that it was wrong for James Earl Ray to assassinate Martin Luther King does
not tell Ray not to do anything, for it is idle to address imperatives to people
in the past. Someone who says that it would be wrong for the president to
pursue a particular policy is not telling the president not to pursue it, unless
she happens to be speaking or writing to the president. Nevertheless, even
5. The early deaths of Janis Joplin and Jimi Hendrix show [14] that drugs are
really [15] dangerous.
6. I think [16] he is out back somewhere.
7. I think [17], therefore [18] I am.
8. I concede [19] that the evidence is hopelessly [20] weak, but [21] I still think
he is guilty.
9. I deny [22] that I had anything [23] to do with it.
10. The wind has shifted to the northeast, which means [24] that snow is
likely [25].
1. Construct three new and interesting examples of statements containing
assuring terms.
2. Do the same for guarding terms.
3. Do the same for discounting terms, and indicate which statement is being
discounted in favor of the other.
4. Do the same for argument markers, and indicate what is presented as a
reason for what.
Exercise VIII
70
CHAPTER 3 ■ The Language of Argument
in cases where evaluations lack direct prescriptive force, they sometimes
have prescriptive force indirectly. Calling Ray’s past action bad or wrong
might, through analogy, be a way of telling someone not to do future acts
like it. Evaluative language is, in these ways, used to perform speech acts of
prescribing action.
Evaluative language is also often used for other speech acts. When a fan
says, “That band is great,” this usually expresses admiration for their music
and a desire to hear more. After a meal, someone who announces, “That was
horrible,” is often expressing aversion or even disgust at the food. To say,
“That’s too bad,” is often to express disappointment or sadness. In these ways,
evaluative language is used to perform speech acts of expressing emotion.
These speech acts do not, however, exhaust the meaning of evaluative
language. For one thing, evaluative language is typically also used to bring
about certain effects. When a mother tells her son that that he ought to keep
his promises, she not only prescribes that her son not lie; she also standardly
intends to have an effect on his behavior—she tries to get him to keep his
promises. And when war protesters express their disapproval by calling a
war immoral, they are normally trying to get anyone listening to feel the
same way about the war. Thus, evaluative language is used to perform con-
versational acts of changing people’s behavior and feelings.
There is still more to the meaning of evaluative language. Our lives con-
sist of a constant stream of choices of varying kinds and of various levels
of importance. Because we often have to make these choices or decisions
under time pressure, it is not possible to make all of them on a case-by-case
basis. It is for this reason, among others, that we come to rely on standards
in making evaluations. In most cases, we call something “good” or “right”
because we believe that it meets or satisfies relevant standards, and we call
something “bad” or “wrong” because we believe that it violates some rele-
vant standard. This is, roughly, the content of the linguistic act of uttering
evaluative language.
On this account, calling something good or bad by itself can be fairly
empty of content. Such remarks gain content—sometimes a very rich
content—by virtue of the particular standards they invoke. This explains
why the word “good” can be applied to so many different kinds of things.
When we say that Hondas are good cars, we are probably applying stan-
dards that involve reliability, efficiency, comfort, and so on. We call someone
a good firefighter because we think the person is skilled at the tasks of a fire-
fighter, is motivated to do those tasks, works well with other firefighters,
and so on. Our standards for calling someone an ethically good person con-
cern honesty, generosity, fairness, and so on. The standards we have for call-
ing something a good car, a good firefighter, and an (ethically) good person
have little in common. Even so, the word “good” functions in the same way
in all three cases: It invokes the standards that are relevant in a given con-
text and indicates that something adequately satisfies these standards.
Because evaluative statements invoke standards, they stand in con-
trast to utterances that merely express personal feelings. If I say that I like
71
E V A L U A T I V E L A N G U A G E
a particular singer, then I am expressing a personal taste. Unless I were
being accused of lying or self-deception, it would be very odd for some-
one to reply, “No, you don’t like that singer.” On the other hand, if I call
someone a good singer (or the best singer in years), then I am going beyond
expressing my personal tastes. I am saying something that others may
accept or reject. Of course, the standards for judging singers may be impre-
cise, and they may shift from culture to culture. Still, to call someone a good
singer is to evaluate that person as a singer, which goes beyond merely
expressing feelings, because it invokes standards and indicates that the
person in question meets them.
The words “good” and “bad” are general evaluative terms. Other eval-
uative terms are more restrictive in their range of application. The word
“delicious” is usually used for evaluating the taste of foods; it means “good-
tasting.” A sin is a kind of wrong action, but, more specifically, it is an action
that is wrong according to religious standards. A bargain has a good price.
An illegal action is one that is legally wrong. Our language contains a great
many specific terms of evaluation like these. Here are a few more examples:
beautiful dangerous wasteful sneaky cute
murder prudent nosy sloppy smart
Each of these words expresses either a positive or a negative evaluation of a
quite specific kind.
Positive and negative evaluations can be subtle. Consider a word like
“clever.” It presents a positive evaluation in terms of quick mental ability. In
contrast, “cunning” often presents a negative evaluation of someone for
misusing mental abilities. It thus makes a difference which one of these
words we choose. It also makes a difference where we apply them. When
something is supposed to be profound and serious, it is insulting to call it
merely clever. Prayers, for example, should not be clever.
Sometimes seemingly innocuous words can shift evaluative force. The
word “too” is the perfect example of this. This word introduces a negative
evaluation, sometimes turning a positive quality into a negative one. Com-
pare the following sentences:
John is smart. John is too smart.
John is honest. John is too honest.
John is ambitious. John is too ambitious.
John is nice. John is too nice.
John is friendly. John is too friendly.
The word “too” indicates an excess, and thereby contains a criticism. If you
look at the items in the second column, you will see that the criticism is some-
times rather brutal—for example, calling someone “too friendly.”
The difference between an evaluative term and a descriptive term is not
always obvious. To see this, consider the terms “homicide” and “murder.”
The words are closely related but do not mean the same thing. “Homicide”
72
CHAPTER 3 ■ The Language of Argument
is a descriptive term meaning “the killing of a human being.” “Murder” is
an evaluative term meaning, in part at least, “the wrongful killing of a human
being.” It takes more to show that something is a murder than it does to
show that something is a homicide.
Just as it is easy to miss evaluative terms because we fail to recognize the
evaluative component built into their meanings, it is also possible to inter-
pret neutral words as evaluative because of positive or negative associations
that the words might evoke. The word “nuclear,” for example, has bad con-
notations for some people because of its association with bombs and wars,
but the word itself is purely descriptive. To call people nuclear scientists is
not to say that they are bad in any way. The test for an evaluative term is
this: Does the word mean that something is good or bad (right or wrong) in
a particular way?
SPIN DOCTORING
Evaluation need not be problematic, but it is often hidden and abused, most
notoriously in spin doctoring. The expression “spin doctor” seems to combine
two metaphors. The first concerns putting the right spin on things—that is,
presenting things in ways that make them look good or bad, depending on
how one wants them to be perceived. The second concerns doctoring things
Image not available due to copyright restrictions
73
E V A L U A T I V E L A N G U A G E
up to accomplish this. Spin doctoring often involves trying to find the right
way of labeling something.
A classic example comes from Shakespeare’s play, Julius Caesar (act III,
scene i). After Brutus and others killed Caesar, Brutus announced, “So are
we Caesar’s friends that have abridged his time of fearing death.” By re-
describing his act in terms of a minor benefit to Caesar, Brutus tries to make
people see his treacherous act of killing his friend as a generous act of doing
his friend a favor!
Spin doctoring is still rampant in politics today. Referring to the Iraq
war, those who favored it spoke of the “liberation” of Iraq, whereas those
who opposed it called it an “invasion” or an “occupation.” Each label in-
volves evaluation, so the disagreement over what to call it was really
about how to evaluate it. This verbal dispute is not purely verbal, because
the label can affect attitudes and behaviors by associating the Iraq war
with good military actions (liberations) or bad military actions (invasions
and occupations).
Often spin doctoring involves attributing questionable views or atti-
tudes to opponents. Supporters of President Bush sometimes refer to his de-
tractors as “the blame-America-first crowd” and describe their policies for
getting out of Iraq as “cut and run.” These negative labels are applied even
to policies, such as partial staged redeployment, that are far from cowardly
in the way suggested by “cut and run.” Similarly, in 2006, conservative com-
mentator David Brooks described opposition to Joe Lieberman’s campaign
as a “liberal inquisition.” Stephen Colbert lampooned this label in a segment
of “The Word” where he joked that, if incumbents like Lieberman continued
to be attacked so harshly, then they would actually be forced to defend their
records. Colbert concluded, “There is only one word for that—inquisition.”
But the sidebar read “democracy.” Sometimes parody is the most effective
response to spin doctoring.
Of course, liberal democrats are often just as guilty of spin doctoring.
When discussing a measure that would repeal a large number of envi-
ronmental regulations, President Clinton sarcastically referred to it as the
“Polluter’s Bill of Rights“—not exactly a generous way of describing a bill
based on the belief that environmental regulations had gone too far. And, to
quote Stephen Colbert again, “Affirmative action is a prime example of
the Leftist campaign to make ideas seem less dangerous than they are,
through the strategic use of positive words. Think about it. How can
something be bad if it is ‘affirmative’? and how can we ignore it if is it
‘action’?” [from I Am America (and So Can You!), New York: Grand Central
Publishing, 2007, p. 174].
Spin doctoring like this makes it harder to discuss the real benefits
and costs of such laws and policies, because nobody wants to argue against
liberation or affirmation or in favor of invasion, inquisition, or pollution.
The task of critical analysis is to see through such slogans to the important
issues that lie behind them, so that we can intelligently address the real val-
ues at stake.
74
CHAPTER 3 ■ The Language of Argument
For each of the following sentences, construct two others—one that reverses
the evaluative force, and one that is as neutral as possible. The symbol “0”
stands for neutral, “+” for positive evaluative force, and “–” for negative eval-
uative force. Try to make as little change as possible in the descriptive content
of the sentence.
Example: – Professor Conrad is rude.
+ Professor Conrad is uncompromisingly honest in his criticisms.
0 Professor Conrad often upsets people with his criticisms.
1. – Larry is a lazy lout. 6. – Walter is a weenie.
2. + Brenda is brave. 7. + Carol is caring.
3. – Sally is a snob. 8. – Bill is bossy.
4. + Bartlett is a blast. 9. – Oprah is opinionated.
5. – George is a goody-goody. 10. – This is a Mickey Mouse
exercise.
Exercise X
Be a spin doctor yourself by writing upbeat, good-sounding titles or descrip-
tions for the following proposals. Remember that, as a professional spin doc-
tor, you should be able to make things you personally hate sound good.
1. Imposing a $1,000 fee on graduating seniors
2. Requiring all students to participate in a twenty-one-meal-per-week food
plan
Exercise XI
Indicate whether the following italicized terms are positively evaluative (E+),
negatively evaluative (E–), or simply descriptive (D). Remember, the evalua-
tions need not be moral evaluations.
1. Janet is an excellent golfer.
2. The group was playing very loudly.
3. The group was playing too loudly.
4. William was rude to his parents.
5. William shouted at his parents.
6. They mistakenly turned right at
the intersection.
7. Fascists ruled Italy for almost
twenty years.
8. That’s a no-no.
Exercise IX
9. Bummer.
10. Debbie lied.
11. Debbie said something false.
12. Joe copped out.
13. Jake is a bully.
14. Mary Lou was a gold medalist.
15. She is sick.
16. He suffers from a hormonal
imbalance.
75
E V A L U A T I V E L A N G U A G E
3. Abolishing coed dormitories
4. Abolishing fraternities
5. Requiring women students to return to their dormitories by midnight
(such rules were once quite common)
6. Abolishing failing grades
7. Restoring failing grades
8. Requiring four years of physical education
9. Abolishing intercollegiate football
10. Introducing a core curriculum in Western civilization
11. Abolishing such a curriculum
12. Abolishing faculty tenure
1. What precisely does “That’s too bad” mean?
2. In the Democratic presidential candidate debate on September 26, 2007,
Representative Dennis Kucinich was asked the following question by
MSNBC correspondent Tim Russert. Was Kucinich’s answer spin
doctoring? Was it legitimate or illegitimate? Why?
Russert: . . . Congressman Kucinich, when you were mayor of Cleveland,
you let Cleveland go into bankruptcy, the first time that happened since
the Depression. The voters of Cleveland rewarded you by throwing you
out of office and electing a Republican mayor of Cleveland. How can you
claim that you have the ability to manage the United States of America,
when you let Cleveland go bankrupt?
Rep. Kucinich: You know, Tim, that was NBC’s story. Now I want the peo-
ple to know what the real story was. I took a stand on behalf of the peo-
ple of Cleveland to save a municipal electric system. The banks and the
utilities in Cleveland, the private utilities, were trying to force me to sell
that system. And so on December 15th, 1978, I told the head of the
biggest bank, when he told me I had to sell the system in order to get the
city’s credit renewed, that I wasn’t going to do it. . . . I put my job on the
line. How many people would be willing to put their job on the line in
the face of pressure from banks and utilities? As this story gets told, peo-
ple will want me to be their next president because they’ll see in me not
only the ability to take a stand, but the ability to live with integrity.
Discussion Questions
This page intentionally left blank
THE ART OF CLOSE ANALYSIS
This chapter will largely be dedicated to a single purpose: the close and careful analy-
sis of a speech drawn from the Congressional Record, using argumentative devices
introduced in Chapter 3. The point of this chapter is to show in detail how these meth-
ods of analysis can be applied to an actual argument of some richness and complexity.
AN EXTENDED EXAMPLE
It is now time to apply all of our previously discussed notions to a genuine
argument. Our example will be a debate that occurred in the House of Rep-
resentatives on the question of whether there should be an increase in the al-
lowance given to members of the House for clerical help—the so-called clerk
hire allowance. The argument against the increase presented by Representa-
tive Kyl (Republican, Iowa) will be examined in detail. We will put it under
an analytic microscope.
The choice of this example may seem odd, for the question of clerk hire
allowance is not one of the burning issues of our time. This, in fact, is one
reason for choosing it. It will be useful to begin with an example about
which feelings do not run high to learn the habit of objective analysis. Later
on we shall examine arguments about which almost everyone has strong
feelings and try to maintain an objective standpoint even there.
The example is good for two other reasons: 1. It contains most of the ar-
gumentative devices we have listed, and 2. relatively speaking, it is quite a
strong argument. This last remark may seem ironic after we seemingly tear
the argument to shreds. However, in comparison to other arguments we
shall examine, it stands up well.
We begin by reading through a section of the Congressional Record (vol.
107, part 3, March 15, 1961, pp. 4059–60) without comment:
CLERK HIRE ALLOWANCE,
HOUSE OF REPRESENTATIVES
Mr. FRIEDEL. Mr. Speaker, by direction of the Committee on House Admin-
istration, I call up the resolution (H. Res. 219) to increase the basic clerk hire
allowance of each Member of the House, and for other purposes, and ask for
its immediate consideration.
4
77
78
CHAPTER 4 ■ The Art of Close Analys is
The Clerk read the resolution as follows:
Resolved, That effective April 1, 1961, there shall be paid out of the contingent
fund of the House, until otherwise provided by law, such sums as may be neces-
sary to increase the basic clerk hire allowance of each Member and the Resident
Commissioner from Puerto Rico by an additional $3,000 per annum, and each
such Member and Resident Commissioner shall be entitled to one clerk in addi-
tion to those to which he is otherwise entitled by law.
Mr. FRIEDEL. Mr. Speaker, this resolution allows an additional $3,000
per annum for clerk hire and an additional clerk for each Member of the
House and the Resident Commissioner from Puerto Rico. Our subcommit-
tee heard the testimony, and we were convinced of the need for this provi-
sion to be made. A few Members are paying out of their own pockets for
additional clerk hire. This $3,000 is the minimum amount we felt was nec-
essary to help Members pay the expenses of running their offices. Of
course, we know that the mail is not as heavy in some of the districts as it is
in others, and, of course, if the Member does not use the money, it remains
in the contingent fund.
Mr. KYL. Mr. Speaker, will the gentleman yield?
Mr. FRIEDEL. I yield to the gentleman from Iowa [Mr. Kyl] for a statement.
Mr. KYL. Mr. Speaker, I oppose this measure. I oppose it first because it is
expensive. I further oppose it because it is untimely.
I do not intend to belabor this first contention. We have been presented a
budget of about $82 billion. We have had recommended to us a whole series
of additional programs or extensions of programs for priming the pump, for
depressed areas, for the needy, for the unemployed, for river pollution proj-
ects, and recreation projects, aid to education, and many more. All are listed
as “must” activities. These extensions are not within the budget. Furthermore,
if business conditions are as deplorable as the newspapers indicate, the
Government’s income will not be as high as anticipated. It is not enough to
say we are spending so much now, a little more will not hurt. What we spend,
we will either have to recover in taxes, or add to the staggering national debt.
The amount of increase does not appear large. I trust, however, there is no
one among us who would suggest that the addition of a clerk would not entail
allowances for another desk, another typewriter, more materials, and it is not
beyond the realm of possibility that the next step would then be a request for
additional office space, and ultimately new buildings. Some will say, “All the
Members will not use their maximum, so the cost will not be great.” And this
is true. If the exceptions are sufficient in number to constitute a valid argument,
then there is no broad general need for this measure. Furthermore, some Mem-
bers will use these additional funds to raise salaries. Competition will force all
salaries upward in all offices and then on committee staffs, and so on. We may
even find ourselves in a position of paying more money for fewer clerks and in
a tighter bind on per person workload.
This measure proposes to increase the allowance from $17,500 base clerical
allowance to $20,500 base salary allowance. No member of this House can tell
79
An Extended Example
us what this means in gross salary. That computation is almost impossible.
Such a completely absurd system has developed through the years on salary
computations for clerical hire that we have under discussion a mathematical
monstrosity. We are usually told that the gross allowed is approximately
$35,000. This is inaccurate. In one office the total might be less than $35,000 and
in another, in complete compliance with the law and without any conscious
padding, the amount may be in excess of $42,000. This is possible because of a
weird set of formulae which determines that three clerks at $5,000 cost less
than five clerks at $3,000. Five times three might total the same as three times
five everywhere else in the world—but not in figuring clerk hire in the House.
This is an application of an absurdity. It is a violation of bookkeeping
principles, accounting principles, business principles and a violation of com-
mon sense. Listen to the formula:
First, 20 percent increase of first $1,200; 10 percent additional from $1,200
to $4,600; 5 percent further additional from $4,600 to $7,000.
Second, after applying the increases provided in paragraph 1, add an ad-
ditional 14 percent or a flat $250, whichever is the greater, but this increase
must not exceed 25 percent.
Third, after applying the increases provided in both paragraphs 1 and 2,
add an additional increase of 10 percent in lieu of overtime.
Fourth, after applying the increases provided in paragraphs 1, 2, and 3,
add an additional increase of $330.
Fifth, after applying the increases provided in paragraphs 1, 2, 3, and 4,
add an additional increase of 5 percent.
Sixth, after applying the increases provided in paragraphs 1, 2, 3, 4, and
5, add an additional increase of 10 percent but not more than $800 nor less
than $300 a year.
Seventh, after applying the increases provided in paragraphs, 1, 2, 3, 4, 5,
and 6, add an additional increase of 71⁄2 percent.
Eighth, after applying the increases provided in paragraphs, 1, 2, 3, 4, 5,
6, and 7, add an additional increase of 10 percent.
Ninth, after applying the increases provided in paragraphs 1, 2, 3, 4, 5, 6,
7, and 8, add an additional increase of 71⁄2 percent.
The Disbursing Office has a set of tables to figure House salaries for office
staffs and for about 900 other employees. It contains 45 sheets with 40
entries per sheet. In the Senate, at least, they have simplified the process
some by figuring their base in multiples of 60, thus eliminating 11 cate-
gories. Committee staffers, incidentally, have an $8,880 base in comparison
to the House $7,000 base limitation.
Now, Mr. Speaker, I have planned to introduce an amendment or a substi-
tute which would grant additional clerk hire where there is a demonstrable
need based on heavier than average population or “election at large” and
possible other factors. But after becoming involved in this mathematical maze,
I realize the folly of proceeding one step until we have corrected this situation.
We can offer all kinds of excuses for avoiding a solution. We cannot offer rea-
sonable arguments that it should not be done or that it cannot be done.
80
CHAPTER 4 ■ The Art of Close Analys is
Someone has suggested that the Members of this great body prefer to
keep the present program because someone back in the home district might
object to the gross figures. I know this is not so. When a Representative is
busy on minimum wage, or aid to education, or civil rights, such matters of
housekeeping seem too picayune to merit attention. The Member simply
checks the table and hires what he can hire under the provisions and then
forgets the whole business. But I know the Members also want the people
back home to realize that what we do here is open and frank and accurate,
and that we set an example in businesslike procedures. The more we can
demonstrate responsibility the greater will be the faith in Congress.
May I summarize. It is obvious that some Members need more clerical help
because of large population and large land area. I have been working for some
time with the best help we can get, on a measure which would take these items
into consideration. Those Members who are really in need of assistance should
realize that this temporary, hastily conceived proposition we debate today will
probably obviate their getting a satisfactory total solution.
First, we should await redistricting of the Nation.
Second, we should consider appropriate allowance for oversize districts
considering both population and total geographic area.
Finally, I hope we can develop a sound and sensible formula for computing
salaries of office clerks and other statutory employees in the same category.
Before going any further, it will be useful to record your general reactions
to this speech. Perhaps you think that on the whole Kyl gives a well-reasoned
argument on behalf of his position. Alternatively, you might think that he is
making a big fuss over nothing, trying to confuse people with numbers, and
just generally being obnoxious. When you are finished examining this argu-
ment in detail, you can look back and ask yourself why you formed this orig-
inal impression and how, if at all, you have changed your mind.
The first step in the close analysis of an argument is to go through the
text, labeling the various argumentative devices we have examined. Here
some abbreviations will be useful:
argument marker M
assuring term A
guarding term G
discounting term D
argumentative performative AP
evaluative term E (+ or –)
rhetorical device R
The last label is a catchall for the various rhetorical devices discussed in
Chapter 1, such as overstatement, understatement, irony or sarcasm,
metaphor, simile, rhetorical questions, and so on.
If you want to make your analysis extra close, it is illuminating to specify
which rhetorical device is deployed whenever you mark something with “R.”
81
An Extended Example
It is also useful to specify whether each argument marker marks a reason or a
conclusion (and what the argument is), which stronger term is replaced by
each guarding term marked “G,” and which objection is discounted whenever
you mark a discounting term with “D.”
This simple process of labeling brings out features of an argument that
could pass by unnoticed. It also directs us to ask sharp critical questions. To
see this, we can look at each part of the argument in detail.
Mr. KYL. Mr. Speaker, I oppose this measure. I oppose it
first because it is expensive. I further oppose it because it is
untimely.
This is a model of clarity. By the use of a performative utterance in the open-
ing sentence, Kyl makes it clear that he opposes the measure. Then by twice
using the argument marker “because,” he gives his two main reasons for
opposing it: It is expensive and it is untimely. We must now see if he makes
good on each of these claims.
The next paragraph begins the argument for the claim that the measure is
expensive:
I do not intend to belabor this first contention. We have been
presented a budget of about $82 billion. We have had recom-
mended to us a whole series of additional programs or ex-
tensions of programs for priming the pump, for depressed
areas, for the needy, for the unemployed, for river pollution
projects, and recreation projects, aid to education, and many
more. All are listed as “ must ” activities. These extensions are
not within the budget. Furthermore , if business conditions
are as deplorable as the newspapers indicate, the Govern-
ment’s income will not be as high as anticipated. It is not
enough to say we are spending so much now, a little more
will not hurt. What we spend, we will either have to recover
in taxes, or add to the staggering national debt.
a. “I do not intend to belabor this first contention.” This is an example of
assuring. The conversational implication is that the point is so obvious that
little has to be said in its support. Yet there is something strange going on
here. Having said that he will not belabor the claim that the bill is expensive,
Kyl actually goes on to say quite a bit on the subject. It is a good idea to look
closely when someone says that he or she is not going to do something, for
often just the opposite is happening. For example, saying “I am not suggest-
ing that Smith is dishonest” is one way of suggesting that Smith is dishon-
est. If no such suggestion is being made, why raise the issue at all?
b. Kyl now proceeds in a rather flat way, stating that the proposed budget
comes to $82 billion and that it contains many new programs and extensions
AP
A
M
R
D
AP
M
M
82
CHAPTER 4 ■ The Art of Close Analys is
of former programs. Because these are matters of public record and nobody
is likely to deny them, there is no need for guarding or assuring. Kyl also
claims, without qualification, that these extensions are not within the
budget. This recital of facts does, however, carry an important conversa-
tional implication: Since the budget is already out of balance, any further ex-
tensions should be viewed with suspicion.
c. Putting the word “must” in quotation marks, or saying it in a sarcastic
tone of voice, is a common device for denying something. The plain sugges-
tion is that some of these measures are not must activities at all. Kyl here
suggests that some of the items already in the budget are not necessary. He
does this, of course, without defending this suggestion.
d. “Furthermore, if business conditions are as deplorable as the newspa-
pers indicate, the Government’s income will not be as high as anticipated.”
The word “furthermore” suggests that an argument is about to come. How-
ever, the following sentence as a whole is an indicative conditional (with the
word “then” dropped out). As such, the sentence does not produce an argu-
ment, but instead provides only a pattern for an argument.
To get an argument from this pattern, one would have to assert the an-
tecedent of the conditional. The argument would then come to this:
(1) If business conditions are as deplorable as the newspapers indicate,
then the Government’s income will not be as high as anticipated.
(2) Business conditions are as deplorable as the newspapers indicate.
(3) The Government’s income will not be as high as anticipated.
The first premise seems perfectly reasonable, so, if Kyl could establish the
second premise, then he would have moved the argument along in an im-
portant way. Yet he never explicitly states that business conditions are so
deplorable. All he says is that “the newspapers indicate” this. Moreover, this
appeal to authority (see Chapter 15) does not mention any specific newspa-
per, so he does not endorse any specific authority. Still, Kyl never questions
what the newspapers claim, and it would be misleading to bring up these
newspaper reports without questioning them if he thought they were way
off the mark. So Kyl does seem to have in mind something like the argu-
ment (1)–(3).
e. “It is not enough to say we are spending so much now, a little more will
not hurt.” The opening phrase is, of course, used to deny what follows it.
Kyl is plainly rejecting the argument that, since we are spending so much
now, a little more will not hurt. Yet his argument has a peculiar twist, for
who would come right out and make such an argument? If you stop to think
for a minute, it should be clear that nobody would want to put it that way.
An opponent, for example, would use quite different phrasing. He might
say something like this: “Considering the large benefits that will flow from
this measure, it is more than worth the small costs.” What Kyl has done is
attribute a bad argument to his opponents and then reject it in an indignant
tone. This is a common device, and when it is used, it is often useful to ask
83
An Extended Example
whether anyone would actually argue or speak in the way suggested. When
the answer to this question is no, as it often is, we have what was called “the
trick of discounting straw men” in Chapter 3 (see also Chapter 17). In such
cases, it is useful to ask what the speaker’s opponent would have said in-
stead. This leads to a further question: Has the arguer even addressed him-
self to the real arguments of his opponents?
So far, Kyl has not addressed himself to the first main point of his argu-
ment: that the measure is expensive. This is not a criticism, because he is re-
ally making the preliminary point that the matter of expense is significant.
Here he has stated some incontestable facts—for example, that the budget is
already out of balance. Beyond this he has indicated, with varying degrees
of strength, that the financial situation is grave. It is against this background
that the detailed argument concerning the cost of the measure is actually
presented in the next paragraph.
The amount of increase does not appear large. I trust,
however, there is no one among us who would suggest that
the addition of a clerk would not entail allowances for
another desk, another typewriter, more materials, and it is
not beyond the realm of possibility that the next step would
then be a request for additional office space, and ultimately
new buildings. Some will say , “All the Members will not
use their maximum, so the cost will not be great.” And this is
true. If the exceptions are sufficient in number to constitute a
valid argument, then there is no broad general need for this
measure. Furthermore, some Members will use these addi-
tional funds to raise salaries. Competition will force all
salaries upward in all offices and then on committee staffs,
and so on. We may even find ourselves in a position of pay-
ing more money for fewer clerks and in a tighter bind on per
person workload.
a. “The amount of increase does not appear large.” Words like “appear”
and “seem” are sometimes used for guarding, but we must be careful not to
apply labels in an unthinking way. The above sentence is the beginning of a
discounting argument. As soon as you hear this sentence, you can feel that a
word like “but” or “however” is about to appear. Sure enough, it does.
b. “I trust, however, there is no one among us who would suggest that the
addition of a clerk would not entail allowances for another desk, another
typewriter, more materials. . . .” This is the beginning of Kyl’s argument that
is intended to rebut the argument that the increase in expenses will not be
large. Appearances to the contrary, he is saying, the increase will be large. He
then ticks off some additional expenses that are entailed by hiring new
clerks. Notice that the whole sentence is covered by the assuring phrase
D A
D
M
G
G
G
84
CHAPTER 4 ■ The Art of Close Analys is
“I trust . . . there is no one among us who would suggest. . . .” This implies
that anyone who would make such a suggestion is merely stupid. But the
trouble with Kyl’s argument so far is this: He has pointed out genuine addi-
tional expenses, but they are not, after all, very large. It is important for him
to get some genuinely large sums of money into his argument. This is the
point of his next remark.
c. “. . . and it is not beyond the realm of possibility that the next step
would then be a request for additional office space, and ultimately new
buildings.” Here, at last, we have some genuinely large sums of money in
the picture, but the difficulty is that the entire claim is totally guarded by the
phrase “it is not beyond the realm of possibility.” There are very few things
that are beyond the realm of possibility. Kyl’s problem, then, is this: There
are certain additional expenses that he can point to without qualification,
but these tend to be small. On the other hand, when he points out genuinely
large expenses, he can only do so in a guarded way. So we are still waiting
for a proof that the expense will be large. (Parenthetically, it should be
pointed out that Kyl’s prediction of new buildings actually came true.)
d. “Some will say, ‘All the Members will not use their maximum, so the
cost will not be great.’ And this is true. If the exceptions are sufficient in
number to constitute a valid argument, then there is no broad general need
for this measure.” This looks like a “trick” argument, and for this reason
alone it demands close attention. The phrase “some will say” is a standard
way of beginning a discounting argument. This is, in fact, a discounting ar-
gument, but its form is rather subtle. Kyl cites what some will say, and then
he adds, somewhat surprisingly: “And this is true.” To understand what is
going on here, we must have a good feel for conversational implication. Kyl
imagines someone reasoning in the following way:
All the Members will not use their maximum.
So, the cost will not be great.
Therefore, we should adopt the measure.
Given the same starting point, Kyl tries to derive just the opposite conclusion
along the following lines:
All the Members will not use their maximum.
If the exceptions are not sufficient, then the cost will be too great.
But if the exceptions are sufficient, there is no broad general need for this
measure.
Therefore, whether it is expensive or not, we should reject this measure.
In order to get clear about this argument, we can put it into schematic form:
Kyl’s argument:
If (1) the measure is expensive, then reject it.
If (2) the measure is inexpensive, then, because that shows there is no
general need, reject it.
85
An Extended Example
The opposite argument:
If (1) the measure is inexpensive, then accept it.
If (2) the measure is expensive, then, because that demonstrates a general
need, accept it.
When the arguments are spread out in this fashion, it should be clear that
they have equal strength. Both are no good. The question that must be set-
tled is this: Does a genuine need exist that can be met in an economically
sound manner? If there is no need for the measure, then it should be re-
jected, however inexpensive. Again, if there is a need, then some expense is
worth paying. The real problem is to balance need against expense and then
decide on this basis whether the measure as a whole is worth adopting.
Kyl’s argument is a sophistry, because it has no tendency to answer the
real question at hand. A sophistry is a clever but fallacious argument intended
to establish a point through trickery. Incidentally, it is one of the marks of a
sophistical argument that, though it may baffle, it almost never convinces.
We think that few readers will have found this argument persuasive even if
they cannot say exactly what is wrong with it. The appearance of a sophisti-
cal argument (or even a complex and tangled argument) is a sign that the
argument is weak. Remember, when a case is strong, people usually argue in
a straightforward way.
e. “Furthermore, some Members will use these additional funds to raise
salaries. Competition will force all salaries upward in all offices and then
on committee staffs, and so on.” The word “furthermore” signals that fur-
ther reasons are forthcoming. Here Kyl returns to the argument that the
measure is more expensive than it might appear at first sight. Although
Kyl’s first sentence is guarded by the term “some,” he quickly drops his
guard and speaks in an unqualified way about all salaries in all offices. Yet
the critic is bound to ask whether Kyl has any right to make these projec-
tions. Beyond this, Kyl here projects a parade of horrors. (See Chapter 13.) He
pictures this measure leading by gradual steps to quite disastrous conse-
quences. Here the little phrase “and so on” carries a great burden in the
argument. Once more, we must simply ask ourselves whether these projec-
tions seem reasonable.
f. “We may even find ourselves in a position of paying more money for
fewer clerks and in a tighter bind on per person workload.” Once more, the
use of a strong guarding expression takes back most of the force of the argu-
ment. Notice that if Kyl could have said straight out that the measure will
put us in a position of paying more money for fewer clerks and in a tighter
bind on per-person workload, that would have counted as a very strong ob-
jection. You can hardly do better in criticizing a position than showing that
it will have just the opposite result from what is intended. In fact, however,
Kyl has not established this; he has only said that this is something that we
“may even find.”
Before we turn to the second half of Kyl’s argument, which we shall see
in a moment is much stronger, we should point out that our analysis has not
86
CHAPTER 4 ■ The Art of Close Analys is
been entirely fair. Speaking before the House of Representatives, Kyl is in an
adversarial situation. He is not trying to prove things for all time; rather, he is
responding to a position held by others. Part of what he is doing is raising
objections, and a sensitive evaluation of the argument demands a detailed
understanding of the nuances of the debate. But even granting this, it should
be remembered that objections themselves must be made for good reasons.
The problem so far in Kyl’s argument is that the major reasons behind his
objections have constantly been guarded in a very strong way.
Turning now to the second part of Kyl’s argument—that the measure is
untimely—we see that he moves along in a clear and direct way with little
guarding.
This measure proposes to increase the allowance from
$17,500 base clerical allowance to $20,500 base salary
allowance. No member of this House can tell us what this
means in gross salary. That computation is almost impossi-
ble. Such a completely absurd system has developed
through the years on salary computations for clerical hire
that we have under discussion a mathematical monstrosity.
We are usually told that the gross allowed is approximately
$35,000. This is inaccurate. In one office the total might be
less than $35,000 and in another, in complete compliance
with the law and without any conscious padding, the
amount may be in excess of $42,000. This is possible because
of a weird set of formulae which determines that three
clerks at $5,000 cost less than five clerks at $3,000. Five times
three might total the same as three times five everywhere
else in the world—but not in figuring clerk hire in the House.
This is an application of an absurdity. It is a violation of
bookkeeping principles, accounting principles, business prin-
ciples and a violation of common sense. Listen to the formula.
The main point of the argument is clear enough: Kyl is saying that the present
system of clerk salary allowance is utterly confusing, and this matter should
be straightened out before any other measures in this area are adopted. There
is a great deal of negative evaluation in this passage. Notice the words and
phrases that Kyl uses:
a completely absurd system
weird set of formulae
violation of common sense
mathematical monstrosity
an absurdity
G
E–
E–
E–
E–
E–
R
87
An Extended Example
There is also a dash of irony in the remark that five times three might total
the same as three times five everywhere else in the world, but not in figur-
ing clerk hire in the House. Remember, there is nothing wrong with using
negative evaluative and expressive terms if they are deserved. Looking at
the nine-step formula in Kyl’s speech, you can decide for yourself whether
he is on strong grounds in using this negative language.
Now, Mr. Speaker, I have planned to introduce an amend-
ment or a substitute which would grant additional clerk hire
where there is a demonstrable need based on heavier than
average population or “election at large” and possible other
factors.
a. This passage discounts any suggestion that Kyl is unaware that a gen-
uine problem does exist in some districts. It also indicates that he is willing
to do something about it.
b. The phrase “and possible other factors” is not very important, but it
seems to be included to anticipate other reasons for clerk hire that should at
least be considered.
But after becoming involved in this mathematical maze, I
realize the folly of proceeding one step until we have cor-
rected this situation.
a. Here Kyl clearly states his reason for saying that the measure is un-
timely. Notice that the reason offered has been well documented and is not
hedged in by qualifications.
b. The phrases “mathematical maze” and “folly” are again negatively
evaluative.
We can offer all kinds of excuses for avoiding a solution.
We cannot offer reasonable arguments that it should not be
done or that it cannot be done.
Notice that the first sentence ridicules the opponents’ arguments by calling
them excuses, a term with negative connotations. The second sentence gives
assurances that such a solution can be found.
Someone has suggested that the Members of this great body
prefer to keep the present program because someone back in
the home district might object to the gross figures. I know
this is not so. When a Representative is busy on minimum
wage, or aid to education, or civil rights, such matters of
housekeeping seem too picayune to merit attention. The
Member simply checks the table and hires what he can hire
under the provisions and then forgets the whole business.
D
E–A
E–
E–
A
A
88
CHAPTER 4 ■ The Art of Close Analys is
But I know the Members also want the people back home
to realize that what we do here is open and frank and accu-
rate, and that we set an example in businesslike procedures.
The more we can demonstrate responsibility the greater will
be the faith in Congress.
a. Once more the seas of rhetoric run high. Someone (though not Kyl him-
self) has suggested that the members of the House wish to conceal informa-
tion. He disavows the very thought that he would make such a suggestion
by the sentence “I know this is not so.” All the same, he has gotten this sug-
gestion into the argument.
b. Kyl then suggests another reason why the members of the House
will not be concerned with this measure: It is “too picayune.” The last two
sentences rebut the suggestion that it is too small to merit close attention.
Even on small matters, the more the House is “open and frank and accu-
rate,” the more it will “set an example in businesslike procedures” and
thus “demonstrate responsibility” that will increase “the faith in Con-
gress.” This is actually an important part of Kyl’s argument, for presum-
ably his main problem is to get the other members of the House to take
the matter seriously.
May I summarize. It is obvious that some Members need
more clerical help because of large population and large
land area. I have been working for some time with the best
help we can get, on a measure which would take these items
into consideration. Those Members who are really in need of
assistance should realize that this temporary, hastily con-
ceived proposition we debate today will probably obviate
their getting a satisfactory total solution.
a. This is a concise summary. Kyl once more assures the House that he is
aware that a genuine problem exists. He also indicates that he is working on it.
b. The phrase “temporary, hastily conceived proposition we debate to-
day” refers back to his arguments concerning untimeliness.
c. The claim that “it will probably obviate their getting a satisfactory total
solution” refers back to the economic argument. Notice, however, that, as
before, the economic claim is guarded by the word “probably.”
First, we should await redistricting of the Nation.
Second, we should consider appropriate allowance for
oversize districts considering both population and total
geographic area.
Finally, I hope we can develop a sound and sensible for-
mula for computing salaries of office clerks and other statu-
tory employees in the same category.
D
E+
A
E–
G
A
A
M
A
E+
E+
E+
89
An Extended Example
This is straightforward except that a new factor is introduced: We should
await redistricting of the nation. This was not mentioned earlier in the argu-
ment, and so seems a bit out of place in a summary. Perhaps the point is so
obvious that it did not need any argument to support it. On the other hand,
it is often useful to keep track of things that are smuggled into the argument
at the very end. If redistricting was about to occur in the near future, this
would give a strong reason for delaying action on the measure. Because the
point is potentially so strong, we might wonder why Kyl has made so little
of it. Here, perhaps, we are getting too subtle.
Now that we have looked at Representative Kyl’s argument in close de-
tail, we can step back and notice some important features of the argument
as a whole. In particular, it is usually illuminating to notice an argument’s
purpose, audience, and standpoint.
First, Kyl’s overall purpose is clear. As his opening sentence indicates,
he is presenting an argument intended to justify his opposition to an in-
crease in the clerk hire allowance. Virtually everything he says is directed
toward this single goal. In other cases, arguers pursue multiple goals, and
sorting things out can be a complex matter. Sometimes it is hard to tell
what an argument is even intended to establish. This is usually a sign that
the person presenting the argument is confused or, perhaps, trying to con-
fuse his audience.
Second, Kyl’s argument is addressed to a specific audience. He is not
speaking to an enemy of the United States who would love to see our gov-
ernment waste its money. Nor is he speaking to clerks or to those U.S.
citizens who might be hired as clerks if the clerk hire allowance were
raised. He is presenting his argument to other representatives in Con-
gress. He is trying to show this group that they and he have reasons to
oppose this increase in the clerk hire allowance. His task, then, is to pres-
ent reasons that they accept—or should accept—for rejecting an increase
in the clerk hire allowance.
Third, Kyl not only addresses his argument to a particular audience, he
also adopts a particular standpoint to it. Good arguments are usually pre-
sented not only to specific audiences but also from particular standpoints.
Kyl’s standpoint is clear and powerful. He puts himself across as a tough-
minded, thoroughly honest person who is willing to stand up against
majority opinion. This, in fact, may be an accurate representation of his
character, but by adopting this standpoint he gains an important argumen-
tative advantage: He suggests that those who disagree with him are a bit
soft-minded, not altogether candid, and, anyway, mere tools of the Demo-
cratic majority that runs the Congress. By adopting this stance, Kyl casts
his opponents in a light that is hardly flattering.
By specifying the purpose, audience, and standpoint of an argument,
we get a clearer sense of what the argument needs to accomplish in order
to succeed in its goals. By looking closely at special words in the argument,
as well as at what is conversationally implied, we get a better idea of how
the argument is supposed to achieve its goals. All of this together helps us
understand the argument. It will sometimes remain unclear how well the
90
CHAPTER 4 ■ The Art of Close Analys is
argument succeeds. It will always require care and skill to apply these
methods. Still, the more you practice, the more you will be able to under-
stand arguments.
Read the following passage. Then, for each of the numbered expressions,
either answer the corresponding question or label the main argumentative
move, if any, using these abbreviations:
M = argument marker
A = assuring term
G = guarding term
D = discounting term
E = negative evaluative term
E+ = positive evaluative term
R = rhetorical device
N = none of the above
This letter to the editor appeared in The Dartmouth on September 23, 1992,
although references to the author’s college have been removed. The author
was president of the student assembly and a member of a single-sex frater-
nity at the time.
GREEKS SHOULD BE CO-ED
by Andrew Beebe
For some time now, people have been asking the question “Why should the
Greek [fraternity and sorority] system go co-ed?” To them, I pose an answer [1]
in a question, “Why not?“ [2]
Learning in college extends beyond the classrooms, onto the athletic fields, into the
art studios, and into our social environs. [3] In fact [4], some [5] say that most [6] of
what we learn at college comes from interaction with people and ideas during
time spent outside of the lecture halls. The concept of segregating students in
their social and residential environments by gender directly contradicts the
ideals [7] of a college experience. This is exactly [8] what the fraternity and
sorority system does.
With all the benefits [9] of a small, closely-bonded group, the poten-
tial for strong social education would seem obvious [10]. But [11] is it fair [12]
for us to remove the other half of our community from that education? [13]
In many colleges, this voluntary segregation exists in fraternities and
sororities.
Exercise I
91
An Extended Example
From the planning of a party or involvement in student activities to the
sharing of living and recreational space, the fraternity and sorority system is a
social environment ripe [14] with educational potential [15]. The idea that
women and men would receive as complete an experience from these environ-
ments while virtually [16] separated is implausible [17].
But [18] what do women and men learn from one another that they don’t already
know? [19] Problems in gender relations between all ages prove [20] that our so-
ciety is plagued by gender-based prejudice [21]. Since [22] prejudice is the ignorance
of one group by another [23], it will best be addressed by education. The ques-
tion then [24] becomes: Which way is best to educate one another?
Sexism, homophobia, date rape, eating disorders, and other social problems
[25] are often [26] connected to gender-relation issues. As campus experience
shows [27], we have a long way to go in combating these problems. Defenders
of fraternities and sororities may [28] argue that they do not, solely by nature
of being single sex, promote sexism or other prejudices. But [29], if we can rec-
ognize that these problems exist in our society, it is not important to find the
blame, but [30] rather to offer a solution. It is clear [31] that separating people
by gender is not the right [32] way to promote better [33] understanding be-
tween the sexes. To the contrary, bringing different people together is the only
way prejudice, no matter what the cause (or result) may be [34], can be overcome.
Acknowledging that breaking down walls of separation may [35] help foster
better understanding, it is important to look at what might [36] change for the
worse. There would be some [37] obvious [38] logistical changes in rush, pledg-
ing, relationships with national organization, and house leadership. But [39]
where are the real consequences? [40] Men could [41] still cultivate strong bonds
with other men. Women could [42] still bond with other women. The difference
is that there would be a well-defined [43] environment where men and women
could [44] create strong, lasting bonds and friendships between one another.
There are many more benefits [45] to a co-ed system than there are sacri-
fices [46]. Men and women could share the responsibilities of running what
is now a predominantly [47] male-controlled social structure. First-year men
and women could interact with older students in a social environment be-
yond the classroom or the dining halls. People in a co-ed system could find a
strong support group that extends beyond their own sex. With these advan-
tages [48] and more, it is clear [49] that the all-co-ed system offers everything
found in a single-sex organization and more. Although [50] there are some [51]
minor sacrifices to be made, they are insignificant in comparison to the gain
[52] for all.
College is the last place we want to isolate ourselves. The entire idea of the
“holistic education” is based on [53] expanding our knowledge, not separating
ourselves from one another. Our fraternity and sorority system includes many
[54] different types of students. So [55] why should some houses refuse women
simply because [56] they are women? Why do some houses refuse men solely
because [57] they are men? The only solution is desegregation of the fraternity
and sorority system. After all [58], when it comes to challenging one another to
learn, what are we afraid of? [59]
(continued)
92
CHAPTER 4 ■ The Art of Close Analys is
QUESTIONS:
[1]: Is this sentence an explicit performative?
[2]: Explain the difference between asking “Why?” and asking “Why
not?” in this context.
[3]: Why does the author begin with this point?
[4]–[12]: Write labels.
[13]: What is the expected answer to this rhetorical question?
[14]: What kind of rhetorical device is this? What is its point?
[15]–[18]: Write labels.
[19]: Who is supposed to be asking this question?
[20]–[22]: Write labels.
[23]: What is the point of this definition?
[24]–[33]: Write labels.
[34]: Why does this author add this dependent clause?
[35]–[39]: Write labels.
[40]: What does this question imply in this context?
[41]–[58]: Write labels.
[59]: What is the expected answer to this rhetorical question?
Read the following passage from The Washington Post (November 25, 1997),
page A19. Then, for each of the numbered expressions, label the main argu-
mentative move, if any, using the same abbreviations as in Exercise I:
A PIECE OF “GOD’S HANDIWORK”
by Robert Redford
Just over a year ago, President Clinton created the Grand Staircase-Escalante Na-
tional Monument to [1] protect [2] once and for all some [3] of Utah’s extraordinary
red rock canyon country. In response to [4] plans of the Dutch company Andalex to
mine coal on the Kaiparowits Plateau, President Clinton used his authority under
the Antiquities Act to establish the new monument, setting aside for protection
what he described as “some of the most remarkable land in the world.” I couldn’t
agree more. [5] For over two decades, many have fought battle after battle [6] to keep
mining conglomerates from despoiling [7] the unique treasures [8] of this stunning
red rock canyon country. Now [9], we thought at least some of it was safe.
Not so. Shocking [10] as it sounds, Clinton’s Bureau of Land Management
(BLM) has approved oil drilling within the monument. BLM has given Conoco
Inc., a subsidiary of the corporate giant DuPont, permission to drill for oil and
Exercise II
93
An Extended Example
gas in the heart [11] of the new monument. You may [12] wonder, as I do, how can
this happen? [13] Wasn’t the whole purpose of creating the monument to
preserve its colorful cliffs, sweeping arches and other extraordinary resources
[14] from large-scale mineral development? Didn’t the president say he was sav-
ing [15] these lands from mining companies for our children and grandchildren?
The BLM says its hands are tied. [16] Why? Because [17] these lands were set aside
subject to “valid existing rights,” and Conoco has a lease that gives it the right to
drill. Sure [18] Conoco has a lease—more than one, in fact [19]—but [20] those leases
were originally issued without sufficient environmental study or public input. As
a result [21], none of them conveyed a valid right to drill. What’s more [22], in decid-
ing to issue a permit to drill now, the BLM did not conduct a full analysis of the en-
vironmental impacts of drilling in these incomparable lands, but instead [23]
determined there would be no significant environmental harm on the basis of an
abbreviated review that didn’t even look at drilling on the other federal leases.
Sounds like [24] Washington double-speak [25] to me. I’ve spent considerable
time on these extraordinary lands for years, and I know [26] that an oil rig in
their midst would have a major impact. What’s more [27], Conoco wants to drill
a well to find oil. Inevitably [28], more rigs, more roads, new pipelines, toxic [29]
wastes and bright lights would follow to get the oil out. The BLM couldn’t see
this, but [30] the U.S. Fish and Wildlife Service and the Environmental Protec-
tion Agency did. Both of those agencies recognized [31] the devastating [32] ef-
fects extensive oil drilling would have on this area and urged the BLM to refuse
to allow it, in order to [33] protect the monument.
Maybe [34] the problem [35] comes from giving management responsibility
for this monument to the BLM. This is the BLM’s first national monument;
almost [36] all the others are managed by the National Park Service. The Park
Service’s mission is to protect the resources [37] under its care while the bureau
has always sought to accommodate economic uses of those under its. Even so
[38], the BLM seemed [39] to be getting off to a good [40] start by enlisting broad
[41] public involvement in developing a management plan for the area. Yet
[42] the agency’s decision to allow oil drilling in the monument completely un-
dercuts [43] this process just as it is beginning.
What we’re talking about is, in the words of President Clinton, a small
piece of “God’s handiwork.” Almost [44] 41/2 million acres of irreplaceable red
rock wilderness remain outside the monument. Let us at least protect what is
within it. The many roadless [45] areas within the monument should [46] remain
so—protected as wilderness. The monument’s designation means little if [47] a
pattern of exploitation is allowed to continue.
Environmentalists—including the Southern Utah Wilderness Alliance, the
Natural Resources Defense Council, and the Wilderness Society—appealed
BLM’s decision to the Interior Department’s Board of Land Appeals. This ap-
peal, however [48], was rejected earlier this month. This is a terrible mistake [49].
We shouldn’t be drilling in our national monuments. Period. As President
Clinton said when dedicating the new monument, “Sometimes progress is
measured in mastering frontiers, but sometimes [50] we must measure progress
in protecting frontiers for our children and children to come.”
Allowing drilling to go forward in the Grand Staircase-Escalante Monu-
ment would permanently stain what might otherwise have been a defining
legacy of the Clinton presidency.
94
CHAPTER 4 ■ The Art of Close Analys is
Read the following advertisement from Equal Exchange (Copyright © 1997,
1998, 1999). For each of the numbered expressions, label the main argumenta-
tive move, if any, using the same abbreviations as in Exercise I. Then state
what you take to be the central conclusions and premises. What criticisms, if
any, do you have of this argument?
It may [1] be a little early in the morning to bring this up, but [2] if [3] you buy
coffee from large corporations [4], you are inadvertently maintaining the system
which keeps small farmers poor [5] while [6] lining the pockets [7] of rich cor-
porations. By [8] choosing Equal Exchange coffee, you can [9] help to make a
change. We believe in trading directly with small farming cooperatives at mutu-
ally agreed-upon prices with a fixed minimum rate. Then [10], should [11] the cof-
fee market decline, the farmers are still guaranteed a fair [12] price. So [13] have a
cup of Equal Exchange Coffee and make a small farmer happy [14]. Of course [15],
your decision to buy Equal Exchange need not be completely altruistic. For [16]
we take as much pride in refining the taste of our gourmet [17] coffees as [18] we
do in helping [19] the farmers who produce them. For [20] more information about
Equal Exchange or to order our line of gourmet, organic, and shade-grown coffee
directly, call 1 800 406 8289.
Exercise III
From Equal Exchange. Advertisement. Copyright 1997, 1998, 1999.
95
An Extended Example
This advertisement appeared in various national magazines in 2008. Circle
and label each of its key argumentative terms. State its central premises and
conclusion, and then put them into standard form. Is the result a good argu-
ment? Why or why not?
Exercise IV
96
CHAPTER 4 ■ The Art of Close Analys is
Provide a close analysis of the following passages by circling and labeling each
of the key argumentative terms. Then state what you take to be the central con-
clusions and premises. What criticisms, if any, do you have of each argument?
PASSAGE 1
The following encomium was written by an art critic for the New York Times, in
which this review appeared (June 6, 1992, pp. 1, 31). If you are not familiar
with Eakins’s work, looking at some reproductions might help you appreciate
this argument. You might also look up the rest of the essay, as only a small part
is reproduced here.
IS EAKINS OUR GREATEST PAINTER?
by John Russell
There never was a painter like Thomas Eakins, who was born in 1844, died in
1916, and is the subject of a great exhibition that has just opened in the
Philadelphia Museum of Art. It is not simply that in his hands painting be-
came an exact science, so that if he paints two men rowing on a river, we can
tell the month, day, and the hour that they passed under a certain bridge. We
admire Eakins for that, but we prize him above all for the new dimension of
moral awareness that he brought to American painting.
The question that he asks is not “What do we look like?” It is “What have
we done to one another?” And it is because he gives that question so full and
so convincing an answer that we ask ourselves whether Thomas Eakins was
not the greatest American painter who ever lived. Even if the question strikes
us as meaningless, we find it difficult after an exhibition such as this to think
of a convincing rival. . . .
PASSAGE 2
The following passage is an excerpt from an essay that appeared in the Boston
Review (vol. 7, no. 5 [October 1982], pp. 13–19), which is funded by the National
Endowment for the Humanities, a government agency that supports scholarship
and the arts. Its author was professor of drama at Stanford University.
BEYOND THE WASTELAND: WHAT AMERICAN
TV CAN LEARN FROM THE BBC
by Martin Esslin
What are the advantages and disadvantages of a public TV service as compared
to a completely commercial system? One of the dangers inherent in a public
service system is paternalism: Some authority decides what the viewers should
Exercise V
97
An Extended Example
see and hear simply on the basis of what it arbitrarily feels would be good for
them. Yet in countries where a highly developed public system exists alongside
a commercial one, that danger is minimized because of the market pressure on
the commercial system to give its audience what it wants. Indeed, in a dual sys-
tem the danger is often that the public service may be tempted to ignore its
stated purpose to serve the public interest and instead pander to mass prefer-
ences because of a sense of competition with the commercial networks.
Another problem that plagues public TV service is that it may run short of
money, which in turn can increase its dependence on the government. The ex-
tent of government dependence is intimately connected to how the public
broadcasting service is financed. In West Germany and Italy, for example, the
public broadcasting service takes advertising, but it is usually confined to a
clearly delimited area of the network’s programming. In Britain the BBC . . .
relies entirely on its annual license fee, which guarantees it a steady income
and allows long-term planning. In periods of severe inflation, the license-fee
system leaves the BBC in a dangerous position vis-a-vis the government, and
the network’s income may decline in real terms. In countries where the public
broadcasting service is financed by an annual allocation in the national
budget, long-term planning becomes more difficult and the dependence on the
government is far greater. Nevertheless, public TV services financed on that
pattern, such as the ABC in Australia and the CBC in Canada, provide pro-
grams of high quality that are genuine alternatives to the fare on the numer-
ous popular and prosperous commercial networks. In Canada, this includes
programs from the three commercial American networks.
One of the most important positive features of services under public con-
trol is their ability to provide planned, high-quality viewing alternatives. The
BBC, for example, has two television channels, BBC 1 and BBC 2. The program
planning on these two networks is closely coordinated so that highly popular
material on one channel is regularly paired with more specialized or demand-
ing fare on the other. And though the percentage of the audience that tunes in
to the challenging programming may be small, the scale of magnitude opera-
tive in the mass media is such that even a small percentage of the viewing au-
dience represents a very large number of people indeed. A popular dramatic
series on BBC 1, for instance, may reach an audience of 20 percent of the adult
population of Britain—about ten million people. A play by Shakespeare on
BBC 2 that may attract an audience of only 5 percent nonetheless reaches
about two-and-a-half million people—a substantial audience for a work of art.
It would take a theater with a seating capacity of 1,000 about seven years, or
2,500 performances, to reach an equivalent number of people! Nor should it be
overlooked that this audience will include people whose influence may be
greater in the long run than that of the ten million who watched the entertain-
ment program. In this system, no segment of the viewing public is forced to
compromise with any other. In our example, not only did BBC 1 provide a
popular entertainment program as an alternative to the Shakespeare, but, in
addition, the commercial network offered still another popular program. By
careful—perhaps paternalistic—planning the general audience satisfaction
was substantially increased.
(continued)
98
CHAPTER 4 ■ The Art of Close Analys is
One of the difficulties of the American situation is that the size of the
United States favors decentralization and the fragmentation of initiatives for
the more ambitious programming of the public service network. A revitalized
PBS would need a strong central governing body that could allocate to local
producing stations the substantial sums of money they require for ambitious
projects—projects that could compete with the best offerings of the rich com-
mercial competitors.
Using existing satellite technology, such a truly national network of public
service television could be made available to the entire country. If a public
service television organization was able to provide simultaneous, alternative
programming along the lines of BBC 1 and BBC 2, the cultural role of televi-
sion in the United States could be radically improved, and the most powerful
communication medium in history could realize its positive potential to in-
form, educate, and provide exposure to diverse cultural ideas.
PASSAGE 3
This article was published in the New York Times (January 15, 2004, p. A33).
LIFE (AND DEATH) ON MARS
by Paul Davies
Sydney, Australia—President Bush’s announcement yesterday that the United
States will soon be pointing its rockets toward Mars will doubtless be greeted
with delight by space scientists. . . . And yet the scientific community’s enthu-
siasm will surely be tempered by skepticism. Scientists, it’s worth remember-
ing, rejoiced when President George H. W. Bush unveiled a Mars project in
1989. The same scientists then despaired when the plan quickly evaporated
amid spiraling projected costs and shifting priorities. . . .
Why is going to Mars so expensive? Mainly it’s the distance from Earth. At
its closest point in orbit, Mars lies 35 million miles away from us, necessitating
a journey of many months, whereas reaching the Moon requires just a few
days’ flight. On top of this, Mars has a surface gravity that, though only 38
percent of Earth’s, is much greater than the Moon’s. It takes a lot of fuel to
blast off Mars and get back home. If the propellant has to be transported there
from Earth, costs of a launching soar. . . .
There is, however, an obvious way to slash the costs and bring Mars within
reach of early manned exploration. The answer lies with a one-way mission.
Most people react with instinctive horror at the suggestion. I recall my own
sense of discomfort when I met an aging American scientist who claimed to
have trained for a one-way mission to the Moon in the pre-Apollo days. And
in the case of the barren Moon, that reaction is largely justified. There is little
on the Moon to sustain human life. Mars, however, is a different story. Because
of the planet’s relatively benign environment, it is theoretically able to support
a permanent human presence. If provided with the right equipment, astronauts
99
An Extended Example
would have a chance of living there for years. A one-way trip to Mars need not
mean a quick demise.
Every two years the orbit of Mars creates a window of opportunity to send
fresh supplies at a reasonable cost. An initial colony of four astronauts, equipped
with a small nuclear reactor and a couple of rover vehicles, could make their
own oxygen, grow some food and even initiate building projects using local raw
materials. Supplemented by food shipments, medical supplies and replacement
gadgets from home, the colony could be sustained indefinitely. To be sure, the
living conditions would be uncomfortable, but the colonists would have the op-
portunity to do ground-breaking scientific work and blaze a trail that would en-
sure them a permanent place in the annals of discovery.
Obviously this strategy carries significant risks in addition to those faced by
a conventional Mars mission. Major equipment failure could leave the colony
without enough power, oxygen or food. An accident might kill or disable an as-
tronaut who provided some vital expertise. A supply drop might fail, condemn-
ing the colonists to starve in a very public way. Even if nothing went wrong, the
astronauts’ lives would certainly be shortened by the harsh conditions. . . .
Would it be right to ask people to accept such conditions for the sake of sci-
ence, or even humanity? The answer has to be yes. We already expect certain
people to take significant risks on our behalf, such as special forces operatives
or test pilots. Some people gleefully dice with death in the name of sport or ad-
venture. Dangerous occupations that reduce life expectancy through exposure
to hazardous conditions or substances are commonplace. . . .
Who would put their hand up for a one-way ticket to Mars? I work among
people who study astrobiology and planetary science, and there is no lack of
eager young scientists who would sign up right now, given half a chance. But
it would make more sense to pick mature, older scientists with reduced life ex-
pectancy. Other considerations, like weight, emotional stability and scientific
credentials, would of course have to be factored in.
The early outpost wouldn’t be left to wither and die. Rather, it would form
the basis for a much more ambitious colonization program. Over the years
new equipment and additional astronauts would be sent to join the original
crew. In time, the colony would grow to the point of being self-sustaining.
When this stage was reached, humanity would have a precious insurance
policy against catastrophe at home. During the next millennium there is a
significant chance that civilization on Earth will be destroyed by an asteroid,
a killer plague or a global war. A Martian colony could keep the flame of civ-
ilization and culture alive until Earth could be reverse-colonized from Mars.
Would NASA entertain a one-way policy for human Mars exploration?
Probably not. But other, more adventurous space agencies in Europe or Asia
might. The next giant leap for mankind won’t come without risk.
PASSAGE 4
The following letter to the editor was written by Charlie Buttrey, a lawyer, and
published in the White River Junction, Vermont, Valley News on Wednesday,
August 29, 2001.
(continued)
100
CHAPTER 4 ■ The Art of Close Analys is
THREE FALLACIES ABOUT GUN RIGHTS
by Charlie Buttrey
In the never-ending debate on these pages over gun ownership, those who fa-
vor expansive gun rights appear to have adopted one (or more) of three lines
of argument:
The Constitution protects the individual right to gun ownership.
The crime rate in Great Britain (where gun ownership is strictly controlled)
is “skyrocketing.”
In colonial days, the vast majority of people owned guns.
Each premise, however, is erroneous.
In our constitutional system, the ultimate arbiter of what the Constitution
says is the Supreme Court. One may not agree with its rulings (for example,
many people do not approve of its ruling extending First Amendment protec-
tions to flag burning or Fourth Amendment protections to the decision to ter-
minate a pregnancy), but, under our system of government, once the Supreme
Court has ruled a certain way, that ruling stands as the law of the land. On the
issue of whether the Second Amendment encompasses an individual’s right to
possess firearms, the Supreme Court has spoken. And the answer is that it
does not. Like it or not, that is the law of the land. It is true that the state con-
stitutions in both Vermont and New Hampshire provide for the protection of
individual gun ownership. In neither instance, however, is such a right ab-
solute. For example, the New Hampshire Supreme Court has ruled that a law
prohibiting felons from possessing firearms is not unconstitutional. Similarly,
a Vermont law prohibiting people from carrying loaded firearms in their vehi-
cles was held to be permissible.
The notion that Great Britain’s crime rate is skyrocketing is belied by the
facts. British government statistics reveal a steady increase in most crime cate-
gories between 1980 and 1995 and a gradual leveling off in the last five years.
Of course, gun-related crimes are minuscule when compared with those in the
United States. Guns are used in robberies in Great Britain less than 5 percent
of the time. And from 1980 to 1999, annual gun-related homicides in Britain
ranged from a low of 7 to a high of 42. In comparison, according to U.S. Bureau
of Justice statistics, an average of 10,000 gun-related homicides occurred each
year in this country during the same period.
Finally, the suggestion that our colonial forebears were almost universally
armed has been thoroughly repudiated. In his recent book, Arming America,
Emory University historian Michael Bellesiles reveals that the average colonial
citizen had virtually no access to or training in the use of firearms, and that the
few guns that did exist were kept under strict control. The fewer than 10 per-
cent of Americans who did possess guns in the years prior to the Civil War
were generally neglectful of the weapons and those guns were expensive,
clumsy, unreliable, and hard to maintain.
There may be sound policy grounds that can be advanced in favor of rela-
tively unrestricted gun ownership. Those arguments should be made, how-
ever, from a perspective of historical truth and legal accuracy.
101
An Extended Example
PASSAGE 5
The following extract comes from Compulsory Mis-Education and the Community
of Scholars (New York: Vintage Books, 1966). Copyright © 1962, 1964 by Paul
Goodman. Used by permission of Random House, Inc.
A PROPOSAL TO ABOLISH GRADING
by Paul Goodman
Let half a dozen of the prestigious universities—Chicago, Stanford, the Ivy
League—abolish grading, and use testing only and entirely for pedagogic pur-
poses as teachers see fit.
Anyone who knows the frantic temper of the present schools will understand
the transvaluation of values that would be effected by this modest innovation.
For most of the students, the competitive grade has come to be the essence. The
naïve teacher points to the beauty of the subject and the ingenuity of the re-
search; the shrewd student asks if he is responsible for that on the final exam.
Let me at once dispose of an objection whose unanimity is quite fascinating.
I think that the great majority of professors agree that grading hinders teaching
and creates a bad spirit, going as far as cheating and plagiarizing. I have before
me the collection of essays, Examining in Harvard College, and this is the consen-
sus. It is uniformly asserted, however, that the grading is inevitable; for how
else will the graduate schools, the foundations, the corporations know whom to
accept, reward, hire? How will the talent scouts know whom to tap? By testing
the applicants, of course, according to the specific task-requirements of the in-
ducting institution, just as applicants for the Civil Service or for licenses in
medicine, law, and architecture are tested. Why should Harvard professors do
the testing for corporations and graduate schools? . . .
There are several good reasons for testing, and kinds of test. But if the aim
is to discover weakness, what is the point of down-grading and punishing,
and thereby inviting the student to conceal his weakness, by faking and
bulling, if not cheating? The natural conclusion of [education] is the insight it-
self, not a grade for having had it. For the important purpose of placement, if
one can establish in the student the belief that one is testing not to grade and
making invidious comparisons but for his own advantage, the student should
normally seek his own level, where he is challenged and yet capable, rather
than trying to get by. If the student dares to accept himself as he is, a teacher’s
grade is a crude instrument compared with a student’s self-awareness. But it
is rare in our universities that students are encouraged to notice objectively
their vast confusion. Unlike Socrates, our teachers rely on power-drives rather
than shame and ingenuous idealism.
Many students are lazy, so teachers try to goad or threaten them by grad-
ing. In the long run this must do more harm than good. Laziness is a character-
defense. It may be a way of avoiding learning, in order to protect the conceit
that one is already perfect. . . . It may be a way of avoiding just the risk of
(continued)
102
CHAPTER 4 ■ The Art of Close Analys is
failing and being down-graded. Sometimes it is a way of politely saying, “I
won’t.” But since it is the authoritarian grown-up demands that have created
such attitudes in the first place, why repeat the trauma? There comes a time
when we must treat people as adults, laziness and all. It is one thing coura-
geously to fire a do-nothing out of your class; it is quite another thing to eval-
uate him with a lordly F.
Most important of all, it is often obvious that balking in doing the work,
especially among bright young people who get to great universities, means
exactly what it says: The work does not suit me, not this subject, or not at this
time, or not in this school, or not in school altogether. The student might not
be bookish; he might be school-tired; perhaps his development ought now to
take another direction. Yet, unfortunately, if such a student is intelligent and is
not sure of himself, he can be bullied into passing, and this obscures every-
thing. My hunch is that I am describing a common situation. What a grim
waste of young life and teacherly effort! Such a student will retain nothing of
what he has “passed” in. Sometimes he must get mononucleosis to tell his
story and be believed.
And, ironically, the converse is also probably commonly true. A student
flunks and is mechanically weeded out, who is really ready and eager to learn
in a scholastic setting, but he has not quite caught on. A good teacher can rec-
ognize the situation, but the computer wreaks its will.
PASSAGE 6
The following op-ed appeared in The Dartmouth on April 30, 2007. The author
was a college student at the time.
MAKE IT 18
by Ben Selznick
Did you get wasted this weekend? Are you under the age of 21? Don’t fret, my
underage friends, you are not alone. Across America, hundreds of thousands
of underage men and women enjoyed booze this weekend—illegally. In re-
sponse to a recent news article in The Dartmouth (“Ex-Middlebury president:
lower drinking age to 18,” April 25), I would like to agree with the former
president of Middlebury College that states should try once again lowering
the drinking age to 18.
We must first get our facts straight. The drinking age is not set by the fed-
eral government; it is set by the states. Currently, each state has its law set at
21. However, this figure neither necessarily reflects research and careful con-
sideration nor appeals to constituents. Instead, it reflects the fact that drinking
age became tied to highway funds somewhere along the way. Because of this,
were a state to lower their drinking age, they would also have to significantly
103
An Extended Example
raise taxes to offset the lost highway funds from the government. This hand-
tying is asinine. If the drinking age is a state decision, it should be made by
states independently of funds for highways. Before any progress can be made,
lawmakers must sever this illogical connection.
Were drinking age to become a true state consideration, states could engage
in open debate about the issue. Several arguments could then be made in
favor of lowering the age to 18.
First: A fair portion of drinkers between the ages of 18 and 21 are at colleges
where alcohol can be easily obtained. Just think about this weekend and all the
people without wristbands—because they forgot their IDs at home I’m sure—
who were able to obtain alcohol. This easy availability leads to a very danger-
ous game of cat and mouse between colleges and their students. In this game,
colleges know full well that underage drinking exists. Students know that they
are drinking underage and therefore illegally. And yet, the game continues
with colleges trying to make sure that no alcohol-related tragedies occur. This
game isn’t the fault of colleges, nor is it the fault of students. Instead, it is the
fault of our law, which classifies part of this college-age set as “underage” and
the rest as “of age.” When it comes to colleges, our law simply doesn’t reflect
reality. It must therefore be changed.
Second: At present, keeping the drinking age at 21 prevents America from
having an open and honest conception of alcohol the way, say, continental
Europe does. In France, where the drinking age is 16, alcohol is understood as a
complement to a meal as well as an intoxicant. Though people certainly drink to
get drunk, the concept of binge drinking beer through a funnel is seen as a bar-
baric abuse of alcohol. More importantly, France’s policy reflects the understand-
ing that people are going to drink alcohol whether the activity is legal or not.
America stubbornly refuses to accept this fact. Perhaps it is our long history of
teetotalers, prohibition and general dislike of alcohol by various groups. What-
ever our past, our present situation is one where binge drinking is the norm and
alcohol education to college students reeks of being too little too late.
Third: The main concern against lowering the drinking age comes from
Mothers Against Drunk Driving. MADD has statistics to back up their claim
that a lower drinking age yields more drunk driving accidents. But if the
drinking age were lowered and then supplemented with honest, not pander-
ing, alcohol education programs in schools, perhaps over time our approach
towards alcohol as a culture would change.
America has the highest drinking age of any country on Earth. Although in-
tended to protect us, in the end this high barrier leads to unsafe abuse of alco-
hol nationwide. States must engage in true conversation with their
constituents and consider arguments from frustrated college administrators as
well as MADD.
In the end, a state may conclude that 21 is a good age. But they may also
conclude that lowering the drinking age is a practical solution to a pressing
problem. After all, by changing the definition of “underage drinking,” we
could alleviate its negative effects.
104
CHAPTER 4 ■ The Art of Close Analys is
Describe the purpose of, intended audience for, and standpoint in each of the
arguments in Exercises I-VI.
Exercise VII
Practice close analysis some more by doing close analyses of:
1. one of the passages in the Discussion Questions at the end of Chapter 1,
2. one of the articles in Part V of this book,
3. an editorial or advertisement from your local paper,
4. something that you read for another course,
5. a lecture by your professor in another course (or this course!), or
6. a paper by you or by a friend in another course.
Exercise VI
1. If, as some social critics have maintained, the pervasive nature of television
has created generation upon generation of intellectually passive
automatons, why study close analysis?
2. Television commercials are often arguments in miniature. Recount several
recent television commercials and identify the argumentative devices at
work.
Discussion Questions
DEEP ANALYSIS
Arguments in everyday life rarely occur in isolation. They usually come in the mid-
dle of much verbiage that is not essential to the argument itself. Everyday arguments
are also rarely complete. Essential premises are often omitted. Many such omissions
are tolerable because we are able to convey a great deal of information indirectly by
conversational implication. Nevertheless, to give a critical evaluation of an argu-
ment, it is necessary to isolate the argument from extraneous surroundings, to make
explicit unstated parts of the argument, and to arrange them in a systematic order.
This puts us in a better position to decide on the soundness or unsoundness of the
argument in question. This chapter will develop methods for reconstructing argu-
ments so that they may be understood and evaluated in a fair and systematic fashion.
These methods will then be illustrated by applying them to a disagreement that
depends on fundamental principles.
GETTING DOWN TO BASICS
To understand an argument, it is useful to put it into standard form. As we
saw in Chapter 3, this is done simply by writing down the numbered prem-
ises, drawing a line, adding “ ” followed by the conclusion, and indicating
which premises are supposed to be reasons for the conclusion. That is all we
write down in standard form, but there is often a lot more in the passage
that includes the argument. It is not uncommon for the stated argument to
stretch over several pages, whereas the basic argument has only a few prem-
ises and a single conclusion.
One reason for this is that people often go off on tangents. They start to
argue for one claim, but that reminds them of something else, so they talk
about that for a while; then they finally return to their original topic. One ex-
ample occurred during the Republican presidential candidates’ debate on
October 9, 2007, when Governor Mitt Romney said,
. . . We’re also going to have to get serious about treating Ahmadinejad [the
President of Iran] like the rogue and buffoon that he is. And it was outrageous
for the United Nations to invite him to come to this country. It was outrageous
for Columbia to invite him to speak at their university. This is a person who
denied the Holocaust, a person who has spoken about genocide, is seeking the
5
105
106
CHAPTER 5 ■ Deep Analys is
means to carry it out. And it is unacceptable to this country to allow that individ-
ual to have control of launching a nuclear weapon. And so we will take the
action necessary to keep that from happening . . .
Romney’s criticisms of the United Nations and Columbia are not really part
of his argument, because they do not support his conclusion that the United
States needs to keep nuclear weapons out of the hands of Ahmadinejad.
Such tangents can be completely irrelevant or unnecessary, and they
often make it hard to follow the argument. Some people even go off on tan-
gents on purpose to confuse their opponents and hide gaping holes in their
arguments. The irrelevant diversion is sometimes called a red herring (after a
man who, when pursued by hounds, threw them off his scent by dragging a
red herring across his trail). More generally, this maneuver might be called
the trick of excess verbiage. It violates the conversational rules of Quantity,
Relevance, or Manner, which were discussed in Chapter 2.
To focus on the argument itself, we need to look carefully at each parti-
cular sentence to determine whether it affects the validity or strength of the
argument or the truth of its premises. If we decide that a sentence is not nec-
essary for the argument, then we should not add it when we list the prem-
ises and conclusion in standard form. Of course, we have to be careful not to
omit anything that would improve the argument, but we also do not want
to include too much, because irrelevant material simply makes it more diffi-
cult to analyze and evaluate the argument.
Another source of extra material is repetition. Consider Senator John
Edwards’s response to a question about the Defense of Marriage Act in the
Democratic presidential candidates’ debate on January 22, 2004:
These are issues that should be left [to the states]. Massachusetts, for example,
has just made a decision—the Supreme Court at least has made a decision—that
embraces the notion of gay marriage. I think these are decisions the states should
have the power to make. And the Defense of Marriage Act, as I understand it—
you’re right, I wasn’t there when it was passed—but as I understand it, would
have taken away that power. And I think that’s wrong—that power should not
be taken away from the states. . . .
Now compare:
These are issues that should be left to the states.
These are decisions that states should have the power to make.
That power should not be taken away from the states.
All three of these sentences say pretty much the same thing, so we do not
need them all.
Why do people repeat themselves like this? Sometimes they just forget that
they already made the point before, but often repetition accomplishes a goal.
Good speakers regularly repeat their main points to remind their audience of
what was said earlier. Repetition is subtler when it is used to explain some-
thing. A point can often be clarified by restating it in a new way. Repetition
107
Gett ing Down to Bas ics
can also function as a kind of assurance, as an expression of confidence, or as
an indication of how important a point is. Some writers seem to think that if
they say something often enough, people will come to believe it. Whether or
not this trick works, if two sentences say equivalent things, there is no need to
list both sentences when the argument is put into standard form. Listing the
same premise twice will not make the argument any better from a logical
point of view.
Sometimes guarding terms can also be dropped. If I say, “I think Miranda
is at home, so we can probably meet her there,” this argument might be rep-
resented in standard form thus:
(1) I think Miranda is at home.
(2) We can probably meet her there. (from 1)
This is misleading. My thoughts are not what make us able to meet Miranda
at home. My thoughts do not even increase the probability that she is at
home or that we can meet her there. It is the fact that Miranda is at home that
provides a reason for the conclusion. Thus, it is clearer to drop the guarding
phrase (“I think”) when putting the argument into standard form. But you
have to be careful, for not all guarding phrases can be dropped. When I say
“We can probably meet her there,” I might not want to say simply, “We can
meet her there.” After all, even if she is there now, we might not be able to
get there before she leaves. Then to drop “probably” from my conclusion
would distort what I meant to say and would make my argument more
questionable, so you should not drop that guarding term if you want to un-
derstand my argument charitably and accurately.
Here’s another example: If a friend says that you ought to buckle your
seat belt because you could have an accident, it would distort her argument
to drop the guarding term (“could”), because she is not claiming that you
definitely will have an accident, or even that you probably will have one.
The chance of an accident is significant enough to show that you ought to
buckle your seat belt, so this guarding term should be kept when the argu-
ment is put into standard form.
It is also possible to drop assuring terms in some cases. Suppose someone
says, “You obviously cannot play golf in Alaska in January, so there’s no
point in bringing your clubs.” There is no need to keep the assuring term
(“obviously”) in the premise. It might even be misleading, because the issue
is whether the premise is true, not whether it is obvious. The argument can-
not be refuted by showing that, even though you in fact cannot play golf in
Alaska in January, this is not obvious, since there might be indoor golf
courses. In contrast, assuring terms cannot be dropped in some other cases.
For example, if someone argues, “We know that poverty causes crime, be-
cause many studies have shown that it does,” then the assuring terms (“We
know that . . .” and “studies have shown that . . .”) cannot be dropped with-
out turning the argument into an empty shell: “Poverty causes crime, be-
cause it does.” The point of this argument is to cite the sources of our
108
CHAPTER 5 ■ Deep Analys is
knowledge (“studies”) and to show that we have knowledge instead of just
a hunch. That point is lost if we drop the assuring terms.
Unfortunately, there is no mechanical method for determining when
guarding or assuring terms and phrases can be dropped, or whether certain
sentences are unnecessary tangents or repetition. We simply have to look
closely at what is being said and think hard about what is needed to support
the conclusion. It takes great skill, care, and insight to pare an argument
down to its essential core without omitting anything that would make it bet-
ter. And that is the goal: If you want to understand someone’s argument,
you should try to make that argument as good as it can be. You should in-
terpret it charitably. Distorting and oversimplifying other people’s argu-
ments might be fun at times and can win points in debates, but it cannot
help us understand or learn from other people’s arguments.
In the quotation above, is it fair to drop “I think” from the start of Edwards’s
sentences “I think these are decisions the states should have the power to make”
and “I think that’s wrong—that power should not be taken away from the
states“? Why or why not? Is this phrase “I think” used for guarding or assuring
or some other purpose in this context? Explain why Edwards adds these words.
Discussion Question
Put the following arguments into standard form and omit anything that does
not affect the validity of the argument or the truth of its premises.
1. Philadelphia is rich in history, but it is not now the capital of the United
States, so the United States Congress must meet somewhere else.
2. Not everybody whom you invited is going to come to your party. Some of
them won’t come. So this room should be big enough.
3. I know that my wife is at home, since I just called her there and spoke to
her. We talked about our dinner plans.
4. I’m not sure, but Joseph is probably Jewish. Hence, he is a rabbi if he is a
member of the clergy.
5. Some students could not concentrate on the lecture, because they did not
eat lunch before class, although I did.
6. The most surprising news of all is that Johnson dropped out of the race
because he thought his opponent was better qualified than he was for
the office.
7. The Democrat is likely to win, since experts agree that more women sup-
port him.
8. It seems to me that married people are happier, so marriage must be a
good thing, or at least I think so.
Exercise I
109
Dissect ing the Argument
CLARIFYING CRUCIAL TERMS
After the essential premises and conclusion are isolated, we often need to clar-
ify these claims before we can begin our logical analysis. The goal here is not
perfect clarity, for there probably is no such thing. It is, however, often neces-
sary to eliminate ambiguity and reduce vagueness before we can give an ar-
gument a fair assessment. In particular, it is usually helpful to specify the
referents of pronouns, because such references can depend on a context that is
changed when the argument is put into standard form. “You are wrong” or
“That’s wrong” can be perfectly clear when said in response to a particular
claim, but they lose their clarity when they are moved into the conclusion of
an argument in standard form. We also often need to specify whether a claim
is about all, most, many, or just some of its subject matter. When people say,
“Blues music is sad,” do they mean all, most, some, or typical blues music?
Another common problem arises when someone argues like this:
You should just say “No” to drugs, because drugs are dangerous.
What counts as a drug? What about penicillin or aspirin? The speaker might
seem to mean “drugs like cocaine,” but “like” them in which respects? Maybe
what is meant is “addictive drugs,” but what about alcohol and nicotine
(which are often addictive)? You might think that the speaker means “danger-
ous drugs,” but then the premise becomes empty: “Dangerous drugs are dan-
gerous.” Or maybe the idea is “illegal drugs,” but that seems to assume that
the law is correct about what is dangerous. In any case, we cannot begin to
evaluate this argument if we do not know the extent of what it claims.
Of course, we should not try to clarify every term in the argument. Even
if this were possible, it would make the argument extremely long and bor-
ing. Instead, our goal is to clarify anything that seems likely to produce con-
fusion later if it is not cleared up now. As our analysis continues, we can
always return and clarify more if the need arises, but it is better to get the
most obvious problems out of the way at the start.
Some problems, however, just won’t go away. Don’t get frustrated if you
cannot figure out how to clarify a crucial term in someone else’s argument.
The fault might lie with the person who gave the argument. Often an argu-
ment leaves a crucial term vague or ambiguous, because serious defects in
the argument would become apparent if its terms were made more precise.
We will discuss such tricks in detail in Chapters 13–14. For now, we just need
to try our best to understand and clarify the essential terms in the argument.
DISSECTING THE ARGUMENT
A single sentence often includes several clauses that make separate claims.
When this happens, it is usually useful to dissect the sentence into its small-
est parts, so that we can investigate each part separately. Because simpler
110
CHAPTER 5 ■ Deep Analys is
steps are easier to follow than complex ones, we can understand the argu-
ment better when it is broken down. Dissection makes us more likely to no-
tice any flaws in the argument. It also enables us to pinpoint exactly where
the argument fails, if it does.
The process of dissecting an argument is a skill that can be learned only
by practice. Let’s start with a simple example:
Joe won his bet, because all he had to do was eat five pounds of oysters,
and he ate nine dozen oysters, which weigh more than five pounds.
The simplest unpacking of this argument yields the following restatement
in standard form:
(1) All Joe had to do was eat five pounds of oysters, and he ate nine
dozen oysters, which weigh more than five pounds.
(2) Joe won his bet. (from 1)
If we think about the premise of this argument, we see that it actually
contains three claims. The argument will be clearer if we separate these
claims into independent premises and add a few words for the sake of clar-
ity. The following, then, is a better representation of this argument:
(1) All Joe had to do (to win his bet) was eat five pounds of oysters.
(2) Joe ate nine dozen oysters.
(3) Nine dozen oysters weigh more than five pounds.
(4) Joe won his bet. (from 1–3)
With the premise split up in this way, it becomes obvious that there are three
separate ways in which the argument could fail. One possibility is that the
first premise is false because Joe had to do more than just eat five pounds of
oysters to win his bet: Maybe what he bet was that he could eat five pounds
in five minutes. Another possibility is that the second premise is false because
Joe did not really eat nine dozen oysters: Maybe he really ate one dozen
oysters cut into nine dozen pieces. A final way in which the argument could
fail is if the third premise is false because nine dozen oysters do not weigh
more than five pounds: Maybe the oysters that Joe ate were very small, or
maybe nine dozen oysters weigh more than five pounds only when they are
still in their shells, but Joe did not eat the shells. In any case, breaking down
complex premises into simpler ones makes it easier to see exactly where the
argument goes wrong, if it does. Consequently, we can be more confident
that an argument does not go wrong if we do not see any problem in it even
after we have broken it down completely.
Although it is a good idea to break down the premises of an argument
when this is possible, we have to be careful not to do this in a way that
changes the logical structure of the argument. Suppose someone argues
like this:
Socialism is doomed to failure because it does not provide the incentives
that are needed for a prosperous economy.
111
Arrang ing Subarguments
The simplest representation of this argument yields the following standard
form:
(1) Socialism does not provide the incentives that are needed for a
prosperous economy.
(2) Socialism is doomed to failure. (from 1)
It is tempting to break up the first premise into two parts:
(1) Socialism does not provide incentives.
(2) Incentives are needed for a prosperous economy.
(3) Socialism is doomed to failure. (from 1–2)
In this form, the argument is open to a fatal objection: Socialism does provide
some incentives. Workers often get public recognition and special privileges
when they produce a great deal in socialist economies. But this does not re-
fute the original argument. The point of the original argument was not that
socialism does not provide any incentives at all, but only that socialism does
not provide enough incentives or the right kind of incentives to create a
prosperous economy. This point is lost if we break up the premise in the way
suggested. A better attempt is this:
(1) Socialism does not provide adequate incentives.
(2) Adequate incentives are needed for a prosperous economy.
(3) Socialism is doomed to failure. (from 1–2)
The problem now is to specify when incentives are adequate. What kinds of
incentives are needed? How much of these incentives? The answer seems to
be “enough for a prosperous economy.” But then premise 2 reduces to
“Enough incentives for a prosperous economy are needed for a prosperous
economy.” This is too empty to be useful. Thus, we are led back to some-
thing like the original premise:
(1) Socialism does not provide enough incentives for a prosperous
economy.
(2) Socialism is doomed to failure. (from 1)
In this case, we cannot break the premise into parts without distorting
the point.
ARRANGING SUBARGUMENTS
When the premises of an argument are dissected, it often becomes clear that
some of these premises are intended as reasons for others. The premises
then form a chain of simpler arguments that culminate in the ultimate con-
clusion, but only after some intermediate steps. Consider this argument:
There’s no way I can finish my paper before the 9 o’clock show, since I have
to do the reading first, so I won’t even start writing until at least 9 o’clock.
112
CHAPTER 5 ■ Deep Analys is
It might seem tempting to put this argument into standard form as:
(1) I have to do the reading first.
(2) I won’t even start writing until at least 9 o’clock.
(3) I can’t finish my paper before the 9 o’clock show. (from 1–2)
This reformulation does include all three parts of the original argument, but it
fails to indicate the correct role for each part. The two argument markers in
the original argument indicate that there are really two conclusions. The word
“since” indicates that what precedes it is a conclusion, and the word “so”
indicates that what follows it is also a conclusion. We cannot represent this
as a single argument in standard form, because each argument in standard
form can have only one conclusion. Thus, the original sentence must have
included two arguments. The relationship between these arguments should
be clear: The conclusion of the first argument functions as a premise or reason
in the second argument. To represent this, we let the two arguments form a
chain. This is the first argument:
(1) I have to do the reading first.
(2) I won’t even start writing until at least 9 o’clock. (from 1)
This is the second argument:
(2) I won’t even start writing until at least 9 o’clock.
(3) I can’t finish my paper before the 9 o’clock show. (from 2)
If we want to, we can then write these two arguments in a chain like this:
(1) I have to do the reading first.
(2) I won’t even start writing until at least 9 o’clock. (from 1)
(3) I can’t finish my paper before the 9 o’clock show. (from 2)
This chain of reasoning can also be diagrammed like this:
(1)
(2)
(3)
The arrows indicate which claims are supposed to provide reasons for
which other claims.
Although it is often illuminating to break an argument into stages and
arrange them in a single series, this can be misleading if done incorrectly. For
example, the first sentences of Kyl’s speech cited in Chapter 4 read as follows:
Mr. Speaker, I oppose this measure. I oppose it first because it is expensive.
I further oppose it because it is untimely.
113
Arrang ing Subarguments
If we try to force this into a simple chain, we might get this:
(1) This measure is expensive.
(2) This measure is untimely. (from 1)
(3) I oppose this measure. (from 2)
This reconstruction suggests that the measure’s being expensive is what
makes it untimely. That might be true (say, during a temporary budget crisis),
but it is not what Kyl actually says. Instead, Kyl is giving two separate reasons
for the same conclusion. First,
(1) This measure is expensive.
(2) I oppose this measure. (from 1)
Second,
(1*) This measure is untimely.
(2) I oppose this measure. (from 1*)
The structure of this argument can now be diagrammed as a branching
tree:
(1) (1*)
(2)
The two arrows indicate that there are two separate reasons for the conclu-
sion. We have to be careful not to confuse branching arguments like this
with chains of arguments that do not branch.
We also need to distinguish this branching structure from cases where
several premises work together to support a single conclusion. Consider
this:
My keys must be either at home or at the office. They can’t be at the
office, because I looked for them there. So they must be at home.
With some clarifications, we can put this argument in standard form:
(1) My keys are either at my home or at my office.
(2) My keys are not at my office.
(3) My keys are at my home. (from 1–2)
Although this argument has two premises, it does not give two separate rea-
sons for its conclusion. Neither premise by itself, without the other, is
enough to give us any reason to believe the conclusion: “My keys are either
at my home or at my office” alone is not enough to support “My keys are at
my home,” and “My keys are not at my office” alone is also not enough to
support “My keys are at my home.” The premises work only when they
work together. Thus, it would be misleading to diagram this argument in the
same way as Kyl’s argument.
114
CHAPTER 5 ■ Deep Analys is
Instead, we need to indicate that the premises work together. Here’s a
simple way:
(1) � (2)
(3)
The symbol “+” with a single arrow indicates that the two premises together
provide a single reason for the conclusion. The line under the premises that
are joined together makes it clear that those are the premises that lead to the
conclusion at the end of the arrow. If three or more premises provided a single
reason, then we could simply add to the list—(1) + (2) + (3), and so on—then
draw a line under the premises to show which ones work together.
The argument that we are diagramming included one part that we have
not incorporated yet:
They can’t be at the office, because I looked for them there.
The standard form is this:
(2*) I looked for my keys at my office.
(2) My keys can’t be at my office. (from 2*)
By itself, this argument has this diagram:
(2*)
(2)
Since the conclusion of this background argument is a premise in the other
part of the argument, we can put the diagrams together like this:
(2*)
(1) � (2)
(3)
The fact that the arrow goes from (2*) to (2) but not to (1) indicates that this
background argument supports premise (2), but not the other premise. In
cases like this, you need to be careful where you draw your arrows.
Argument structures can get very complex, but we can diagram most
arguments by connecting the simple forms that we illustrated. Begin by iden-
tifying the premises and conclusions. Give each different claim a different
number. When two premises work together to support a single conclusion,
put a “+” between the premises and a line under them connected to a single
arrow that points to the conclusion. When two or more premises (or sets of
115
Arrang ing Subarguments
premises) provide separate reasons for a conclusion, draw separate arrows
from each reason to the conclusion. When a conclusion of one argument is
a premise in another, put it in the middle of a chain. The whole diagram
together will then show how the parts of the argument fit together.
Put the following arguments into standard form. Break up the premises and
form chains of arguments wherever this can be done without distorting the
argument. Then diagram the argument.
1. I know that Pat can’t be a father, because she is not a male. So she can’t be
a grandfather either.
2. Either Jack is a fool or Mary is a crook, because she ended up with all of
his money.
3. Our team can’t win this Saturday, both because they are not going to play,
and because they are no good, so they wouldn’t win even if they did play.
4. Mercury is known to be the only metal that is liquid at room temperature,
so a pound of mercury would be liquid in this room, which is at room
temperature, and it would also conduct electricity, since all metals do.
Therefore, some liquids do conduct electricity.
5. Since he won the lottery, he’s rich and lucky, so he’ll probably do well in
the stock market, too, unless his luck runs out.
6. Joe is not a freshman, since he lives in a fraternity, and freshmen are not
allowed to live in fraternities. He also can’t be a senior, since he has not
declared a major, and every senior has declared a major. And he can’t be a
junior, because I never met him before today, and I would have met him
before now if he were a junior. So Joe must be a sophomore.
7. Since many newly emerging nations do not have the capital resources
necessary for sustained growth, they will continue to need help from
industrial nations to avoid mass starvation.
Exercise II
In “A Piece of ‘God’s Handiwork’” (Exercise II in Chapter 4), Robert Redford
argues that the Bureau of Land Management (BLM) should not allow Conoco
to drill for oil in Utah’s Grand Staircase-Escalante National Monument. The
following passage is a crucial part where Redford answers an objection.
Arrange its subarguments in standard form so as to reveal the structure of his
argument. Then diagram the overall argument.
The BLM says its hands are tied. Why? Because these lands were set aside subject
to “valid existing rights,” and Conoco has a lease that gives it the right to drill.
Sure Conoco has a lease—more than one, in fact—but those leases were originally
Exercise III
(continued)
116
CHAPTER 5 ■ Deep Analys is
issued without sufficient environmental study or public input. As a result, none of
them conveyed a valid right to drill. What’s more, in deciding to issue a permit to
drill now, the BLM did not conduct a full analysis of the environmental impacts of
drilling in these incomparable lands, but instead determined there would be no
significant environmental harm on the basis of an abbreviated review that didn’t
even look at drilling on the other federal leases. Sounds like Washington double-
speak to me.
During the Republican candidates’ debate on October 9, 2007, Chris Matthews
asked Senator John McCain, “. . . Do you believe that Congress has to autho-
rize a strategic attack, not an attack on—during hot pursuit, but a strategic
attack on weaponry in Iran—do you need congressional approval as com-
mander and chief?” Read McCain’s response, then arrange its subarguments
in standard form so as to reveal the structure of his argument. Then diagram
the overall argument.
McCain: We’re dealing, of course, with hypotheticals. If the situation is that it
requires immediate action to ensure the security of the United States of America,
that’s what you take your oath to do, when you’re inaugurated as president of the
United States. If it’s a long series of build-ups, where the threat becomes greater and
greater, of course you want to go to Congress; of course you want to get approval, if
this is an imminent threat to the security of the United States of America. So it obvi-
ously depends on the scenario. But I would, at minimum, consult with the leaders of
Congress because there may come a time when you need the approval of Congress.
And I believe that this is a possibility that is, maybe, closer to reality than we are
discussing tonight.
Exercise IV
SUPPRESSED PREMISES
Arguments in everyday life are rarely completely explicit. They usually de-
pend on unstated assumptions that are understood by those involved in the
conversation. Thus, if we are told that Chester Arthur was a president of the
United States, we have a right to conclude a great many things about him—
for example, that at the time he was president, he was a live human being.
Appeals to facts of this kind lie behind the following argument:
Benjamin Franklin could not have been our second president, because he
died before the second election was held.
This argument obviously turns on a question of fact: Did Franklin die before
the second presidential election was held? (He did.) The argument would
not be sound if this explicit premise were not true. But the argument
117
Suppressed Prem ises
also depends on a more general principle that ties the premise and conclu-
sion together:
The dead cannot be president.
This new premise is needed to make the argument valid in the technical
sense.
This new premise is also needed to explain why the premise supports the
conclusion. You could have made the original argument valid simply by
adding this:
If Franklin died before the second election was held, then he could not
have been our second president.
Indeed, you can always make an argument valid simply by adding a con-
ditional whose antecedent is the premises and whose consequent is the
conclusion. However, this trick is often not illuminating; it does not reveal
how the argument works. In our example, there is nothing special about
Franklin, so it is misleading to add a conditional that mentions Franklin in
particular. In contrast, when we add the general principle, “The dead cannot
be president,” this new premise not only makes the argument valid but also
helps us understand how the conclusion is supposed to follow from the
premise.
Traditionally, logicians have called premises that are not stated but are
needed (to make the argument valid and explain how it works) suppressed prem-
ises. An argument depending on suppressed premises is called an enthymeme
and is said to be enthymematic. If we look at arguments that occur in daily life,
we discover that they are, almost without exception, enthymematic. Therefore,
to trace the pathway between premises and conclusion, it is usually necessary
to fill in these suppressed premises that serve as links between the stated prem-
ises and the conclusion.
CONTINGENT FACTS
Suppressed premises come in several varieties. They often concern facts
or conventions that might have been otherwise—that are contingent rather
than necessary. Our example assumed that the dead are not eligible for the
presidency, but we can imagine a society in which the deceased are elected to
public office as an honor (something like posthumous induction into the
Baseball Hall of Fame). Our national government is not like that, however,
and this is something that most Americans know. This makes it odd to come
right out and say that the deceased cannot hold public office. In most
settings, this would involve a violation of the conversational rule of Quantity,
because it says more than needs to be said.
Even though it would be odd to state it, this fact plays a central role in the
argument. To assert the conclusion without believing the suppressed premise
would involve a violation of the conversational rule of Quality, because the
speaker would not have adequate reasons for the conclusion. Furthermore, if
118
CHAPTER 5 ■ Deep Analys is
this suppressed premise were not believed to be true, then to give the explicit
premise as a reason for the conclusion would violate the conversational rule of
Relevance (just as it would be irrelevant to point out that Babe Ruth is dead
when someone asks whether he is in the Baseball Hall of Fame). For these
reasons, anyone who gives the original argument conversationally implies
a commitment to the suppressed premise.
Suppressed premises are not always so obvious. A somewhat more com-
plicated example is this:
Arnold Schwarzenegger cannot become president of the United States,
because he was born in Austria.
Why should being from Austria disqualify someone from being president?
It seems odd that the Founding Fathers should have something against that
particular part of the world. The answer is that the argument depends on a
more general suppressed premise:
Only a natural-born United States citizen may become president of the
United States.
It is this provision of the United States Constitution that lies at the heart of
the argument. Knowing this provision is, of course, a more specialized piece
of knowledge than knowing that you have to be alive to be president. For
this reason, more people will see the force of the first argument (about
Franklin) than the second (about Schwarzenegger). The second argument
assumes an audience with more specialized knowledge.
The argument still has to draw a connection between being born in Aus-
tria and being a natural-born United States citizen. So it turns out that the
argument has three stages:
(1) Schwarzenegger was born in Austria.
(2) Austria has never been part of the United States.
(3) Schwarzenegger was born outside of the United States. (from 1–2)
(4) Anyone who was born outside of the United States is not a natural-
born United States citizen.
(5) Schwarzenegger is not a natural-born United States citizen. (from 3–4)
(6) Only a natural-born United States citizen may become president of
the United States.
(7) Schwarzenegger cannot become president of the United States.
(from 5–6)
With the addition of suppressed premises (2), (4), and (6), the argument is
technically valid, for, if (1)–(2) are true, (3) must be true; if (3)–(4) are true,
(5) must also be true; and if (5)–(6) are true, then (7) must be true.
The argument is still not sound, however, because some of the suppressed
premises that were added are not true. In particular, there is an exception to
the suppressed premise about who is a natural-born United States citizen.
This exception is well known to United States citizens who live overseas.
119
Suppressed Prem ises
People who were born in Austria are United States citizens if their parents
were United States citizens. They also seem to count as natural-born citizens,
since they are not naturalized. This is not completely settled, but it does not
matter here, as Arnold Schwarzenegger’s parents were not United States citi-
zens when he was born. Thus, the second stage of the above argument can be
reformulated as follows:
(3) Schwarzenegger was born outside of the United States.
(4*) Schwarzenegger’s parents were not United States citizens when
he was born.
(4**) Anyone who was born outside of the United States and whose
parents were not United States citizens at the time is not a
natural-born United States citizen.
(5) Schwarzenegger is not a natural-born United States citizen.
(from 3, 4*, and 4**)
This much of the argument is now sound.
An argument with a single premise has grown to include three stages
with at least four suppressed premises. Some of the added premises are ob-
vious, but others are less well known, so we cannot assume that the person
who gave the original argument had the more complete argument in mind.
Many people would be convinced by the original argument even without all
these added complexities. Nonetheless, the many suppressed premises are
necessary to make the argument sound. Seeing this brings out the assump-
tions that must be true for the conclusion to follow from the premises. This
process of making everything explicit enables us to assess these background
assumptions directly.
There is one obscure exception to the premise that only a natural-born citizen
may become president of the United States. The Constitution does allow a
person who is not a natural-born citizen to become president if he or she was
“a citizen of the United States at the time of the adoption of this Constitution.”
This exception is said to have been added to allow Alexander Hamilton to run
for president, but it obviously does not apply to Schwarzenegger or to anyone
else alive today. Nonetheless, this exception keeps the argument from being
sound in its present form. Reformulate the final stage of the argument to make
it sound.
Exercise V
LINGUISTIC PRINCIPLES
Often an argument is valid, but it is still not clear why it is valid. It is not clear
how the conclusion follows from the premises. Arguments are like pathways
between premises and conclusions, and some of these pathways are more
complicated than others. Yet even the simplest arguments reveal hidden
120
CHAPTER 5 ■ Deep Analys is
complexities when examined closely. For example, there is no question that
the following argument is valid:
(1) Harriet is in New York with her son.
(2) Harriet’s son is in New York.
It is not possible for the premise to be true and the conclusion false. If asked
why this conclusion follows from the premise, it would be natural to reply that:
You cannot be someplace with somebody unless that person is there, too.
This is not something we usually spell out, but it is the principle that takes
us from the premise to the conclusion.
One thing to notice about this principle is that it is quite general—that is,
it does not depend on any special features of the people or places involved. It
is also true that if Benjamin is in St. Louis with his daughter, then Benjamin’s
daughter is in St. Louis. Although the references have changed, the general
pattern that lies behind this inference will seem obvious to anyone who
understands the words used to formulate it. For this reason, principles of this
kind are basically linguistic in character.
If we look at arguments as they occur in everyday life, we will discover that
almost all of them turn on unstated linguistic principles. To cite just one more
example: Alice is taller than her husband, so there is at least one woman who
is taller than at least one man. This inference relies on the principles that hus-
bands are men and wives are women. We do not usually state these linguistic
principles, for to do so will often violate the rule of Quantity. (Try to imagine a
context in which you would come right out and say, “Husbands, you know,
are men.” Unless you were speaking to someone just learning the language,
this would be a peculiar remark.) Nonetheless, even if it would usually be pe-
culiar to come right out and state such linguistic principles, our arguments
still typically presuppose them. This observation reveals yet another way in
which our daily use of language moves within a rich, though largely unno-
ticed, framework of linguistic rules, as we emphasized in Chapter 2.
EVALUATIVE SUPPRESSED PREMISES
We have examined two kinds of suppressed premises, factual and linguistic.
Many arguments also contain unstated evaluative premises. As we saw in
Chapter 3, evaluation comes in many kinds. The following argument involves
moral evaluation:
It is immoral to buy pornography, because pornography leads to
violence toward women.
This argument clearly relies on the moral principle that it is immoral to buy
anything that leads to violence toward women. A different example contains
religious premises:
You shouldn’t take the name of the Lord in vain, for this shows disrespect.
121
Suppressed Prem ises
The suppressed premise here is that you should not do anything that
shows disrespect (to the Lord). One more example is about economics:
It is unwise to invest all of your money in one stock, since this increases
the risk that you will lose everything.
The suppressed premise here is that it is unwise to increase the risk that you
will lose everything. More examples could be given, but the point should be
clear. Most arguments depend on unstated assumptions, and many of these
assumptions are evaluative in one way or another.
USES AND ABUSES OF SUPPRESSED PREMISES
Talk about suppressed premises may bring to mind suppressing a rebellion
or an ugly thought, and using hidden premises may sound somewhat
sneaky. However, the way we are using them, these expressions do not
carry such negative connotations. A suppressed or hidden premise is sim-
ply an unstated premise. It is often legitimate to leave premises unstated. It
is legitimate if (1) those who are given the argument can easily supply these
unstated premises for themselves, and (2) the unstated premises are not
themselves controversial. If done properly, the suppression of premises can
add greatly to the efficiency of language. Indeed, without the judicious
suppression of obvious premises, many arguments would become too cum-
bersome to be effective.
On the other hand, suppressed premises can also be used improperly.
People sometimes suppress questionable assumptions so that their
opponents will not notice where an argument goes astray. For example,
when election debates turn to the topic of crime, we often hear arguments
like this:
My opponent is opposed to the death penalty, so he must be soft on
crime.
The response sometimes sounds like this:
Since my opponent continues to support the death penalty, he must not
have read the most recent studies, which show that the death penalty
does not deter crime.
The first argument assumes that anyone who is opposed to the death penalty
is soft on crime, and the second argument assumes that anyone who read the
studies in question would be convinced by them and would turn against
the death penalty. Both of these assumptions are questionable, and the ques-
tions they raise are central to the debate. If we want to understand these
issues and address them directly, we have to bring out these suppressed
premises openly.
122
CHAPTER 5 ■ Deep Analys is
The following arguments depend for their validity on suppressed premises of
various kinds. For each of them, list enough suppressed premises to make the
argument valid and also to show why it is valid. This might require several
suppressed premises of various kinds.
Example: Carol has no sisters, because all her siblings are
brothers.
Suppressed Premises: A sister would be a sibling.
A brother is not a sister.
1. Britney Spears is under age thirty-five. Therefore, she cannot run for
president of the United States.
2. Nixon couldn’t have been president in 1950 because he was still in the Senate.
3. 81 is not a prime number, because 81 is divisible by 3.
4. There’s no patient named Rupert here; we have only female patients.
5. Columbus did not discover the New World because the Vikings explored
Newfoundland centuries earlier.
6. There must not be any survivors, since they would have been found by now.
7. Lincoln could not have met Washington, because Washington was dead
before Lincoln was born.
8. Philadelphia cannot play Los Angeles in the World Series, since they are
both in the National League.
9. Mildred must be over forty-three, since she has a daughter who is thirty-
six years old.
10. He cannot be a grandfather because he never had children.
11. That’s not modern poetry; you can understand it.
12. Harold can’t play in the Super Bowl, because he broke his leg.
13. Shaquille must be a basketball player, since he is so tall.
14. Dan is either stupid or very cunning, so he must be stupid.
15. Susan refuses to work on Sundays, which shows that she is lazy and
inflexible.
16. Jim told me that Mary is a professor, so she can’t be a student, since
professors must already have degrees.
17. This burglar alarm won’t work unless we are lucky or the burglar uses the
front door, so we can’t count on it.
18. His natural talents were not enough; he still lost the match because he had
not practiced sufficiently.
Exercise VI
THE METHOD OF RECONSTRUCTION
We can summarize the discussion so far by listing the steps to be taken in re-
constructing an argument. The first two steps were discussed in Chapters 4
and 3, respectively.
123
The Method of Reconstruct ion
1. Do a close analysis of the passage containing the argument.
2. List all explicit premises and the conclusion in standard form.
3. Clarify the premises and the conclusion where necessary.
4. Break up the premises and the conclusion into smaller parts where this
is possible.
5. Arrange the parts of the argument into a chain or tree of subarguments
where this is possible.
6. Assess each argument and subargument for validity.1
7. If any argument or subargument is not valid, or if it is not clear why
it is valid, add suppressed premises that will show how to get from the
premises to the conclusion.
8. Assess the truth of the premises.
Remember that the goal of reconstruction is not just technical validity but is,
instead, to understand why and how the conclusion is supposed to follow
from the premises.
After reconstructing the argument, it is often helpful to add some indica-
tion of its structure. This can be done by numbering the premises and then,
after each conclusion, listing the premises from which that conclusion
follows. (We did this in our examples.) The argument’s structure can also be
shown by a diagram like those discussed above. Either way, we need to
make it clear exactly how the separate parts of the argument are supposed
to fit together.
This method is not intended to be mechanical. Each step requires care and
intelligence. As a result, a given argument can be reconstructed in various
ways with varying degrees of illumination and insight. The goal of this
method is to reveal as much of the structure of an argument as possible and
to learn from it as much as you can. Different reconstructions approach this
goal more or less closely.
The whole process is more complex than our discussion thus far has sug-
gested. This is especially clear in the last three steps of reconstruction, which
must be carried out simultaneously. In deciding whether an argument is
acceptable, we try to find a set of true suppressed premises that, if added
to the stated premises, yields a sound argument for the conclusion. Two
problems typically arise when we make this effort:
1. We find a set of premises strong enough to support the conclusion, but
at least one of these premises is false.
2. We modify the premises to avoid falsehood, but the conclusion no
longer follows from them.
The reconstruction of an argument typically involves shifting back and forth
between the demand for a valid argument and the demand for true prem-
ises. Eventually, either we show the argument to be sound or we abandon
the effort. In the latter case, we conclude that the argument in question has
no sound reconstruction. It is still possible that we were at fault in not
124
CHAPTER 5 ■ Deep Analys is
finding a reconstruction that showed the argument to be sound. Perhaps we
did not show enough ingenuity in searching for a suppressed premise that
would do the trick. There is, in fact, no purely formal or mechanical way of
dealing with this problem. A person presenting an argument may reason-
ably leave out steps, provided that they can easily be filled in by those to
whom the argument is addressed. So, in analyzing an argument, we should
be charitable, but our charity has limits. After a reasonable search for those
suppressed premises that would show the argument to be sound, we should
not blame ourselves if we fail to find them. Rather, the blame shifts to the
person who formulated the argument for not doing so clearly.
Reconstruct and diagram the main arguments in:
1. The passages in the Discussion Questions at the end of Chapter 1.
2. The passages in Exercises I and V in Chapter 4.
3. An editorial from your local paper.
4. Your last term paper or a friend’s last term paper.
5. Part of one of the readings in Part V.
Exercise VII
Not all arguments are serious or good. The following silly argument comes
from a famous scene in Monty Python and the Holy Grail. Reconstruct the argu-
ment that is supposed to show that the woman is a witch.
Crowd: We have found a witch. May we burn her? . . .
Woman: I’m not a witch! I’m not a witch! . . .
Leader: What makes you think she is a witch?
Man #1: She turned me into a newt!
Leader: A newt?
Man #1: I got better.
Crowd: Burn her anyway!
Leader: Quiet! Quiet! There are ways of telling whether she is a witch.
Crowd: Are there? What are they? Tell us. Do they hurt?
Leader: Tell me, what do you do with witches?
Crowd: Burn them!
Leader: And what do you burn apart from witches?
Man #2: More witches!
Man #3: Wood.
Leader: So, why do witches burn?
Exercise VIII
125
Digg ing Deeper
Man #1: ’Cause they’re made of wood.
Leader: Good! . . . So, how do we tell whether she is made of wood?
Crowd: Build a bridge out of her.
Leader: Ah, but can you not also make bridges out of stone?
Crowd: Oh yeah.
Leader: Does wood sink in water?
Crowd: No, it floats. Throw her into the pond!
Leader: What also floats in water?
Crowd: Bread. Apples. Very small rocks. Cider! Great gravy. Cherries.
Mud. Churches. Lead.
Arthur: A duck!
Leader: Exactly. So, logically, —
Man #3: If she weighs the same as a duck, she’s made of wood.
Leader: And therefore?
Crowd: A witch! . . . A duck. A duck. Here’s a duck!
Leader: We shall use my largest scales.
Crowd: Burn the witch! (Woman is placed on scales opposite a duck.)
Leader: Remove the supports. (Woman balances duck.)
Crowd: A witch!
Woman: It’s a fair cop.
DIGGING DEEPER
After we have reconstructed an argument as well as we can, doubts still might
arise about its premises. If we agree with its premises, others might deny them
or ask why they are true. If we disagree with its premises, we might be able to
understand the source of our disagreement better if we determine why those
premises are believed by other people—including the person who gave the ar-
gument. Either way, it is useful to try to construct supporting arguments for
the premises that may be questioned. These supporting arguments are not
parts of the explicit argument or even of its reconstruction, but are arguments
further back in a chain of arguments.
When we look for further arguments to support the premises in an argu-
ment, we might then wonder whether the premises of each new argument can
be accepted without supporting arguments as well. We seem faced with the
unpleasant task of producing endless chains of argument. (A similar problem
was mentioned in Chapter 3.) When pressed in this way to give reasons for
our reasons, and reasons for our reasons for our reasons, we eventually come
upon fundamental principles—principles for which we cannot give any more
basic argument. These fundamental principles can concern morality, religion,
126
CHAPTER 5 ■ Deep Analys is
politics, or our general views concerning the nature of the world. We often ar-
gue within a framework of such principles without actually stating them, be-
cause we assume (sometimes incorrectly) that others accept them as well.
When someone argues that environmental destruction should be stopped
because it will lead to the annihilation of the human race, he or she will
not feel called on to say explicitly, “And the annihilation of the human race
would be a very bad thing.” That, after all, is something that most people take
for granted.
Though fundamental principles are often obvious and generally accepted,
at times it is not clear which principles are being assumed and just how accept-
able they really are. Then we need to make our assumptions explicit and look
for deeper arguments, continuing the process as far as we can. There is a limit
to how far we can go, but the deeper we go, the better we will understand both
our own views and the views of our opponents.
The following arguments depend on suppressed premises for their validity.
(a) State what these underlying premises might be. In some cases, there might
be more than one. (b) Indicate whether these premises are fundamental princi-
ples in the sense just described. (c) For each premise that is not fundamental,
give a supporting argument that is as plausible as you can make it. Remember,
you do not have to accept an argument to detect its underlying principles and
to understand the kind of argument that could be used to support it.
Example: General Snork has no right to rule, because he
came to power by a military coup.
Suppressed Premise: Someone who came to power by a military coup
has no right to rule.
Supporting Argument: Someone has a right to rule only if he or she has
been elected by the people.
Someone who comes to power by a military coup
has not been elected by the people.
Someone who came to power by a military coup
has no right to rule.
(Further support for the first premise might be provided by some general theory
of democracy, such as that all rights to use power come from the people, and the
only legitimate way for the people to delegate these rights is in elections.)
1. You shouldn’t call Kirk guilty, because he has not even been tried yet.
2. Cows cannot live in a desert, because they eat grass.
3. The liquid in this glass must not be water, since the sugar I put in it isn’t
dissolving.
4. People can vote and be drafted at age eighteen, so they should also be al-
lowed to drink at eighteen.
Exercise IX
127
An Example of Deep Analys is : Cap ital Pun ishment
5. We have no right to attack Muslim dictatorships if we support Christian
dictatorships. So we should not attack Muslim dictatorships.
6. That chair won’t hold you, since it almost broke a few minutes ago when
your little sister sat in it.
7. Bringing down our deficits should be a high priority, because inflation
will return if high deficits continue.
8. The thought of eating ostrich meat will seem strange to most people, so
your ostrich farm is bound to go broke.
9. Getting good grades must be hard, since, if it were easy, more people
would get good grades.
10. Morris does not deserve his wealth, for he merely inherited it.
11. Frank should not be punished, because it is wrong to punish someone just
to make an example of him.
12. There can’t be UFOs (unidentified flying objects), because there is no life
on other planets.
13. The sky is red tonight, so it isn’t going to rain tomorrow.
14. Parents take care of their children when they are young, so their children
should take care of them when they get old.
How can you tell when you have reached a fundamental principle? Must
every argument start with some basic claim for which no further argument can
be given or for which you can give no argument? Why or why not?
Discussion Question
AN EXAMPLE OF DEEP ANALYSIS:
CAPITAL PUNISHMENT
We can illustrate these methods of deep analysis by examining the difficult
question of the constitutionality of capital punishment. It has been argued
that the Supreme Court should declare the death penalty unconstitutional
because it is a cruel and unusual punishment. The explicitly stated argu-
ment has the following basic form:
(1) The death penalty is a cruel and unusual punishment.
(2) The death penalty should be declared unconstitutional. (from 1)
The argument plainly depends on two suppressed premises:
SP1: The Constitution prohibits cruel and unusual punishments.
SP2: Anything that the Constitution prohibits should be declared
unconstitutional.
128
CHAPTER 5 ■ Deep Analys is
So the argument, more fully spelled out, looks like this:
(1) The death penalty is a cruel and unusual punishment.
(2) SP: The Constitution prohibits cruel and unusual punishments.
(3) The Constitution prohibits the death penalty. (from 1–2)
(4) SP: Anything that the Constitution prohibits should be declared
unconstitutional.
(5) The death penalty should be declared unconstitutional. (from 3–4)
This reconstruction seems to be a fair representation of the intent of the
original argument.
We can now turn to an assessment of this argument. First, the argument
is valid: Given the premises, the conclusion does follow. All that remains is
to determine the truth of the premises one by one.
Premise 4 seems uncontroversial. Indeed, it might sound like a truism to
say that anything that violates a constitutional provision should be declared
unconstitutional. But in fact, this notion was once controversial, for nothing
in the Constitution explicitly gives the courts the right to declare acts of legis-
lators unconstitutional and hence void. The courts have acquired and consol-
idated this right in the years since 1789, and it is still sometimes challenged
by those who think that it gives the courts too much power. But even if the
judiciary’s power to declare laws unconstitutional is not itself a constitution-
ally stated power, it is so much an accepted part of our system that no one
would challenge it in a courtroom proceeding today.
The second premise is clearly true, for the Constitution does, in fact,
prohibit cruel and unusual punishments. Its Eighth Amendment reads,
“Excessive bail shall not be required, nor excessive fines imposed, nor cruel
and unusual punishments inflicted.” It is not clear, however, just what this
prohibition amounts to. In particular, does the punishment have to be both
cruel and unusual to be prohibited, or is it prohibited whenever it is either
cruel or unusual? This would make a big difference if cruel punishments
were usual, or if some unusual punishments were not cruel. For the moment,
let us interpret the language as meaning “both cruel and unusual.”
The first premise—“The death penalty is a cruel and unusual punish-
ment“—obviously forms the heart of the argument. What we would expect,
then, is a good supporting argument to be put forward on its behalf. The
following argument by Supreme Court Justice Potter Stewart (in Furman v.
Georgia, 408 U.S. 239 at 309–310 [1972]) was intended to support this claim in
particular cases in which the death penalty was imposed for rape and murder:
In the first place, it is clear that these sentences are “cruel” in the sense that they
excessively go beyond, not in degree but in kind, the punishments that the state
legislatures have determined to be necessary. . . . In the second place, it is equally
clear that these sentences are “unusual” in the sense that the penalty of death
is infrequently imposed for murder, and that its imposition for rape is extraordi-
narily rare. But I do not rest my conclusion upon these two propositions alone.
These death sentences are cruel and unusual in the same way that being struck
129
An Example of Deep Analys is : Cap ital Pun ishment
by lightning is cruel and unusual. For, of all the people convicted of rapes and
murders in 1967 and 1968, many just as reprehensible as these, the petitioners
are among a capriciously selected random handful upon whom the sentence of
death has in fact been imposed. My concurring brothers [the Justices who agree
with Stewart] have demonstrated that, if any basis can be discerned for the selec-
tion of these few to be sentenced to die, it is the constitutionally impermissible
basis of race.
The first sentence argues that the death penalty is cruel. The basic idea is that
punishments are cruel if they inflict harms that are much worse than what is
necessary for any legitimate and worthwhile purpose. Stewart then seems to
accept the state legislatures’ view that the death penalty does go far beyond
what is necessary. This makes it cruel.
Now let us concentrate on the part of this argument intended to show that
the death penalty is an unusual punishment. Of course, in civilized nations,
the death penalty is reserved for a small range of crimes, but this is hardly the
point at issue. The point of the argument is that the death penalty is unusual
even for those crimes that are punishable by death, including first-degree
murder. Moreover, Stewart claims that, among those convicted of crimes
punishable by death, who actually receives a death sentence is determined
either capriciously or on the basis of race. The point seems to be that whether
a person who is convicted of a capital crime will be given the death penalty
depends on the kind of legal aid he or she receives, the prosecutor’s willing-
ness to offer a plea bargain, the judge’s personality, the beliefs and attitudes
of the jury, and many other considerations. At many points in the process,
choices that affect the outcome could be based on mere whim or caprice,
or even on the race of the defendant or the victim. Why are these factors
mentioned? Because, as Stewart says, it is unconstitutional for sentencing to
be based on caprice or race.
We can then restate this supporting argument more carefully:
(1) Very few criminals who were found guilty of crimes that are
punishable by death are actually sentenced to death.
(2) Among those found guilty of crimes punishable by death, who is
sentenced to death depends on caprice or race.
(3) It is unconstitutional for sentencing to depend on caprice or race.
(4) A punishment is unusual if it is imposed infrequently and on an
unconstitutional basis.
(5) The death penalty is an unusual punishment. (from 1–4)
This conclusion is part of the first premise in our original argument. Now
we can spread the entire argument out before us:
(1) An act is cruel if it inflicts harms that are much worse than what is
necessary for any legitimate and worthwhile purpose.
(2) The death penalty inflicts harms that are much worse than what is
necessary for any legitimate and worthwhile purpose.
(3) The death penalty is cruel. (from 1–2)
130
CHAPTER 5 ■ Deep Analys is
(4) Very few criminals who were found guilty of crimes that are
punishable by death are sentenced to death.
(5) Among those found guilty of crimes punishable by death, who is
sentenced to death depends on caprice or race.
(6) It is unconstitutional for sentencing to depend on caprice or race.
(7) A punishment is unusual if it is imposed infrequently and on an
unconstitutional basis.
(8) The death penalty is an unusual punishment. (from 4–7)
(9) The death penalty is both cruel and unusual. (from 3 and 8)
(10) The Constitution prohibits cruel and unusual punishments.
(11) The Constitution prohibits the death penalty. (from 9–10)
(12) Anything that the Constitution prohibits should be declared
unconstitutional.
(13) The death penalty should be declared unconstitutional. (from 11–12)
These propositions provide at least the skeleton of an argument with some
force. The conclusion does seem to follow from the premises, and the prem-
ises themselves seem plausible. We have produced a charitable reconstruc-
tion of the argument.
We can now see how an opponent might respond to it. One particularly
probing objection goes like this:
It is sadly true that caprice and race sometimes determine who, among those
found guilty of crimes punishable by death, is given the death sentence.
However, this fact reflects badly not on the law but on its administration.
If judges and juries met their obligations, these factors would not affect who
receives the death penalty, and this punishment would no longer be unusual
in any relevant sense. What is needed, then, is judicial reform and not the
removal of the death penalty on constitutional grounds.
This response is probing because it insists on a distinction between a law it-
self and the effects of its application—or, more pointedly, its misapplication.
Because this distinction was not drawn in the argument above, it is not clear
which premise is denied in this response. Probably the best interpretation is
that this response denies premise 7, because it is not the death penalty itself
that is unusual in the relevant sense when the conditions in premise 7 are
met. Instead, it is the present administration of the death penalty that is
problematic.
To meet this objection, the original argument could be strengthened in the
following way:
A law should not be judged in isolation from the likely effects of implementing it.
Because of the very nature of our system of criminal justice, for the foreseeable
future, the death penalty will almost certainly continue to be applied in a capri-
cious and racially discriminatory manner, and who receives it will be determined
partly by factors that the Constitution forbids as a basis for sentencing. The death
penalty will therefore remain an unusual punishment, so it should be declared
unconstitutional.
131
An Example of Deep Analys is : Cap ital Pun ishment
This argument suggests ways to avoid the above objection by strengthening
premises 5–7 of the above argument. The new versions of these premises can
be spelled out in the following way:
(5*) It is very likely in the foreseeable future that, among those found
guilty of crimes punishable by death, who is given the death
sentence will continue to depend on either caprice or race.
(6) It is unconstitutional for sentencing to depend on caprice or race.
(7*) A punishment is unusual if it is very likely in the foreseeable
future that it will continue to be imposed infrequently and on
an unconstitutional basis.
(8) The death penalty is an unusual punishment. (from 5*–7*)
Of course, an opponent can still respond that these premises are false if he
or she can show that there is some way to avoid the problems that are raised
by premise (5*). This will not be easy to show, however, if the argument is
right about “the very nature of our system of criminal justice.”
Another kind of question is raised by premise (7*). Should a law be
declared unconstitutional whenever there is a good chance that it will
be abused in ways that infringe on constitutional rights? Many laws have this
potential—for example, all laws involving police power. This is why certain
police powers have been limited by court rulings. Strict rules governing in-
terrogations and wiretaps are two results. But only an extremist would sug-
gest that we should abolish all police powers because of the inevitable risk of
unconstitutional abuse. Accordingly, those who argue in favor of the death
penalty might try to show that the problems in the application of the death
penalty are not sufficiently important to be constitutionally intolerable.
The supporter of the death penalty also might take a very different tack:
Those who argue against the constitutionality of the death penalty on the grounds
that it is a cruel and unusual punishment use the expression “cruel and unusual” in a
way wholly different from that intended by the framers of the Eighth Amendment.
By “cruel” they had in mind punishments that involved torture. By “unusual” they
meant bizarre or ghoulish punishments of the kind that often were part of public
spectacles in barbaric times. Modern methods of execution are neither cruel nor un-
usual in the constitutionally relevant senses of these words. Therefore, laws demand-
ing the death penalty cannot be declared unconstitutional on the grounds that they
either directly or indirectly involve a punishment that is cruel and unusual.
The core of this counterargument can be expressed as follows:
(1) In appeals to the Constitution, its words should be taken as they
were originally intended.
(2) Modern methods of carrying out a death penalty are neither
“cruel” nor “unusual” if these words are interpreted as they
were originally intended.
(3) The death penalty should not be declared unconstitutional on the
ground that it violates the prohibition against cruel and unusual
punishments. (from 1–2)
132
CHAPTER 5 ■ Deep Analys is
The second premise of this argument states a matter of historical fact that
might not be altogether easy to verify. Chances are, however, that it comes
close to the truth, because the Constitution refers to the death penalty with-
out criticism, and the death penalty was rarely questioned at the time. Given
this, the opponent of the death penalty must either attack the first premise
or find some other grounds for holding that the death penalty should be de-
clared unconstitutional.
The first premise might seem like a truism, for how can a document
guide conduct if anyone can reinterpret its words regardless of what was
intended? Isn’t the literal meaning of the document simply what was meant
at the time, with everything else being interpretation? Of course, there are
times when it is not easy to discover what its meaning was. (In the present
case, for example, it is not clear whether the Eighth Amendment prohibits
punishments that are either cruel or unusual or only those that are both
cruel and unusual.) It seems unlikely, however, that those who drafted the
Eighth Amendment used either the word “cruel” or the word “unusual” in
the ways in which they are employed in the argument against the death
penalty.
Does this last concession end the debate in favor of those who reject the
anti–capital punishment argument that we have been examining? The argu-
ment certainly seems to be weakened, but there are those who would take
a bold course by simply denying the first premise of the argument used to
refute them. They would deny, that is, that we are bound to read the Constitu-
tion in the way intended by its framers. An argument in favor of this position
might look something like this:
The great bulk of the Constitution was written in an age almost wholly
different from our own. To cite just two examples: Women were denied
fundamental rights of full citizenship, and slavery was a constitutionally
accepted feature of national life. The Constitution has remained a live and
relevant document just because it has undergone constant reinterpretation.
So, even if it is true that the expression “cruel and unusual” meant something
quite special to those who framed the Eighth Amendment, plainly a humane
desire to make punishment more civilized lay behind it. The present reading
of this amendment is in the spirit of its original intention and simply makes it
applicable to our own times.
The argument has now moved to an entirely new level, one concerning
whether the Constitution should be read strictly in accord with the original
intentions of those who wrote it or more freely to accommodate modern
realities.
We shall not pursue the discussion further into these complex areas. In-
stead, we should consider how we were led into them. Recall that our
original argument did not concern the general question of whether capi-
tal punishment is morally right or wrong. The argument turned on a
much more specific point: Does the death penalty violate the prohibition
against cruel and unusual punishments in the Eighth Amendment to the
133
An Example of Deep Analys is : Cap ital Pun ishment
Constitution? The argument with which we began seemed to be a
straightforward proof that it does. Yet as we explored principles that lay
in back of this deceptively simple argument, the issue became broader
and more complex. We finally reached a point at which the force of the
original argument was seen to depend on what we consider the proper
way to interpret the Constitution—strictly or more freely. If we now go on
to ask which method of interpretation is best, we will have to look at the
role of the courts and, more generally, the purpose of government. Even-
tually we will come to fundamental principles for which we can give no
further argument.
In examining the question of the constitutionality of capital punishment,
we have had to compress a complicated discussion into a few pages. We
have only begun to show how the issues involved in this complex debate
can be sorted out and then addressed intelligently. There is, however, no
guarantee that these procedures, however far they are carried out, will even-
tually settle this or any other fundamental dispute. It is entirely possible that
the parties to a dispute may reach a point where they encounter a funda-
mental or rock-bottom disagreement that they cannot resolve. They simply
disagree and cannot conceive of any deeper principles that could resolve
their disagreement. But even if this happens, they will at least understand
the source of their disagreement. They will not be arguing at cross-purposes,
as so often happens in the discussion of important issues. Finally, even
if they continue to disagree, they may come to appreciate that others
may view things quite differently from the way they do. This may in turn
help them deal with their basic disagreements in an intelligent, humane, and
civilized way.
What is the best argument that Justice Stewart could give in support of the
premise that the death penalty “excessively go[es] beyond” what is necessary
for any legitimate and worthwhile purpose? Is this argument adequate to jus-
tify this premise? (For one such argument, see Justice Brennan’s opinion in
Furman v. Georgia.)
Exercise X
How could one best argue in support of premise (5*), that it is very likely in
the foreseeable future that, among those found guilty of crimes punishable by
death, who is given the death sentence will continue to depend on either
caprice or race? How could defenders of the death penalty try to refute this
argument?
Exercise XI
134
CHAPTER 5 ■ Deep Analys is
Formulate the best argument you can in support of the premise that, in appeals
to the Constitution, its words should be taken as they were originally intended.
Is this argument adequate to justify this premise? Why or why not?
Exercise XII
The final argument in our examination of whether or not the death penalty vio-
lates the Constitution attempts to show that the Constitution must be read in a
free or liberal way that makes it relevant to present society. Filling in suppressed
premises where necessary, restate this argument as a sequence of explicit steps.
After you have given the argument the strongest restatement you can, evaluate
it for soundness.
Exercise XIII
To solve a mystery, you need to determine which facts are crucial and then
argue from those facts to a solution. Solve the following mysteries and recon-
struct your own argument for your solution. These stories come from Five-
Minute Whodunits, by Stan Smith (New York: Sterling, 1997). The first passage
introduces our hero:
Even those acquainted with Thomas P. Stanwick are often struck by his ap-
pearance. A lean and lanky young man, he stands six feet two inches tall.
His long, thin face is complemented by a full head of brown hair and a droopy
mustache. Though not husky in build, he is surprisingly strong and enjoys
ruggedly good health. His origins and early life are obscure. He is undeniably
well educated, however, for he graduated with high honors from Dartmouth
College as a philosophy major. . . .
MYSTERY 1: A MERE MATTER OF DEDUCTION
Thomas P. Stanwick, the amateur logician, removed a pile of papers from
the extra chair and sat down. His friend Inspector Matthew Walker had just
returned to his office from the interrogation room, and Stanwick thought he
looked unusually weary.
“I’m glad you dropped by, Tom,” said Walker. “We have a difficult case on
hand. Several thousand dollars’ worth of jewelry was stolen from Hoffman’s
Jewel Palace yesterday morning. From some clues at the scene and a few
handy tips, we have it narrowed down to three suspects: Addington, Burke,
and Chatham. We know that at least one of them was involved, and possibly
more than one.”
Exercise XIV
135
An Example of Deep Analys is : Cap ital Pun ishment
“Burke has been suspected in several other cases, hasn’t he?” asked Stanwick
as he filled his pipe.
“Yes, he has,” Walker replied, “but we haven’t been able to nail him yet.
The other two are small potatoes, so what we really want to know is whether
Burke was involved in this one.”
“What have you learned about the three of them?”
“Not too much. Addington and Burke were definitely here in the city
yesterday. Chatham may not have been. Addington never works alone, and
carries a snub-nosed revolver. Chatham always uses an accomplice, and he
was seen lurking in the area last week. He also refuses to work with Adding-
ton, who he says once set him up.”
“Quite a ragamuffin crew!” Stanwick laughed. “Based on what you’ve said,
it’s not too hard to deduce whether Burke was involved.”
Was Burke involved or not?
MYSTERY 2: MURDER IN A LONDON FLAT
Lord Calinore was gunned down in his London flat by a robber, who then
ransacked the flat. The case was placed in the capable hands of Inspector
Gilbert Bodwin of Scotland Yard. Bodwin’s investigation revealed that one
man had planned the crime, another had carried it out, and a third had acted
as lookout.
Bodwin discussed the case at length one evening over dinner at his club with
an old friend, Thomas P. Stanwick, the amateur logician, visiting from America.
“It’s quite a case,” Stanwick remarked. “Have you any suspects?”
Bodwin sliced his roast beef with relish. “Yes, indeed. Four. We have con-
clusive evidence that three of those four were responsible for the crime.”
“Really! That’s remarkable progress. What about the fourth?”
“He had no prior knowledge of the crime and is completely innocent. The
problem is that we’re not sure which of the four are the planner, the gunman, the
lookout, and the innocent bystander.”
“I see.” Stanwick took more Yorkshire pudding. “What do you know about
them at this point?”
“Well, the names of the four are Merrick, Cross, Llewellyn, and Halifax.
Halifax and Cross play golf together every Saturday. They’re an odd pair!
Halifax can’t drive, and Cross has been out of Dartmoor Prison for only a
year.”
“What was he in for?”
“Forgery. We know that Merrick and Halifax kept the flat under surveil-
lance for several days just before the day the crime was committed, the 17th.
Llewellyn and Merrick, with their wives, had dinner together on the Strand
on the 12th.”
“An interesting compilation,” said Stanwick, “but hardly conclusive. Is that
all of it?”
(continued)
136
CHAPTER 5 ■ Deep Analys is
“Not quite. We know that the gunman spent the week before the crime in
Edinburgh, and that the innocent bystander was acquainted with the plan-
ner and the gunman, but not with the lookout.”
“That is very helpful,” said Stanwick with a smile. He raised his wine
glass. “Bodwin, old fellow, your case is complete.”
Who are the planner, the gunman, and the lookout?
MYSTERY 3: A THEFT AT THE ART MUSEUM
The theft of several valuable paintings from the Royston Art Museum created
a sensation throughout New England. Two days after Stanwick’s return from
a visit to Scotland, he was visited by Inspector Matt Walker, who was in charge
of the case. As Stanwick poured tea, Walker quickly brought his friend up to
date on the case.
“We’ve identified the gang of five thieves who must have done this job,”
Walker reported. “Archie McOrr, who never finished high school, is married
to another one of the five, Charlayne Trumbull. The other three are Beverly
Cuttle, Ed Browning, and Douglas Stephens.”
“I thought you told me earlier that only four people were involved in the
robbery,” said Stanwick.
“That’s right. One stayed in the car as the driver, another waited outside and
acted as lookout, and two others entered the museum and carried out the actual
theft. One of the five gang members was not involved in this particular job at all.”
“And the question, I hope,” said Stanwick with a smile, “is who played
what part, if any, in the theft.”
“Exactly.” Walker flipped open his notebook. “Though I’m glad to say that
our investigation is already bearing some fruit. For example, we have good
reason to believe that the lookout has a Ph.D. in art history, and that the
driver was first arrested less than two years ago.”
“A remarkable combination,” Stanwick chuckled.
“Yes, indeed. We know that Douglas was on the scene during the robbery.
One of the actual thieves (who entered the museum) is the sister of Ed Brown-
ing. The other thief is either Archie or his wife.”
“What else do you have on Douglas?” asked Stanwick.
“Not much. Although he’s never learned to drive, he used to be a security
guard at the Metropolitan Museum of Art in New York.”
“Interesting. Please go on.”
“The rest is mainly odds and ends.” Walker thumbed through a few more
pages of notes. “Charlayne, an only child, is very talented on the saxophone. Bev-
erly and Ed both have criminal records stretching back a decade or more. We’ve
also learned that the driver has a brother who is not a member of the gang.”
“Most interesting indeed,” remarked Stanwick. He handed Walker a mug
of tea and sat down with his own. “Your investigation has made excellent
progress. So much, in fact, that you already have enough to tell who the
thieves, the lookout, and the driver are.”
Who are they?
137
An Example of Deep Analys is : Cap ital Pun ishment
MYSTERY 4: TRIVIA AND SIGNIFICA
“For April, this is starting out to be a pretty quiet month,” remarked Inspector
Walker as he rummaged in his desk drawer for a cigar.
Thomas P. Stanwick, the amateur logician, finished lighting his pipe and
leaned back in his chair, stretching his long legs forward.
“That is indeed unusual,” he said. “Spring usually makes some young fan-
cies turn to crime. The change is welcome.”
“Not that we police have nothing to do.” Walker lit his cigar. “A couple of
the youth gangs, the Hawks and the Owls, have been screeching at each other
lately. In fact, we heard a rumor that they were planning to fight each other
this Wednesday or Thursday, and we’re scrambling around trying to find out
whether it’s true.”
“The Hawks all go to Royston North High, don’t they?” asked Stanwick.
“That’s right. The Owls are the street-smart dropouts who hang out at
Joe’s Lunch Cafe on Lindhurst. You know that only those who eat at Joe’s col-
lect green matchbooks?”
Stanwick blinked and smiled. “I beg your pardon?”
“That’s right.” Walker picked up a few papers from his desk. “That’s the
sort of trivia I’m being fed in my reports. Not only that, but everyone at Roys-
ton North High wears monogrammed jackets. What else have I got here? Only
kids who hang out on Laraby Street fight on weekdays. Laraby is three blocks
from Lindhurst. The Hawks go out for pizza three times a week.”
“Keep going,” chuckled Stanwick. “It’s wonderful.”
“A hog for useless facts, eh? No one who eats at Joe’s wears a mono-
grammed jacket. The Owls elect a new leader every six months, the Hawks
every year. Elections! Furthermore, everyone who hangs out on Laraby
Street collects green matchbooks. Finally, the older (but not wiser) Owls buy
beer at Johnny’s Package Store.”
Stanwick laughed heartily. “Lewis Carroll,” he said, “the author of Symbolic
Logic and the ‘Alice in Wonderland’ books, taught logic at Oxford, and he used
to construct soriteses, or polysyllogisms [that is, chains of categorical syllo-
gisms], out of material like that. In fact, his were longer and much wilder and
more intricate, but of course they were fiction.
“As it is, the information you’ve cited should ease your worries. Those
gangs won’t get together to fight until at least Saturday.”
How does he know?
NOTE
1 We assess inductive arguments for strength instead of validity, but here we focus on deductive
arguments. Inductive arguments will be examined in Part III, Chapters 8–12.
How to Evaluate
Arguments: Deductive
Standards
After isolating, laying out, and filling in an argument, the next step is to determine
whether that uncovered argument is any good. This assessment, like other evalua-
tions, requires standards. There are two main standards for evaluating arguments:
the deductive standard of validity and the inductive standard of strength. Part II
(which includes Chapters 6 and 7) will investigate the deductive standard of valid-
ity. Part III (which includes Chapters 8–12) will then explore the inductive stan-
dard of strength.
We already saw in Chapter 3 that an argument is valid in our technical sense if
and only if it is not possible that its premises are true and its conclusion false. That
standard sounds simple, but it is not so easy to say how to determine whether this
combination of truth values is or is not possible in a particular case. Sometimes the
validity of an argument can be seen simply by looking at the premises and conclu-
sion viewed as whole propositions. That is the approach of propositional (or senten-
tial) logic, which is the topic of Chapter 6. Another possibility is that the validity of
an argument can be seen only by looking inside premises and conclusions to their
parts, including their subjects and predicates. That is the approach of categorical (or
syllogistic) logic, which is the topic of Chapter 7.
These relatively simple examples of formal logic do not, of course, exhaust the
possibilities. There are many more kinds of formal logic. Many arguments remain
valid, even though their validity is not captured by either propositional or categori-
cal logic. That creates problems that we will face throughout Chapters 6 and 7. Still,
by exploring some simple ways in which arguments can be valid by virtue of their
form alone, we can gain greater insight into the nature of validity and, thereby, into
the standards for assessing arguments.
II
139
This page intentionally left blank
PROPOSITIONAL LOGIC
This chapter begins our investigation of evaluating arguments by means of formal
deductive logic. The first part of the chapter will show how the crucial standard of
validity, which was introduced in Chapter 3, can be developed rigorously in one
area—what is called propositional logic. This branch of logic deals with connectives
such as “and” and “or,” which allow us to build up compound propositions from
simpler ones. Throughout most of the chapter, the focus will be theoretical rather
than immediately practical. It is intended to provide insight into the concept of
validity by examining it in an ideal setting. The chapter will close with a discussion
of the relationship between the ideal language of symbolic logic and the language we
ordinarily speak.
THE FORMAL ANALYSIS OF ARGUMENTS
When we carry out an informal analysis of an argument, we pay close at-
tention to the key words used to present the argument and then ask our-
selves whether these key terms have been used properly. So far, we have
no exact techniques for answering the question of whether a word is used
correctly. We rely, instead, on linguistic instincts that, on the whole, are
fairly good.
In a great many cases, people can tell whether an argument marker, such
as “therefore,” is used correctly in indicating that one claim follows from an-
other. However, if we go on to ask the average intelligent person why one
claim follows from the other, he or she will probably have little to say except,
perhaps, that it is just obvious. In short, it is often easy to see that one claim
follows from another, but to explain why can be difficult. The purpose of this
chapter is to provide such an explanation for some arguments.
This quality of “following from” is elusive, but it is related to the techni-
cal notion of validity, which was introduced in Chapter 3. The focus of our
attention will be largely on the concept of validity. We are not, for the time be-
ing at least, interested in whether this or that argument is valid; we want to
understand validity itself. To this end, the arguments we will examine are so
simple that you will not be able to imagine anyone not understanding them
6
141
142
CHAPTER 6 ■ Propos it ional Log ic
at a glance. Who needs logic to deal with arguments of this kind? There is,
however, good reason for dealing with simple—trivially simple—arguments
at the start. The analytic approach to a complex issue is first to break it down
into subissues, repeating the process until we reach problems simple enough
to be solved. After these simpler problems are solved, we can reverse the
process and construct solutions to larger and more complex problems. When
done correctly, the result of such an analytic process may seem dull and
obvious—and it often is. The discovery of such a process, in contrast, often
demands the insight of genius.
The methods of analysis to be discussed here are formal in a specific way.
In Chapter 3, we gave the following argument as an example of a valid ar-
gument: “All Senators are paid, and Sam is a Senator, so Sam is paid.” The
point could have been made just as well with many similar examples:
(a) “All Senators are paid, and Sally is a Senator, so Sally is paid.” (b) “All
plumbers are paid, and Sally is a plumber, so Sally is paid.” (c) “All
plumbers are dirty, and Sally is a plumber, so Sally is dirty.” These argu-
ments are all valid (though not all are sound). Thus, we can change the per-
son we are talking about, the group that we say the person is in, and the
property that we ascribe to the person and to the group, all without affect-
ing the validity of the argument at all. That flexibility shows that the valid-
ity of this argument does not depend on the particular content of its
premises and conclusion. Instead, the validity of this argument results
solely from its form. Formal validity of this kind is what formal logics try
to capture.
BASIC PROPOSITIONAL CONNECTIVES
CONJUNCTION
The first system of formal logic that we will examine concerns proposi-
tional (or sentential) connectives. Propositional connectives are terms that
allow us to build new propositions from old ones, usually combining two
or more propositions into a single proposition. For example, given the
propositions “John is tall” and “Harry is short,” we can use the term “and”
to conjoin them, forming a single compound proposition: “John is tall and
Harry is short.”
Let us look carefully at the simple word “and” and ask how it functions.
“And” is a curious word, for it does not seem to stand for anything, at least
in the way in which a proper name (“Churchill”) and a common noun
(“dog”) seem to stand for things. Instead of asking what this word stands for,
we can ask a different question: What truth conditions govern this connective?
That is, under what conditions are propositions containing this connective
true? To answer this question, we imagine every possible way in which the
component propositions can be true or false. Then, for each combination, we
143
Bas ic Propos it ional Connect ives
decide what truth value to assign to the entire proposition. This may sound
complicated, but an example will make it clear:
John is tall. Harry is short. John is tall and Harry is short.
T T T
T F F
F T F
F F F
Here the first two columns cover every possibility for the component propo-
sitions to be either true or false. The third column states the truth value of
the whole proposition for each combination. Clearly, the conjunction of two
propositions is true if both of the component propositions are true; other-
wise, it is false.
Our reflections have not depended on the particular propositions in our
example. We could have been talking about dinosaurs instead of people, and
we still would have come to the conclusion that the conjunction of two propo-
sitions is true if both propositions are true, but false otherwise. This neglect of
the particular content of propositions is what makes our account formal.
To reflect the generality of our concerns, we can drop the reference to par-
ticular sentences altogether and use variables instead. Just as the lowercase
letters “x,” “y,” and “z” can be replaced by any numbers in mathematics, so
we can use the lowercase letters “p,” “q,” “r,” “s,” and so on as variables that
can be replaced by any propositions in logic. We will also use the symbol
“&” (called an ampersand) for “and.”
Consider the expression “p & q.” Is it true or false? There is obviously no
answer to this question. This is not because we do not know what “p” and
“q” stand for, for in fact “p” and “q” do not stand for any proposition at all.
Just as “x + y” is not any particular number in mathematics, so “p & q” is not
a proposition. Instead, “p & q” is a pattern for a whole series of propositions.
To reflect this, we will say that “p & q” is a propositional form. It is a pattern,
or form, for a whole series of propositions, including “John is tall and Harry
is short” as well as many other propositions.
To specify precisely which propositions have the form “p & q,” we need a
little technical terminology. The central idea is that we can pass from a propo-
sition to a propositional form by replacing propositions with propositional
variables.
Proposition Propositional Form
John is tall and Harry is short. p & q
When we proceed in the opposite direction by uniformly substituting
propositions for propositional variables, we get what we will call a substitu-
tion instance of that propositional form.
Propositional Form Substitution Instance
p & q Roses are red and violets are blue.
144
CHAPTER 6 ■ Propos it ional Log ic
Thus, “John is tall and Harry is short” and “Roses are red and violets are
blue” are both substitution instances of the propositional form “p & q.”
To get clear about these ideas, it is important to notice that “p” is also a
propositional form, with every proposition, including “Roses are red and
violets are blue,” among its substitution instances. There is no rule against
substituting compound propositions for propositional variables. Perhaps a
bit more surprisingly, our definitions allow “Roses are red and roses are
red” to be a substitution instance of “p & q.” This example makes sense if
you compare it to variables in mathematics. Using only positive integers,
how many solutions are there to the equation “x + y = 4“? There are three:
3 + 1, 1 + 3, and 2 + 2. The fact that “2 + 2” is a solution to “x + y = 4” shows
that “2” can be substituted for both “x” and “y” in the same solution.
That’s just like allowing “Roses are red” to be substituted for both “p” and
“q,” so that “Roses are red and roses are red” is a substitution instance of
“p & q” in propositional logic.
In general, then, we get a substitution instance of a propositional form by
uniformly replacing the same variable with the same proposition throughout,
but different variables do not have to be replaced with different propositions.
The rule is this:
Different variables may be replaced with the same proposition, but
different propositions may not be replaced with the same variable.
According to this rule:
“Roses are red and violets are blue” is a substitution instance of “p & q.”
“Roses are red and violets are blue” is also a substitution instance of “p.”
“Roses are red and roses are red” is a substitution instance of “p & q.”
“Roses are red and roses are red” is a substitution instance of “p & p.”
“Roses are red and violets are blue” is not a substitution instance of “p & p.”
“Roses are red” is not a substitution instance of “p & p.”
We are now in a position to give a perfectly general definition of conjunction
with the following truth table, using propositional variables where previously
we used specific propositions.
p q p & q
T T T
T F F
F T F
F F F
There is no limit to the number of propositions we can conjoin to form a new
proposition. “Roses are red and violets are blue; sugar is sweet and so are you”
is a substitution instance of “p & q & r & s.” We can also use parentheses to
145
Bas ic Propos it ional Connect ives
group propositions. This last example could be treated as a substitution instance
of “(p & q) & (r & s)“—that is, as a conjunction of two conjunctions. Later we will
see that, just as in mathematics, parentheses can make an important difference
to the meaning of a total proposition.
One cautionary note: The word “and” is not always used to connect two
distinct sentences. Sometimes a sentence has to be rewritten for us to see
that it is equivalent to a sentence of this form. For example,
Serena and Venus are tennis players.
is simply a short way of saying
Serena is a tennis player, and Venus is a tennis player.
At other times, the word “and” is not used to produce a conjunction of
propositions. For example,
Serena and Venus are playing each other.
does not mean that
Serena is playing each other, and Venus is playing each other.
That does not even make sense, so the original sentence cannot express a
conjunction of two propositions. Instead, it expresses a single proposition
about two people taken as a group. Consequently, it should not be symbol-
ized as “p & q.” Often, unfortunately, it is unclear whether a sentence ex-
presses a conjunction of propositions or a single proposition about a group.
The sentence
Serena and Venus are playing tennis.
could be taken either way. Maybe Serena and Venus are playing each other.
If that is what it means, then the sentence expresses a single proposition
about a group, so it should not be symbolized as “p & q.” But maybe Serena
is playing one match, while Venus is playing another. If that would make it
true, then the sentence expresses a conjunction of propositions, so it may be
symbolized as “p & q.”
When a sentence containing the word “and” expresses the conjunction of
two propositions, we will say that it expresses a propositional conjunction.
When a sentence containing “and” does not express the conjunction of two
propositions, we will say that it expresses a nonpropositional conjunction. In
this chapter we are concerned only with sentences that express proposi-
tional conjunctions. A sentence should be translated into the symbolic form
“p & q” only if it expresses a propositional conjunction. There is no mechani-
cal procedure that can be followed to determine whether a certain sentence
expresses a conjunction of two propositions. You must think carefully about
what the sentence means and about the context in which that sentence is
used. This takes practice.
146
CHAPTER 6 ■ Propos it ional Log ic
Which of the following propositions is a substitution instance of “p & q & q“?
1. The night is young, and you’re so beautiful, and my flight leaves in thirty
minutes.
2. The night is young, and you’re so beautiful, and my flight leaves in thirty
minutes, and my flight leaves in thirty minutes.
3. You’re so beautiful, and you’re so beautiful, and you’re so beautiful.
Exercise II
For each of the following propositions, give three different propositional forms
of which that proposition is a substitution instance.
1. The night is young, and you’re so beautiful, and my flight leaves in thirty
minutes.
2. The night is young, and you’re so beautiful, and you’re so beautiful.
Exercise III
Indicate whether each of the following sentences expresses a propositional con-
junction or a nonpropositional conjunction—that is, whether or not it expresses
a conjunction of two propositions. If the sentence could be either, then specify a
context in which it would naturally be used to express a propositional conjunc-
tion and a different context in which it would naturally be used to express a
nonpropositional conjunction.
1. A Catholic priest married John and Mary.
2. Fred had pie and ice cream for dessert.
3. The winning presidential candidate rarely loses both New York
and California.
Exercise IV
The proposition “The night is young, and you’re so beautiful” is a substitution
instance of which of the following propositional forms?
1. p
2. q
3. p & q
4. p & r
Exercise I
5. p & q & r
6. p & p
7. p or q
147
Bas ic Propos it ional Connect ives
Now we can look at an argument involving conjunction. Here is one that
is ridiculously simple:
Harry is short and John is tall.
Harry is short.
This argument is obviously valid. But why is it valid? Why does the conclusion
follow from the premise? The answer in this case seems obvious, but we will
spell it out in detail as a guide for more difficult cases. Suppose we replace
these particular propositions with propositional forms, using a different vari-
able for each distinct proposition throughout the argument. This yields what
we will call an argument form. For example:
p & q
p
This is a pattern for endlessly many arguments, each of which is called a sub-
stitution instance of this argument form. Every argument that has this general
form will also be valid. It really does not matter which propositions we put into
this schema; the resulting argument will be valid—so long as we are careful to
substitute the same proposition for the same variable throughout.
Let’s pursue this matter further. If an argument has true premises and a
false conclusion, then we know at once that it is invalid. But in saying that an
argument is valid, we are not only saying that it does not have true premises
and a false conclusion; we are also saying that the argument cannot have a
false conclusion when the premises are true. Sometimes this is true because
the argument has a structure or form that rules out the very possibility of true
premises and a false conclusion. We can appeal to the notion of an argument
form to make sense of this idea. A somewhat more complicated truth table
will make this clear:
PREMISE CONCLUSION
p q p & q p
T T T T
T F F T
F T F F
F F F F
4. Susan got married and had a child.
5. Jane speaks both French and English.
6. Someone who speaks both French and English is bilingual.
7. Ken and Naomi are two of my best friends.
8. Miranda and Nick cooked dinner.
9. I doubt that John is poor and happy.
148
CHAPTER 6 ■ Propos it ional Log ic
The first two columns give all the combinations for the truth values of the
propositions that we might substitute for “p” and “q.” The third column
gives the truth value of the premise for each of these combinations. (This
column is the same as the definition for “&” given above.) Finally, the fourth
column gives the truth value for the conclusion for each combination. (Here,
of course, this merely involves repeating the first column. Later on, things
will become more complicated and interesting.) If we look at this truth table,
we see that no matter how we make substitutions for the variables, we never
have a case in which the premise is true and the conclusion is false. In
the first line, the premise is true and the conclusion is also true. In the re-
maining three lines, the premise is not true, so the possibility of the premise
being true and the conclusion false does not arise.
Here it is important to remember that a valid argument can have false
premises, for one proposition can follow from another proposition that is
false. Of course, an argument that is sound cannot have a false premise, be-
cause a sound argument is defined as a valid argument with true premises.
But our subject here is validity, not soundness.
Let’s summarize this discussion. In the case we have examined, validity
depends on the form of an argument and not on its particular content. A first
principle, then, is this:
An argument is valid if it is an instance of a valid argument form.
Hence, the argument “Harry is short and John is tall; therefore, Harry is
short” is valid because it is an instance of the valid argument form “p & q;
p.”
Next we must ask what makes an argument form valid. The answer to
this is given in this principle:
An argument form is valid if and only if it has no substitution instances
in which the premises are true and the conclusion is false.
We have just seen that the argument form “p & q; p” passes this test. The
truth table analysis showed that. Incidentally, we can use the same truth
table to show that the following argument is valid:
John is tall.
Harry is short.
John is tall and Harry is short.
p
q
p & q
The argument on the left is a substitution instance of the argument form on
the right. A glance at the truth table will show that there can be no cases for
which all the premises could be true and the conclusion false. This pretty
well covers the logical properties of conjunction.
Notice that we have not said that every argument that is valid is so in virtue
of its form. There may be arguments in which the conclusion follows from the
premises but we cannot show how the argument’s validity is a matter of logical
149
Bas ic Propos it ional Connect ives
form. There are, in fact, some obviously valid arguments that have yet to be
shown to be valid in terms of their form. Explaining validity by means of logi-
cal form has long been an ideal of logical theory, but there are arguments—
many of them quite common—where this ideal has yet to be adequately
fulfilled. Many arguments in mathematics fall into this category. At present,
however, we will only consider arguments in which the strategy we used for
analyzing conjunction continues to work.
Are the following arguments valid by virtue of their propositional form? Why
or why not?
1. Donald owns a tower in New York and a palace in Atlantic City.
Therefore, Donald owns a palace in Atlantic City.
2. Tom owns a house. Therefore, Tom owns a house and a piece of land.
3. Ilsa is tall. Therefore, Ilsa is tall, and Ilsa is tall.
4. Bernie has a son and a daughter. Bernie has a father and a mother.
Therefore, Bernie has a son and a mother.
5. Mary got married and had a child. Therefore, Mary had a child and got
married.
6. Bess and Katie tied for MVP. Therefore, Bess tied for MVP.
Exercise V
For each of the following claims, determine whether it is true or false. Defend
your answers.
1. An argument that is a substitution instance of a valid argument form is
always valid.
2. An argument that is a substitution instance of an invalid argument form is
always invalid.
3. An invalid argument is always a substitution instance of an invalid argu-
ment form.
Exercise VI
Is a valid argument always a substitution instance of a valid argument form?
Why or why not?
Discussion Question
150
CHAPTER 6 ■ Propos it ional Log ic
DISJUNCTION
Just as we can form a conjunction of two propositions by using the connective
“and,” we can form a disjunction of two propositions by using the connective
“or,” as in the following compound sentence:
John will win or Harry will win.
Again, it is easy to see that the truth of this whole compound proposition
depends on the truth of the component propositions. If they are both false,
then the compound proposition is false. If just one of them is true, then the
compound proposition is true. But suppose they are both true. What shall
we say then?
Sometimes when we say “either-or,” we seem to rule out the possibility of
both. When a waiter approaches your table and tells you, “Tonight’s dinner
will be chicken or steak,” this suggests that you cannot have both. In other
cases, however, it does not seem that the possibility of both is ruled out—for
example, when we say to someone, “If you want to see tall mountains, go to
California or Colorado.”
One way to deal with this problem is to say that the English word “or” has
two meanings: one exclusive, which rules out both, and one inclusive, which
does not rule out both. Another solution is to claim that the English word “or”
always has the inclusive sense, but utterances with “or” sometimes conversa-
tionally imply the exclusion of both because of special features of certain con-
texts. It is, for example, our familiarity with common restaurant practices that
leads us to infer that we cannot have both when the waiter says, “Tonight’s din-
ner will be chicken or steak.” If we may have both, then the waiter’s utterance
would not be as informative as is required for the purpose of revealing our op-
tions, so it would violate Grice’s conversational rule of Quantity (as discussed
in Chapter 2). That explains why the waiter’s utterance seems to exclude both.
Because such explanations are plausible, and because it is simpler as well
as traditional to develop propositional logic with the inclusive sense of
“or,” we will adopt that inclusive sense. Where necessary, we will define
the exclusive sense using the inclusive sense as a starting point. Logicians
symbolize disjunctions using the connective “ ” (called a wedge). The truth
table for this connective has the following form:
p q p q
T T T
T F T
F T T
F F F
We will look at some arguments involving this connective in a moment.
NEGATION
With conjunction and disjunction, we begin with two propositions and con-
struct a new proposition from them. There is another way in which we can
construct a new proposition from just one proposition—by negating it.
151
Bas ic Propos it ional Connect ives
Given the proposition “John is clever,” we can get a new proposition, “John
is not clever,” simply by inserting the word “not” in the correct place in the
sentence.
What, exactly, does the word “not” mean? This can be a difficult question
to answer. Does it mean “nothing” or, maybe, “nothingness“? Although
some respectable philosophers have sometimes spoken in this way, it is im-
portant to see that the word “not” does not stand for anything at all. It has an
altogether different function in the language. To see this, think about how
conjunction and disjunction work. Given two propositions, the word “and”
allows us to construct another proposition that is true only when both origi-
nal propositions are true, and false otherwise. With disjunction, given two
propositions, the word “or” allows us to construct another proposition that is
false only when both of the original propositions are false, and true other-
wise. (Our truth table definitions reflect these facts.) Using these definitions
as models, how should we define negation? A parallel answer is that the
negation of a proposition is true just in the cases in which the original propo-
sition is false, and it is false just in the cases in which original proposition is
true. Using the symbol “~” (called a tilde) to stand for negation, this gives us
the following truth table definition:
p ~p
T F
F T
Negation might seem as simple as can be, but people quite often get con-
fused by negations. If Diana says, “I could not breathe for a whole minute,”
she might mean that there was a minute when something made her unable
to breathe (maybe she was choking) or she might mean that she was able to
hold her breath for a whole minute (say, to win a bet). If “A” symbolizes
“Diana could breathe sometime during this minute,” then “~A” symbolizes
the former claim (that Diana was unable to breathe for this minute). Conse-
quently, the latter claim (that Diana could hold her breath for this minute)
should not also be symbolized by “~A.” Indeed, this interpretation of the
original sentence is not a negation, even though the original sentence did in-
clude the word “not.” Moreover, some sentences are negations even though
they do not include the word “not.” For example, “Nobody owns Mars” is
the negation of “Somebody owns Mars.” If the latter is symbolized as “A,”
the former can be symbolized as “~A,” even though the former does not
include the word “not.”
The complexities of negation can be illustrated by noticing that the simple
sentence “Everyone loves running” can include negation at four distinct
places: “Not everyone loves running,“ “Everyone does not love running,“
“Everyone loves not running,“ and the colloquial “Everyone loves running—
not!” Some of these sentences can be symbolized in propositional logic as
negations of “Everyone loves running,“ but others cannot.
To determine whether a sentence can be symbolized as a negation in
propositional logic, it is often useful to reformulate the sentence so that it
152
CHAPTER 6 ■ Propos it ional Log ic
starts with “It is not the case that . . . .” For example, “I did none of the
homework” would be reformulated as “It is not the case that I did any of the
homework.” If the resulting sentence means the same as the original (as it
does in this example), then the original sentence can be symbolized as a
propositional negation. In contrast, “I promise not to leave you” means
something very different from “It is not the case that I promise to leave
you,“ so “I promise not to leave you” should not be symbolized as a propo-
sitional negation.
Unfortunately, this test will not always work. There is no complete mechan-
ical procedure for determining whether an English sentence can be symbolized
as a negation. All you can do is think carefully about the sentence’s meaning
and context. The best way to get good at this is to practice.
Put each of the following sentences in symbolic form. Be sure to specify ex-
actly which sentence is represented by each capital letter, and pay special at-
tention to the placement of the negation. If the sentence could be interpreted
Exercise IX
Negative terms or prefixes can often be interpreted in more than one way.
Explain two ways to interpret each of the following sentences. Describe a con-
text in which it would be natural to interpret it in each way.
1. You may not go to the meeting.
2. I cannot recommend him too highly.
3. He never thought he’d go to the Himalayas.
4. Have you not done all of your homework?
5. All of his friends are not students.
6. I will not go to some football games next season.
7. No smoking section available.
8. The lock on his locker was unlockable.
Exercise VIII
Explain the differences in meaning among “Not everyone loves running,”
“Everyone does not love running,” “Everyone loves not running,” and
“Everyone loves running—not!” For each, is it a negation of “Everyone loves
running“? Why or why not?
Exercise VII
153
Bas ic Propos it ional Connect ives
PROCESS OF ELIMINATION
Using only negation and disjunction, we can analyze the form of one common
pattern of reasoning, which is called process of elimination or, more technically,
disjunctive syllogism. As an example, consider this argument:
Her phone line is busy, so she must either be talking on the phone or using
her modem. She is not using her modem, since I just tried to e-mail her and
she did not respond. So she must be talking on the phone.
After trimming off assurances and subarguments that support the premises,
the core of this argument can be put in standard form:
(1) She is either using her modem or talking on the phone.
(2) She is not using her modem.
(3) She is talking on the phone. (from 1–2)
This core argument is then an instance of this argument form:
1. p q
~p
q
It does not matter if we change the order of the disjuncts so that the first
premise is “She must be either talking on the phone or using her modem.”
Then the argument takes this form:
2. p q
~q
p
Both of these argument forms are valid, so the core of the original argument is
also valid.
in more than one way, symbolize each interpretation and describe a context in
which it would be natural to interpret it in each way.
1. It won’t rain tomorrow.
2. It might not rain tomorrow.
3. There is no chance that it will rain tomorrow.
4. I believe that it won’t rain tomorrow.
5. Joe is not too smart or else he’s very clever.
6. Kristin is not smart or rich.
7. Sometimes you feel like a nut; sometimes you don’t. (from an advertisement
for Mounds and Almond Joy candies, which are made by the same company
and are exactly alike except that only one of them has a nut)
154
CHAPTER 6 ■ Propos it ional Log ic
Process of elimination is sometimes confused with a similar but crucially
different pattern of reasoning, which can be called affirming a disjunct. This
pattern includes both of these forms:
3. p q
p
~q
Explain why argument forms 1–2 are valid. Use common language that would
be understandable to someone who has not read this chapter.
Exercise X
Give other instances of argument forms 3–4 that are not valid. Explain why
these instances are invalid and why they show that the general argument form
is invalid.
Exercise XI
HOW TRUTH-FUNCTIONAL CONNECTIVES WORK
We have now defined conjunction, disjunction, and negation. That, all by it-
self, is sufficient to complete the branch of modern logic called propositional
logic. The definitions themselves may seem peculiar. They do not look like the
definitions we find in a dictionary. But the form of these definitions is impor-
tant, for it tells us something interesting about the character of such words as
“and,” “or,” and “not.” Two things are worth noting: (1) These expressions are
used to construct a new proposition from old ones; (2) the newly constructed
proposition is always a truth function of the original propositions—that is, the
4. p q
q
~p
These forms of argument are invalid. This can be shown by the following
single instance:
She is either using her modem or talking on the phone.
She is using her modem.
She is not talking on the phone.
This argument might seem valid if one assumes that she cannot talk on the
phone while using her modem. The premises, however, do not specify that she
has only one phone line. If she talks on one phone line while using her modem
on a different phone line, then the premises are true and the conclusion is false.
Because this is possible, the argument is invalid, and so is its form, 3. Moreover,
this argument would remain invalid if the disjuncts were listed in a different
order, so that the argument took the form of 4. Thus, affirming a disjunct is a
fallacy.
155
Bas ic Propos it ional Connect ives
truth value of the new proposition is always determined by the truth value of
the original propositions. For this reason, these connectives are called truth-
functional connectives. (Of course, with negation, we start with a single proposi-
tion, so there are not really two things to connect.) For example, suppose that
“A” and “B” are two true propositions and “G” and “H” are two false propo-
sitions. We can then determine the truth values of more complex propositions
built from them using conjunction, disjunction, and negation. Sometimes the
correct assignment is obvious at a glance:
A & B True
A & G False
~A False
~G True
A H True
G H False
~A & G False
As noted earlier, parentheses can be used to distinguish groupings. Some-
times the placement of parentheses can make an important difference, as in
the following two expressions:
~A & G
~(A & G)
Notice that in one expression the negation symbol applies only to the propo-
sition “A,” whereas in the other expression it applies to the entire proposition
“A & G.” Thus, the first expression above is false, and the second expression
is true. Only the second expression can be translated as “Not both A and G.”
Both of these expressions are different from “~A & ~G,” which means “Nei-
ther A nor G.”
As expressions become more complex, we reach a point where it is no
longer obvious how the truth values of the component propositions determine
the truth value of the entire proposition. Here a regular procedure is helpful.
The easiest method is to fill in the truth values of the basic propositions and
then, step-by-step, make assignments progressively wider, going from the in-
side out. For example:
~((A G) & ~(~H & B))
~((T F) & ~(~F & T))
~((T F) & ~(T & T))
~(T & ~(T))
~(T & F)
~(F)
T
With a little practice, you can master this technique in dealing with other
very complex examples.
156
CHAPTER 6 ■ Propos it ional Log ic
TESTING FOR VALIDITY
What is the point of all this? In everyday life, we rarely run into an expres-
sion as complicated as the one in our example at the end of the previous
section. Our purpose here is to sharpen our sensitivity to how truth-
functional connectives work and then to express our insights in clear ways.
This is important because the validity of many arguments depends on the
logical features of these truth-functional connectives. We can now turn di-
rectly to this subject.
Earlier we saw that every argument with the form “p & q; p” will be
valid. This is obvious in itself, but we saw that this claim could be justified
by an appeal to truth tables. A truth table analysis shows us that an argu-
ment with this form can never have an instance in which the premise is true
and the conclusion is false. We can now apply this same technique to argu-
ments that are more complex. In the beginning, we will examine arguments
that are still easy to follow without the use of technical help. In the end, we
will consider some arguments that most people cannot follow without
guidance.
Consider the following argument:
Valerie is either a doctor or a lawyer.
Valerie is neither a doctor nor a stockbroker.
Valerie is a lawyer.
We can use the following abbreviations:
D = Valerie is a doctor.
L = Valerie is a lawyer.
S = Valerie is a stockbroker.
Given that “A,” “B,” and “C” are true propositions and “X,” “Y,” and “Z” are
false propositions, determine the truth values of the following compound
propositions:
1. ~X Y 9. ~(A (Z X))
2. ~(X Y) 10. ~(A ~(Z X))
3. ~(Z Z) 11. ~A ~(Z X)
4. ~(Z ~Z) 12. ~Z (Z & A)
5. ~ ~(A B) 13. ~(Z (Z & A))
6. (A Z) & B 14. ~((Z Z) & A)
7. (A X) & (B Z) 15. A ((~B & C) ~(~B ~(Z B)))
8. (A & Z) (B & Z) 16. A & ((~B & C) ~(~B ~(Z B)))
Exercise XII
157
Bas ic Propos it ional Connect ives
Using these abbreviations, the argument and its counterpart argument
form look like this:
D L
~(D S)
L
PREMISE PREMISE CONCLUSION
p q r (p q) (p r) ~(p r) q
T T T T T F T
T T F T T F T
T F T T T F F
T F F T T F F
F T T T T F T
F T F T F T T OK
F F T F T F F
F F F F F T F
Notice that there is only one combination of truth values for which both
premises are true, and in that case the conclusion is true as well. So the origi-
nal argument is valid because it is an instance of a valid argument form—that
is, an argument form with no substitution instances for which true premises
are combined with a false conclusion.
This last truth table may need some explaining. First, why do we get eight
rows in this truth table where before we got only four? The answer to this is
that we need to test the argument form for every possible combination of truth
values for the component propositions. With two variables, there are four
possible combinations: (TT), (TF), (FT), and (FF). With three variables, there
are eight possible combinations: (TTT), (TTF), (TFT), (TFF), (FTT), (FTF),
(FFT), and (FFF). The general rule is this: If an argument form has n vari-
ables, the truth table used in its analysis must have 2n rows. For four vari-
ables there will be sixteen rows; for five variables, thirty-two rows; for six
variables, sixty-four rows; and so on. You can be sure that you capture all
possible combinations of truth values by using the following pattern in con-
structing the columns of your truth table under each individual variable:
First column Second column Third column . . .
First half Ts, First quarter Ts, First eighth Ts,
second half Fs. second quarter Fs, second eighth Fs,
and so on. and so on.
A glance at the earlier examples in this chapter will show that we have been
using this pattern, and it is the standard way of listing the possibilities.
p q
~(p r)
q
The expression on the right gives the argument form of the argument pre-
sented on the left. To test the argument for validity, we ask whether the argu-
ment form is valid. The procedure is cumbersome, but perfectly mechanical:
158
CHAPTER 6 ■ Propos it ional Log ic
Of course, as soon as an argument becomes at all complex, these truth tables
become very large indeed. But there is no need to worry about this, because
we will not consider arguments with many variables. Those who do so turn
to a computer for help.
The style of the truth table above is also significant. The premises are
plainly labeled, and so is the conclusion. A line is drawn under every row
in which the premises are all true. (In this case, there is only one such row—
row 6.) If the conclusion on this line is also true, it is marked “OK.” If every
line in which the premises are all true is OK, then the argument form
is valid. Marking all this may seem rather childish, but it is worth doing.
First, it helps guard against mistakes. More importantly, it draws one’s
attention to the purpose of the procedure being used. Cranking out truth
tables without understanding what they are about—or even why they
might be helpful—does not enlighten the mind or elevate the spirit.
For the sake of contrast, we can next consider an invalid argument:
(1) Valerie is either a doctor or a lawyer.
(2) Valerie is not both a lawyer and a stockbroker.
(3) Therefore, Valerie is a doctor.
Using the same abbreviations as earlier, this becomes:
D L p q
~(L & S) ~(q & r)
D p
The truth table for this argument form looks like this:
PREMISE PREMISE CONCLUSION
p q r (p q) (q & r) ˜(q & r) p
T T T T T F T
T T F T F T T OK
T F T T F T T OK
T F F T F T T OK
F T T T T F F
F T F T F T F Invalid
F F T F F T F
F F F F F T F
This time, we find four rows in which all the premises are true. In three cases
the conclusion is true as well, but in one of these cases (row 6), the conclusion
is false. This line is marked “Invalid.” Notice that every line in which all of
the premises are true is marked either as “OK” or as “Invalid.” If even one
row is marked “Invalid,” then the argument form as a whole is invalid. The
argument form under consideration is thus invalid, because it is possible for
it to have a substitution instance in which all the premises are true and the
conclusion is false.
159
Bas ic Propos it ional Connect ives
The labeling not only shows that the argument form is invalid, it also shows
why it is invalid. Each line that is marked “Invalid” shows a combination of
truth values that makes the premises true and the conclusion false. Row 6
presents the combination in which Valerie is not a doctor, is a lawyer, and is
not a stockbroker. With these assignments, it will be true that she is either a
doctor or a lawyer (premise 1), and also true that she is not both a lawyer and
a stockbroker (premise 2), yet false that she is a doctor (the conclusion). It is
this possibility that shows why the argument form is not valid.
In sum, we can test a propositional argument form for validity by following
these simple steps:
1. Provide a column for each premise and the conclusion.
2. Fill in truth values in each column.
3. Underline each row where all of the premises are true.
4. Mark each row “OK” if the conclusion is true on that row.
5. Mark each row “Invalid” if the conclusion is false on that row.
6. If any row is marked “Invalid,” the argument form is invalid.
7. If no row is marked “Invalid,” the argument form is valid.
Using the truth table technique outlined above, show that argument forms 1–2
in the above section on process of elimination are valid and that argument
forms 3–4 in the same section are invalid.
Exercise XIII
Using the truth table technique outlined above, explain why the “Tricky Case”
that was mentioned in Chapter 2 is valid.
Exercise XIV
Using the truth table technique outlined above, test the following argument
forms for validity:
1. ~ p q
p
~q
2. ~ (p q)
~q
Exercise XV
3. ~ (p q)
p
q
4. ~ (p q)
p
r
(continued)
160
CHAPTER 6 ■ Propos it ional Log ic
SOME FURTHER CONNECTIVES
We have developed the logic of propositions using only three basic notions cor-
responding (perhaps roughly) to the English words “and,” “or,” and “not.”
Now let us go back to the question of the two possible senses of the word “or“:
one exclusive and the other inclusive. Sometimes “or” seems to rule out the
possibility that both alternatives are true; at other times “or” seems to allow this
possibility. This is the difference between exclusive and inclusive disjunction.
Suppose we use the symbol “ ” to stand for exclusive disjunction. This is
the same as the symbol for inclusive disjunction except that it is underlined.
(After this discussion, we will not use it again.) We could then give two truth
table definitions, one for each of these symbols:
INCLUSIVE EXCLUSIVE
p q p q p q
T T T F
T F T T
F T T T
F F F F
We could also define this new connective in the following way:
(p q) = (by definition) ((p q) & ~(p & q))
It is not hard to see that the expression on the right side of this definition
captures the force of exclusive disjunction. Because we can always define
exclusive disjunction when we want it, there is no need to introduce it into
our system of basic notions.
5. ~ (p & q)
q
~ p
6. ~ (p & q)
~q
p
7. (p & q) (p & r)
p & (q r)
Construct a truth table analysis of the expression on the right side of the pre-
ceding definition, and compare it with the truth table definition of exclusive
disjunction.
Exercise XVI
8. (p q) & (p r)
p & (q r)
9. p & q
(p r) & (q r)
10. p q
(p & r) (q & r)
161
Bas ic Propos it ional Connect ives
Actually, in analyzing arguments we have been defining new logical con-
nectives without thinking about it much. For example, “not both p and q” was
symbolized as “~(p & q).” “Neither p nor q” was symbolized as “~(p q).” Let
us look more closely at the example “~(p q).” Perhaps we should have sym-
bolized it as “~p & ~q.” In fact, we could have used this symbolization, because
the two expressions amount to the same thing. Again, this may be obvious, but
we can prove it by using a truth table in yet another way. Compare the truth
table analysis of these two expressions:
p q ˜p ˜q ˜p & ˜q (p q) ˜(p q)
T T F F F T F
T F F T F T F
F T T F F T F
F F T T T F T
Under “~p & ~q” we find the column (FFFT), and we find the same sequence
under “~(p q).” This shows that, for every possible substitution we make,
these two expressions will yield propositions with the same truth value. We
will say that these propositional forms are truth-functionally equivalent. The
above table also shows that the expressions “~q” and “~p & ~q” are not
truth-functionally equivalent, because the columns underneath these two
expressions differ in the second row, so some substitutions into these expres-
sions will not yield propositions with the same truth value.
Given the notion of truth-functional equivalence, the problem of more
than one translation can often be solved. If two translations of a sentence are
truth-functionally equivalent, then it does not matter which one we use in
testing for validity. Of course, some translations will seem more natural than
others. For example, “p q” is truth-functionally equivalent to
~((~p & ~p) & (~q ~q))
Despite this equivalence, the first form of expression is obviously more nat-
ural than the second when translating sentences, such as “It is either cloudy
or sunny.”
Use truth tables to test the following argument forms for validity:
1. p
p q
2. p q
p
~q
3. p & q
~(p q)
Exercise XVII
4. ~(p & q)
p q
5. p q
p q
6. p q
p q
162
CHAPTER 6 ■ Propos it ional Log ic
Use truth tables to test which of the following propositional forms are truth-
functionally equivalent to each other:
1. ~(p q)
2. ~(~p ~q)
3. ~p & ~q
4. p & q
Exercise XVIII
Use truth tables to determine whether the expressions in each of the following
pairs are truth-functionally equivalent:
1. “p” and “p & p” 9. “~(p q)” and “~p q”
2. “p” and “p p” 10. “~(p q)” and “~p & ~q”
3. “p ~p” and “~(p & ~p)” 11. “~~(p q)” and “~~p & ~~q”
4. “p” and “p & (q ~q)“ 12. “~(p & q)” and “~p q”
5. “p” and “p & (q & ~q)” 13. “~~(p & q)” and “~~p ~~q”
6. “p” and “p (q & ~q)” 14. “~~p ~~q” and “~(~p & ~q)”
7. “p & (q r)” and “p (q & r)” 15. “~~p & ~~q” and “~(~p ~q)”
8. “p & (q & r)” and “(p & q) & r” 16. “p & ~~q” and “~~p & q”
Exercise XIX
CONDITIONALS
So far in this chapter we have seen that by using conjunction, disjunction,
and negation, it is possible to construct compound propositions out of sim-
ple propositions. A distinctive feature of compound propositions con-
structed in these three ways is that the truth of the compound proposition is
always a function of the truth of its component propositions. Thus, these
three notions allow us to construct truth-functionally compound proposi-
tions. Some arguments depend for their validity simply on these truth-
functional connectives. When this is so, it is possible to test for validity in a
purely mechanical way. This can be done through the use of truth tables.
Thus, in this area at least, we are able to give a clear account of validity and
to specify exact procedures for testing for validity.
This truth-functional approach might seem problematic in another area:
conditionals. We will argue that an important group of conditionals can be
handled in much the same way as negation, conjunction, and disjunction.
We separate conditionals from the other connectives only because a truth-
functional treatment of conditionals is more controversial and faces prob-
lems that are instructive.
163
Cond it ionals
Conditionals have the form “If , then .” What goes in the first
blank of this pattern is called the antecedent of the conditional; what goes in
the second blank is called its consequent. Sometimes conditionals appear in
the indicative mood:
If it rains, then the crop will be saved.
Sometimes they occur in the subjunctive mood:
If it had rained, then the crop would have been saved.
There are also conditional imperatives:
If a fire breaks out, then call the fire department first!
And there are conditional promises:
If you get into trouble, then I promise to help you.
Indeed, conditionals get a great deal of use in our language, often in argu-
ments. It is important, therefore, to understand them.
Unfortunately, there is no general agreement among experts concerning
the correct way to analyze conditionals. We will simplify matters and avoid
some of these controversies by considering only indicative conditionals. We
will not examine conditional imperatives, conditional promises, or subjunc-
tive conditionals. Furthermore, at the start, we will examine only what we
will call propositional conditionals. We get a propositional conditional by sub-
stituting indicative sentences that express propositions—something either
true or false—into the schema “If , then .” Or, to use technical
language already introduced, a propositional conditional is a substitution
instance of “If p, then q” in which “p” and “q” are propositional variables. Of
the four conditional sentences listed above, only the first is clearly a propo-
sitional conditional.
Even if we restrict our attention to propositional conditionals, this will not
avoid all controversy. Several competing theories claim to provide the correct
analysis of propositional conditionals, and no consensus has been reached
concerning which is right. It may seem surprising that theorists disagree
about such a simple and fundamental notion as the if-then construction, but
they do. In what follows, we will first describe the most standard treatment
of propositional conditionals, and then consider alternatives to it.
TRUTH TABLES FOR CONDITIONALS
For conjunction, disjunction, and negation, the truth table method provides
an approach that is at once plausible and effective. A propositional condi-
tional is also compounded from two simpler propositions, and this suggests
that we might be able to offer a truth table definition for these conditionals
as well. What should the truth table look like? When we try to answer this
question, we get stuck almost at once, for it is unclear how we should fill in
the table in three out of four cases.
164
CHAPTER 6 ■ Propos it ional Log ic
p q If p, then q
T T ?
T F F
F T ?
F F ?
It seems obvious that a conditional cannot be true if the antecedent is true
and the consequent is false. We record this by putting “F” in the second row.
But suppose “p” and “q” are replaced by two arbitrary true propositions—
say, “Two plus two equals four” and “Chile is in South America.” Consider
what we shall say about the conditional:
If two plus two equals four, then Chile is in South America.
This is a very strange statement, because the arithmetical remark in the an-
tecedent does not seem to have anything to do with the geographical remark
in the consequent. So this conditional is odd—indeed, extremely odd—but is it
true or false? At this point, a reasonable response is bafflement.
Consider the following argument, which is intended to solve all these prob-
lems by providing reasons for assigning truth values in each row of the truth
table. First, it seems obvious that, if “If p, then q” is true, then it is not the case
that both “p” is true and “q” is false. That in turn means that “~(p & ~q)” must
be true. The following, then, seems to be a valid argument form:
If p, then q.
~(p & ~q)
Second, we can also reason in the opposite direction. Suppose we know that
“~(p & ~q)” is true. For this to be true, “p & ~q” must be false. We know this
from the truth table definition of negation. Next let us suppose that “p” is
true. Then “~q” must be false. We know this from the truth table definition of
conjunction. Finally, if “~q” is false, then “q” itself must be true. This line of
reasoning is supposed to show that the following argument form is valid:
~(p & ~q)
If p, then q.
The first step in the argument was intended to show that we can validly
derive “~(p & ~q)” from “If p, then q.” The second step was intended to
show that the derivation can be run in the other direction as well. But if each
of these expressions is derivable from the other, this suggests that they are
equivalent. We use this background argument as a justification for the
following definition:
If p, then q = (by definition) not both p and not q.
We can put this into symbolic notation using “)” (called a horseshoe) to
symbolize the conditional connective:
p ) q = (by definition) ~(p & ~q)
165
Cond it ionals
Given this definition, we can now construct the truth table for propositional
conditionals. It is simply the truth table for “~(p & ~q)“:
p q ~(p & ~q) p ) q ~p q
T T T T T
T F F F F
F T T T T
F F T T T
Notice that “~(p & ~q)” is also truth-functionally equivalent to the expres-
sion “~p q.” We have cited it here because “~p q” has traditionally been
used to define “p ) q.” For reasons that are now obscure, when a conditional
is defined in this truth-functional way, it is called a material conditional.
Let’s suppose, for the moment, that the notion of a material conditional cor-
responds exactly to our idea of a propositional conditional. What would fol-
low from this? The answer is that we could treat conditionals in the same way
in which we have treated conjunction, disjunction, and negation. A proposi-
tional conditional would be just one more kind of truth-functionally com-
pound proposition capable of definition by truth tables. Furthermore, the
validity of arguments that depend on this notion (together with conjunction,
disjunction, and negation) could be settled by appeal to truth table techniques.
Let us pause for a moment to examine this.
One of the most common patterns of reasoning is called modus ponens. It
looks like this:
If p, then q.
p
q
p ) q
p
q
The truth table definition of a material conditional shows at once that this
pattern of argument is valid:
PREMISE PREMISE CONCLUSION
p q p ) q q
T T T T OK
T F F F
F T T T
F F T F
The argument form called modus tollens looks like this:
p ) q
~q
~p
Use truth tables to show that this argument form is valid.
Exercise XX
166
CHAPTER 6 ■ Propos it ional Log ic
A second standard fallacy is called affirming the consequent. It looks like this:
p ) q
q
p
Use truth tables to show that this argument form is invalid.
Exercise XXI
These same techniques allow us to show that one of the traditional falla-
cies is, indeed, a fallacy. It is called the fallacy of denying the antecedent, and it
has this form:
p ) q
~p
~q
The truth table showing the invalidity of this argument form looks like this:
PREMISE PREMISE CONCLUSION
p q p ) q ~p ~q
T T T F F
T F F F T
F T T T F Invalid
F F T T T OK
Image not available due to copyright restrictions
167
Cond it ionals
The relations among these last four argument forms can be seen in this
diagram:
Antecedent Consequent
Affirming Affirming the Antecedent = Affirming the Consequent
Modus Ponens (valid) (invalid)
Denying Denying the Antecedent Denying the Consequent =
(invalid) Modus Tollens (valid)
Another argument form that has been historically significant is called a
hypothetical syllogism:
p ) q
q ) r
p ) r
Because we are dealing with an argument form containing three variables, we
must perform the boring task of constructing a truth table with eight rows:
PREMISE PREMISE CONCLUSION
p q r p ) q q ) r p ) r
T T T T T T OK
T T F T F F
T F T F T T
T F F F T F
F T T T T T OK
F T F T F T
F F T T T T OK
F F F T T T OK
This is fit work for a computer, not for a human being, but it is important to
see that it actually works.
In his radio address to the nation on April 17, 1982, President Ronald Reagan ar-
gued that the United States should not accept a treaty with the Soviet Union that
would mutually freeze nuclear weapons at current levels, because he believed
that the United States had fallen behind. Here is a central part of his argument:
It would be wonderful if we could restore the balance of power with the Soviet Union
without increasing our military power. And, ideally, it would be a long step towards
assuring peace if we could have significant and verifiable reductions of arms on both
sides. But let’s not fool ourselves. The Soviet Union will not come to any conference
table bearing gifts. Soviet negotiators will not make unilateral concessions. To achieve
parity, we must make it plain that we have the will to achieve parity by our own effort.
Put Reagan’s central argument into standard form. Then symbolize it and its
form. Does his argument commit any fallacy? If so, identify it.
Exercise XXII
168
CHAPTER 6 ■ Propos it ional Log ic
Why is it important to see that these techniques work? Most people, after all,
could see that hypothetical syllogisms are valid without going through all of
this tedious business. We seem only to be piling boredom on top of triviality.
This protest deserves an answer. Suppose we ask someone why he or she thinks
that the conclusion follows from the premises in a hypothetical syllogism. The
person might answer that anyone can see that—which, by the way, is false. Be-
yond this, he or she might say that it all depends on the meanings of the words
or that it is all a matter of definition. But if we go on to ask, “which words?”
and “what definitions?” then most people will fall silent. We have discovered
that the validity of some arguments depends on the meanings of such words
as “and,” “or,” “not,” and “if-then.” We have then gone on to give explicit defi-
nitions of these terms—definitions, by the way, that help us see how these
terms function in an argument. Finally, by getting all these simple things right,
we have produced what is called a decision procedure for determining the valid-
ity of every argument depending only on conjunctions, disjunctions, negations,
and propositional conditionals. Our truth table techniques give us a mechani-
cal procedure for settling questions of validity in this area. In fact, truth table
techniques have practical applications, for example, in computer program-
ming. But the important point here is that, through an understanding of how
these techniques work, we can gain a deeper insight into the notion of validity.
Using the truth table techniques employed above, test the following argument
forms for validity. (For your own entertainment, guess whether the argument
form is valid or invalid before working it out.)
1. p ) q
q ) p
2. p ) q
~q ) ~p
Exercise XXIV
Two more classic, common, and useful argument forms combine conditionals
with disjunction. Using truth tables, test them for validity.
Constructive Dilemma
p q
p ) r
q ) r
r
Exercise XXIII
Destructive Dilemma
~p ~q
r ) p
r ) q
~r
3. ~q ) ~p
p ) q
4. p ) q
q ) r
p ) (q & r)
5. p ) q
q ) r
~r
~p
6. p ) q
q ) r
~r ) ~p
7. p q
p ) q
q ) r
r
8. p ) (q r)
~q
~r
~p
9. (p q) ) r
p ) r
10. (p & q) ) r
p ) r
11. p ) (q ) r)
(p & q) ) r
12. (p & q) ) r
p ) (q ) r)
169
Cond it ionals
13. p ) (q ) r)
q
~r
~p
14. p ) (q ) r)
p ) q
r
15. (p q) & (p r)
~r
~q
16. (p ) q) & (p ) ~r)
q & r
~p
17. (p q) )p
~q
18. (p q) ) (p & q)
(p ) q) & (q ) p)
19. (p & q) ) (p q)
(p ) q) (q ) p)
20. r
(p ) q) (q ) p)
LOGICAL LANGUAGE AND EVERYDAY LANGUAGE
Early in this chapter we started out by talking about such common words as
“and” and “or,” and then we slipped over to talking about conjunction and
disjunction. The transition was a bit sneaky, but intentional. To understand
what is going on here, we can ask how closely these logical notions we have
defined match their everyday counterparts. We will start with conjunction,
and then come back to the more difficult question of conditionals.
At first sight, the match between conjunction as we have defined it and the
everyday use of the word “and” may seem fairly bad. To begin with, in
everyday discourse, we do not go about conjoining random bits of informa-
tion. We do not say, for example, “Two plus two equals four and Chile is in
South America.” We already know why we do not say such things, for unless
170
CHAPTER 6 ■ Propos it ional Log ic
the context is quite extraordinary, this is bound to violate the conversational
rule of Relevance. But if we are interested in validity, the rule of Relevance—
like all other conversational (or pragmatic) rules—is simply beside the point.
When dealing with validity, we are interested in only one question: If the
premises of an argument are true, must the conclusion be true as well? Con-
versational rules, as we saw in Chapter 2, do not affect truth.
The truth-functional notion of conjunction is also insensitive to another im-
portant feature of our everyday discourse: By reducing all conjunctions to
their bare truth-functional content, the truth-functional notion often misses
the argumentative point of a conjunction. As we saw in Chapter 3, each of the
following remarks has a different force in the context of an argument:
The ring is beautiful, but expensive.
The ring is expensive, but beautiful.
These two remarks point in opposite directions in the context of an actual
argument, but from a purely truth-functional point of view, we treat them
as equivalent. We translate the first sentence as “B & E” and the second as
“E & B.” Their truth-functional equivalence is too obvious to need proof.
Similar oddities arise for all discounting terms, such as “although,”
“whereas,” and “however.”
It might seem that if formal analysis cannot distinguish an “and” from a
“but,” then it can hardly be of any use at all. This is not true. A formal analysis
of an argument will tell us just one thing: whether the argument is valid or
not. If we expect the analysis to tell us more than this, we will be sorely disap-
pointed. It is important to remember two things: (1) We expect deductive ar-
guments to be valid, and (2) usually we expect much more than this from an
argument. To elaborate on the second point, we usually expect an argument
to be sound as well as valid; we expect the premises to be true. Beyond this,
we expect the argument to be informative, intelligible, convincing, and so
forth. Validity, then, is an important aspect of an argument, and formal analy-
sis helps us evaluate it. But validity is not the only aspect of an argument that
concerns us. In many contexts, it is not even our chief concern.
We can now look at our analysis of conditionals, for here we find some
striking differences between the logician’s analysis and everyday use. The
following argument forms are both valid:
1. p
q )p
2. ~p
p ) q
Though valid, both argument forms seem odd—so odd that they have actu-
ally been called paradoxical. The first argument form seems to say this: If a
Check the validity of the argument forms above using truth tables.
Exercise XXV
171
Cond it ionals
proposition is true, then it is implied by any proposition whatsoever. Here is an
example of an argument that satisfies this argument form and is therefore valid:
Lincoln was president.
If the moon is made of cheese, Lincoln was president.
This is a peculiar argument to call valid. First, we want to know what the moon
has to do with Lincoln’s having been president. Beyond this, how can his
having been president depend on a blatant falsehood? We can give these ques-
tions even more force by noticing that even the following argument is valid:
Lincoln was president.
If Lincoln was not president, then Lincoln was president.
Both arguments are instances of the valid argument form “p; q ) p.”
The other argument form is also paradoxical. It seems to say that a false
proposition implies any proposition whatsoever. The following is an instance
of this argument form:
Columbus was not president.
If Columbus was president, then the moon is made of cheese.
Here it is hard to see what the falsehood that Columbus was president has
to do with the composition of the moon.
At this point, nonphilosophers become impatient, whereas philosophers be-
come worried. We started out with principles that seemed to be both obvious
and simple. Now, quite suddenly, we are being overwhelmed with a whole se-
ries of peculiar results. What in the world has happened, and what should be
done about it? Philosophers remain divided in the answers they give to these
questions. The responses fall into two main categories: (1) Simply give up the
idea that conditionals can be defined by truth-functional techniques and search
for a different and better analysis of conditionals that avoids the difficulties in-
volved in truth-functional analysis; or (2) take the difficult line and argue that
there is nothing wrong with calling the aforementioned argument forms valid.
The first approach is highly technical and cannot be pursued in detail in
this book, but the general idea is this: Instead of identifying “If p, then q”
with “Not both p and not q,” identify it with “Not possibly both p and not q.”
This provides a stronger notion of a conditional and avoids some—though
not all—of the problems concerning conditionals. This theory is given a sys-
tematic development by offering a logical analysis of the notion of possibil-
ity. This branch of logic is called modal logic, and it has shown remarkable
development in recent decades.
The second line has been taken by Paul Grice, whose theories played a
prominent part in Chapter 2. He acknowledges—as anyone must—that the
two argument forms above are decidedly odd. He denies, however, that this
oddness has anything to do with validity. Validity concerns one thing and
one thing only: a relationship between premises and conclusion. An argu-
ment is valid if the premises cannot be true without the conclusion being
true as well. The above arguments are valid by this definition of “validity.”
172
CHAPTER 6 ■ Propos it ional Log ic
Of course, arguments can be defective in all sorts of other ways. Look at the
first argument form: (1) p; q ) p. Because “q” can be replaced by any propo-
sition (true or false), the rule of Relevance will often be violated. It is worth
pointing out violations of the rule of Relevance, but, according to Grice, this
issue has nothing to do with validity. Beyond this, arguments having this form
can also involve violations of the rule of Quantity. A conditional will be true
whenever the consequent is true. Given this, it does not matter to the truth of
the whole conditional whether the antecedent is true or false. Yet it can be mis-
leading to use a conditional on the basis of this logical feature. For example, it
would be misleading for a museum guard to say, “If you give me five dollars,
then I will let you into the exhibition,” when, in fact, he will admit you in any
case. For Grice, this is misleading because it violates the rule of Quantity. Yet
strictly speaking, it is not false. Strictly speaking, it is true.
The Grice line is attractive because, among other things, it allows us to ac-
cept the truth-functional account of conditionals, with all its simplicity. Yet
sometimes it is difficult to swallow. Consider the following remark:
If God exists, then there is evil in the world.
If Grice’s analysis is correct, even the most pious person will have to admit that
this conditional is true provided only that he or she is willing to admit that
there is evil in the world. Yet this conditional plainly suggests that there is some
connection between God’s existence and the evil in the world—presumably,
that is the point of connecting them in a conditional. The pious will wish
to deny this suggestion. All the same, this connection is something that is con-
versationally implied, not asserted. So, once more, this conditional could be
misleading—and therefore is in need of criticism and correction—but it is still,
strictly speaking, true.
Philosophers and logicians have had various responses to Grice’s posi-
tion. No consensus has emerged on this issue. The authors of this book find
it adequate, at least in most normal cases, and therefore have adopted it.
This has two advantages: (1) The appeal to conversational rules fits in well
with our previous discussions, and (2) it provides a way of keeping the logic
simple and within the range of a beginning student. Other philosophers and
logicians continue to work toward a definition superior to the truth table
definition for indicative conditionals.
OTHER CONDITIONALS IN ORDINARY LANGUAGE
So far we have considered only one form in which propositional condition-
als appear in everyday language: the conditional “If p, then q.” But proposi-
tional conditionals come in a variety of forms, and some of them demand
careful treatment.
We can first consider the contrast between constructions using “if” and
those using “only if“:
1. I’ll clean the barn if Hazel will help me.
2. I’ll clean the barn only if Hazel will help me.
173
Cond it ionals
Adopting the following abbreviations:
B = I’ll clean the barn
H = Hazel will help me
the first sentence is symbolized as follows:
H ) B
Notice that in the prose version of item 1, the antecedent and consequent
appear in reverse order; “q if p” means the same thing as “If p, then q.”
How shall we translate the second sentence? Here we should move
slowly and first notice what seems incontestable: If Hazel does not help me,
then I will not clean the barn. This is translated in the following way:
~H ) ~B
And that is equivalent to:
B ) H
If this equivalence is not obvious, it can quickly be established using a
truth table.
A more difficult question arises when we ask whether an implication runs
the other way. When I say that I will clean the barn only if Hazel will help
me, am I committing myself to cleaning the barn if she does help me? There
is a strong temptation to answer the question “yes” and then give a fuller
translation of item 2 in the following way:
(B ) H) & (H ) B)
Logicians call such two-way implications biconditionals, and we will discuss
them in a moment. But adding this second conjunct is almost surely a mistake,
for we can think of parallel cases where we would not be tempted to include
it. A government regulation might read as follows:
A student may receive a New York State Scholarship only if the student
attends a New York State school.
From this it does not follow that anyone who attends a New York State school
may receive a New York State Scholarship. There may be other requirements
as well—for example, being a New York State resident.
Why were we tempted to use a biconditional in translating sentences
containing the connective “only if“? Why, that is, are we tempted to think
that the statement “I’ll clean the barn only if Hazel will help me” implies
“If Hazel helps me, then I will clean the barn“? The answer turns on the
notion of conversational implication first discussed in Chapter 2. If I am
not going to clean the barn whether Hazel helps me or not, then it will be
misleading—a violation of the rule of Quantity—to say that I will clean the
barn only if Hazel helps me. For this reason, in many contexts, the use of a
sentence of the form “p only if q” will conversationally imply a commitment
to “p if and only if q.”
174
CHAPTER 6 ■ Propos it ional Log ic
We can next look at sentences of the form “p if and only if q“—so-called
biconditionals. If I say that I will clean the barn if and only if Hazel will help
me, then I am saying that I will clean it if she helps and I will not clean it if
she does not. Translated, this becomes:
(H ) B) & (~H ) ~B)
This is equivalent to:
(H ) B) & (B ) H)
We thus have an implication going both ways—the characteristic form of a
biconditional. In fact, constructions containing the expression “if and only if”
do not often appear in everyday speech. They appear almost exclusively in
technical or legal writing. In ordinary conversation, we capture the force of a
biconditional by saying something like this:
I will clean the barn, but only if Hazel helps me.
The decision whether to translate a remark of everyday conversation into a
conditional or a biconditional is often subtle and difficult. We have already
noticed that the use of sentences of the form “p only if q” will often conversa-
tionally imply a commitment to the biconditional “p if and only if q.” In the
same way, the use of the conditional “p if q” will often carry this same impli-
cation. If I plan to clean the barn whether Hazel helps me or not, it will cer-
tainly be misleading—again, a violation of the rule of Quantity—to say that I
will clean the barn if Hazel helps me.
We can close this discussion by considering one further, rather difficult
case. What is the force of saying “p unless q“? Is this a biconditional, or just a
conditional? If it is just a conditional, which way does the implication go?
There is a strong temptation to treat this as a biconditional, but the follow-
ing example shows this to be wrong:
McCain will lose the election unless he carries the South.
To appreciate the complexities of the little word “only,” it is useful to notice
that it fits at every point in the sentence “I hit him in the eye“:
Only I hit him in the eye.
I only hit him in the eye.
I hit only him in the eye.
I hit him only in the eye.
I hit him in only the eye.
I hit him in the only eye.
I hit him in the eye only.
Explain what each of these sentences means.
Exercise XXVI
175
Cond it ionals
This sentence clearly indicates that McCain will lose the election if he does
not carry the South. Using abbreviations, we get the following:
N = McCain will carry the South.
L = McCain will lose the election.
~N ) L
The original statement does not imply—even conversationally—that McCain
will win the election if he does carry the South. Thus,
p unless q = ~q ) p
In short, “unless” means “if not.” We can also note that “~p unless q” means
the same thing as “p only if q,” and they both are translated thus:
p ) q
Our results can be diagrammed as follows:
Translates As Often Conversationally Implies
p if q q ) p (p ) q) & (q ) p)
p only if q p ) q (p ) q) & (q ) p)
p unless q ~q ) p (p ) ~q) & (~q ) p)
Translate each of the following sentences into symbolic notation, using the
suggested symbols as abbreviations.
1. The Reds will win only if the Dodgers collapse. (R, D)
2. The Steelers will win if their defense holds up. (S, D)
3. If it rains or snows, the game will be called off. (R, S, O)
4. If she came home with a trophy and a prize, she must have won the
tournament. (T, P, W)
5. If you order the dinner special, you get dessert and coffee. (S, D, C)
6. If you order the dinner special, you get dessert; but you can have coffee
whether or not you order the dinner special. (S, D, C)
7. If the house comes up for sale, and if I have the money in hand, I will bid
on it. (S, M, B)
8. If you come to dinner, I will cook you a lobster, if you want me to. (D, L, W)
9. You can be a success if only you try. (S, T)
10. You can be a success only if you try. (S, T)
11. Only if you try can you be a success. (S, T)
12. You can be a success if you are the only one who tries. (S, O)
Exercise XXVII
(continued)
176
CHAPTER 6 ■ Propos it ional Log ic
13. Unless there is a panic, stock prices will continue to rise. (P, R)
14. I won’t scratch your back unless you scratch mine. (I, Y)
15. You will get a good bargain provided you get there early. (B, E)
16. You cannot lead a happy life without friends. (Let H = You can lead a
happy life, and let F = You have friends.)
17. The only way that horse will win the race is if every other horse drops
dead. (Let W = That horse will win the race, and let D = Every other horse
drops dead.)
18. You should take prescription drugs if, but only if, they are prescribed for
you. (T, P)
19. The grass will die without rain. (D, R = It rains.)
20. Given rain, the grass won’t die. (R, D = The grass will die.)
21. Unless it doesn’t rain, the grass won’t die. (R, D = The grass will die.)
Exercise XXVIII
(a) Translate each of the following arguments into symbolic notation. Then (b)
test each argument for truth-functional validity using truth table techniques,
and (c) comment on any violations of conversational rules.
Example: Harold is clever; so, if Harold isn’t clever, then Anna isn’t clever either. (H, A)
(a) H
~H )~A
p
~p )~q
(b) PREMISE CONCLUSION
p q ~p ~q ~p ) ~q
T T F F T OK
T F F T T OK
F T T F F
F F T T T
(c) The argument violates the rule of Relevance, because Anna’s cleverness is
irrelevant to Harold’s cleverness.
1. Jones is brave, so Jones is brave or Jones is brave. (J )
2. The Republicans will carry either New Mexico or Arizona; but, since they
will carry Arizona, they will not carry New Mexico. (A, N)
3. The Democrats will win the election whether they win Idaho or not.
Therefore, they will win the election. (D, I)
4. The Democrats will win the election. Therefore, they will win the election
whether they win Idaho or not. (D, I)
5. The Democrats will win the election. Therefore, they will win the election
whether they win a majority or not. (D, M)
177
Cond it ionals
6. If Bobby moves his queen there, he will lose her. Bobby will not lose his
queen. Therefore, Bobby will not move his queen there. (M, L)
7. John will play only if the situation is hopeless. But the situation is
hopeless. So John will play. (P, H)
8. Although Brown will pitch, the Rams will lose. If the Rams lose, their
manager will get fired. So their manager will get fired. (B, L, F)
9. America will win the Olympics unless China does. China will win the
Olympics unless Germany does. So America will win the Olympics unless
Germany does. (A, R, E)
10. If you dial 0, you will get the operator. So, if you dial 0 and do not get the
operator, then there is something wrong with the telephone. (D, O, W)
11. The Democrats will run either Jones or Borg. If Borg runs, they will lose
the South. If Jones runs, they will lose the North. So the Democrats will
lose either the North or the South. (J, B, S, N)
12. I am going to order either the fish special or the meat special. Either way, I
will get soup. So I’ll get soup. (F, M, S)
13. The grass will die if it rains too much or it does not rain enough. If it does
not rain enough, it won’t rain too much. If it rains too much, then it won’t
not rain enough. So the grass will die. (D = The grass will die, M = It rains
too much, E = It rains enough.)
14. If you flip the switch, then the light will go on. But if the light goes on,
then the generator is working. So if you flip the switch, then the generator
is working. (F, L, G) (This example comes from Charles L. Stevenson.)
1. If “~p unless q” is translated as “p ) q,” then “p unless q” can be translated
as “p q.” Why?
2. Symbolize the following argument and give its form. Does this example
show that modus ponens is not always valid? Why or why not?
Opinion polls taken just before the 1980 election showed the Republican Ronald
Reagan decisively ahead of the Democrat Jimmy Carter, with the other Republican
in the race, John Anderson, a distant third. Those apprised of the poll results
believed, with good reason:
1. If a Republican wins the election, then if it’s not Reagan who wins it will be
Anderson.
2. A Republican will win the election.
Yet they did not have reason to believe:
3. If it’s not Reagan who wins, it will be Anderson.1
Discussion Questions
(continued)
178
CHAPTER 6 ■ Propos it ional Log ic
3. Symbolize the following argument and give its form. Does this example
show that modus tollens is not always valid? Why or why not?
(1) If it rained, it didn’t rain hard.
(2) It rained hard.
(3) It didn’t rain.2
4. In order to avoid logical mistakes, it is useful to study our own psychologi-
cal tendencies. One experiment asked subjects whether the following argu-
ments are valid:
If the card has an “A” on the left, then it has a “3” on the right. The card has an “A”
on the left. Therefore, the card has a “3” on the right. (95–100 percent)
If the card has an “A” on the left, it has a “3” on the right. The card does not have a
“3” on the right. Therefore, it does not have an “A” on the left. (70–75 percent)
If the card does not have an “A” on the left, then it has a “3” on the right. The card
does not have a “3” on the right. Therefore, it has an “A” on the left. (40–50 percent)
The figures in parentheses give the percentage of people who correctly identi-
fied that argument as valid. In another experiment, the indicated percentage
of subjects gave the correct answer to these questions:
If she meets her friend, she will go to a play. She meets her friend. What follows?
(96 percent)
If she meets her friend, she will go to a play. If she has enough money, she will go
to a play. She meets her friend. What follows? (38 percent)
Again, subjects often deny the validity of arguments with implausible conclu-
sions, like this:
If her pet is a fish, then it is a phylone. If her pet is a phylone, then it is a whale. So, if
her pet is a fish, then it is a whale.
Finally, the Wason Selection Task uses cards with a capital letter on one side and a
single-figure number on the other side. Four such cards are placed on a table with,
say, “B,” “L,” “2,” and “9” on the top side in this order, then subjects are asked:
Which cards need to be turned over to check whether the following rule is true or false?
(1) If a card has a “B” on one side, it has a “2” on other side. (10 percent)
(2) If a card has a “B” on one side, it does not have a “2” on other side. (100 percent)
The figure in parentheses indicates how many subjects on average give the
correct answer for each of the rules. What are the correct answers?
How can you explain why so many people make these mistakes? How can
you avoid making these mistakes yourself?
NOTES
1 Vann McGee, “A Counterexample to Modus Ponens,” Journal of Philosophy 82, no. 9 (September
1985): 462. See also Walter Sinnott-Armstrong, James Moor, and Robert Fogelin, “A Defense of
Modus Ponens,” Journal of Philosophy 83, no. 5 (May 1986): 296–300.
2 Ernest Adams, “Modus Tollens Revisited,” Analysis 48, no. 3 (1988): 122–28. See also Walter
Sinnott-Armstrong, James Moor, and Robert Fogelin, “A Defense of Modus Tollens,” Analysis 50,
no. 1 (1990): 9–16.
Categorical Logic
In Chapter 6, we saw how validity can depend on the external connections among
propositions. This chapter will demonstrate how validity can depend on the internal
structure of propositions. In particular, we will examine two types of categorical
arguments—immediate inferences and syllogisms—whose validity or invalidity
depends on relations among the subject and predicate terms in their premises and
conclusions. Our interest in these kinds of arguments is mostly theoretical. Under-
standing the theory of the syllogism deepens our understanding of validity, even if this
theory is, in some cases, difficult to apply directly to complex arguments in daily life.
BEYOND PROPOSITIONAL LOGIC
Armed with the techniques developed in Chapter 6, let’s look at the follow-
ing argument:
All squares are rectangles.
All rectangles have parallel sides.
All squares have parallel sides.
It is obvious at a glance that the conclusion follows from the premises, so
this argument is valid. Furthermore, it seems to be valid in virtue of its form.
But it is not yet clear what the form of this argument is. To show the form of
this argument, we might try something of the following kind:
p ) q
q ) r
p ) r
But this is a mistake—and a bad mistake. We have been using the letters
“p,” “q,” and “r” as propositional variables—they stand for arbitrary proposi-
tions. But the proposition “All squares are rectangles” is not itself composed
of two propositions. Nor does it contain “if,” “then” or any other proposi-
tional connective. In fact, if we properly translate the above argument into
the language of propositional logic, we get the following result:
p
q
r
7
179
180
CHAPTER ■ 7 Categor ical Log ic
This, of course, is not a valid argument form. But if we look back at the
original argument, we see that it is obviously valid. This shows that propo-
sitional logic—however adequate it is in its own area—is not capable of
explaining the validity of all valid arguments. There is more to logic than
propositional logic.
CATEGORICAL PROPOSITIONS
To broaden our understanding of the notion of validity, we will examine a
modern version of a branch of logic first developed in ancient times—
categorical logic. Categorical logic concerns immediate inferences and
syllogisms that are composed of categorical propositions, so we need to
begin by explaining what a categorical proposition is.
In the argument above, the first premise asserts some kind of relation-
ship between squares and rectangles; the second premise asserts some kind
of relationship between rectangles and things with parallel sides; finally, in
virtue of these asserted relationships, the conclusion asserts a relationship
between squares and things having parallel sides. Our task is to understand
these relationships as clearly as possible so that we can discover the basis
for the validity of this argument. Again, we shall adopt the strategy of start-
ing from simple cases and then use the insights gained there for dealing
with more complicated cases.
A natural way to represent the relationships expressed by the proposi-
tions in an argument is through diagrams. Suppose we draw one circle
standing for all things that are squares and another circle standing for all
things that are rectangles. The claim that all squares are rectangles may be
represented by placing the circle representing squares completely inside the
circle representing rectangles.
Rectangles Squares
Squares Rectangles1 2 3
Another way of representing this relationship is to begin with overlapping
circles.
We then shade out the portions of the circles in which nothing exists, ac-
cording to the proposition we are diagramming. If all squares are rectangles,
181
Categor ical Propos it ions
Squares Rectangles1 2 3
Aliens Spies1 2 3
Triangles Squares1 2 3
Either method of representation seems plausible. Perhaps the first seems more
natural. We shall, however, use the system of overlapping circles, because
they will work better when we get to more complex arguments. They are
called Venn diagrams, after their inventor, John Venn, a nineteenth-century
English logician.
Having examined one relationship that can exist between two classes, it is
natural to wonder what other relationships might exist. Going to the opposite
extreme from our first example, two classes may have nothing in common.
This relationship could be expressed by saying, “All triangles are not
squares,” but it is more common and natural to say, “No triangles are
squares.” We diagram this claim by indicating that there is nothing in the
overlapping region of things that are both triangles and squares:
This is one of the relationships that could not be diagrammed by putting one
circle inside another. (Just try it!)
In these first two extreme cases, we have indicated that one class is either
completely included in another (“All squares are rectangles”) or completely
excluded from another (“No triangles are squares”). Sometimes, however,
we claim only that two classes have at least some things in common. We
might say, for example, “Some aliens are spies.” How shall we indicate this
relationship in the following diagram?
In this case, we do not want to cross out any whole region. We do not want to
cross out region 1 because we are not saying that all aliens are spies. Plainly,
we do not want to cross out region 2, for we are actually saying that some per-
sons are both aliens and spies. Finally, we do not want to cross out region 3,
for we are not saying that all spies are aliens. Saying that some aliens are spies
there is nothing that is a square that is not a rectangle—that is, there is noth-
ing in region 1. So our diagram looks like this:
182
CHAPTER ■ 7 Categor ical Log ic
does not rule out the possibility that some spies are homegrown. So we need
some new device to represent claims that two classes have at least some mem-
bers in common. We shall do this in the following way:
Aliens Spies*
Here the asterisk indicates that there is at least one person who is both an alien
and a spy. Notice, by the way, that we are departing a bit from an everyday
way of speaking. “Some” is usually taken to mean “more than one“; here we
let it mean “at least one.” This makes things simpler and will cause no trouble,
so long as we remember that this is what we are using “some” to mean.
Given this new method of diagramming class relationships, we can
immediately think of other possibilities. The following diagram indicates
that there is someone who is an alien but not a spy. In more natural lan-
guage, it represents the claim that some aliens are not spies.
Aliens Spies*
Next we can indicate that there is someone who is a spy but not an alien.
More simply, the claim is that some spies are not aliens, and it is represented
like this:
Aliens Spies*
These last three claims are, of course, compatible, because there might be
some aliens who are spies, some aliens who are not spies, and some spies
who are not aliens.
THE FOUR BASIC CATEGORICAL FORMS
Although two classes can be related in a great many different ways, it is pos-
sible to examine many of these relationships in terms of four basic proposi-
tional forms:
A: All S is P. E: No S is P.
I: Some S is P. O: Some S is not P.
183
Categor ical Propos it ions
These forms are called categorical forms, and propositions with these forms
are called categorical propositions.
As with the propositional forms discussed in the previous chapter, the
A, E, I, and O forms for categorical propositions are not themselves
propositions, so they are neither true nor false. Instead, they are patterns
for whole groups of propositions. We get propositions from these forms
by uniformly replacing the variables S and P with terms that refer to
classes of things. For example, “Some spies are not aliens” is a substitu-
tion instance of the O propositional form. Nonetheless, we will refer to
propositions with the A, E, I, or O form simply as A, E, I, or O proposi-
tions, except where this might cause confusion.
A and E propositions are said to be universal propositions (because they
are about all S), and I and O propositions are called particular propositions
(because they are about some S). A and I propositions are described as
affirmative propositions (because they say what is P), and E and O propo-
sitions are referred to as negative propositions (because they say what is not
P). Thus, these four basic propositional forms can be described this way:
A = Universal Affirmative E = Universal Negative
I = Particular Affirmative O = Particular Negative
These four forms fit into the following table:
Affirmative Negative
Universal A: All S is P. E: No S is P.
Particular I: Some S is P. O: Some S is not P.
Here are the Venn diagrams for the four basic categorical forms:
S P S P
S P*
A: All S is P. E: No S is P.
I: Some S is P. O: Some S is not P.
S P*
These basic categorical forms, together with their labels, classifications, and
diagrams, should be memorized, because they will be referred to often in
the rest of this chapter.
184
CHAPTER ■ 7 Categor ical Log ic
Using just the four basic categorical forms, indicate what information is given
in each of the following diagrams:
Example:
S P
No S is P.
Some S is not P.
Some P is not S.
No P is S.
* *
Exercise I
S P S P
S P*
1. 2.
S P*
* * *
*
3. 4.
* *
S P
5.
S P
6.
*
S P
7.
S P
8.
*
TRANSLATION INTO THE BASIC
CATEGORICAL FORMS
Propositions with the specific A, E, I, and O forms do not appear often in
everyday conversations. Normal people rarely say things like “All whales are
mammals. All mammals breathe air. Therefore, all whales breathe air.” Most
people talk more like this: “Whales breathe air, since they’re mammals.” Thus,
185
Categor ical Propos it ions
if our logical apparatus could be applied only to propositions with the explicit
forms of A, E, I, and O, then it would apply to few arguments in everyday life.
Fortunately, however, many common statements that are not explicitly
in a categorical form can be translated into a categorical form. For example,
when someone says, “Whales are mammals,” the speaker presumably
means to refer to all whales, so this statement can be translated into “All
whales are mammals,” which is an A proposition. We need to be careful,
however. If someone says, “Whales are found in the North Atlantic,” the
speaker probably does not mean to refer to all whales, because there are
many whales in the Pacific as well. Similarly, if someone says, “A whale is a
mammal,” this can usually be translated as “All whales are mammals,”
which is an A proposition, but this translation would be inappropriate for
“A whale is stranded on the beach,” which seems to mean “One whale is
stranded on the beach.” Thus, we can be misled badly if we look only at the
surface structure of what people say. We also need to pay attention to the
context when we translate everyday talk into the basic categorical forms.
Despite these complications, it is possible to give some rough-and-ready
guides to help in translating many common forms of expression into propo-
sitions with the A, E, I, and O forms. Let’s begin with one problem that arises
for all these categorical forms: They all require a class of things as a predicate.
Thus, “All whales are big” and “No whales live on land” should strictly be
reformulated as “All whales are big things” and “No whales are things that
live on land” or “No whales are land dwellers.” This much is easy.
Things get more complicated when we look at the word “all” in A proposi-
tions. We have already seen that the word “all” is sometimes dropped in every-
day conversation, as in “Whales are mammals.” The word “all” can also be
moved away from the start of a sentence. “Democrats are all liberal” usually
means “All Democrats are liberal,” which is an A proposition. Moreover, other
words can be used in place of “all.” Each of the following claims can, in stan-
dard contexts, be translated into an A proposition with the form “All S is P“:
Every Republican is conservative.
Any investment is risky.
Anyone who is human is mortal.
Each ant is precious to its mother.
To translate such claims, we sometimes need to construct noun phrases out of
adjectives and verbs. These transformations are often straightforward, but
sometimes they require ingenuity, and even then they can seem somewhat
contorted. For example, both “Only a fool would bungee jump” and “Nobody
but a fool would bungee jump” can usually be translated into “All people
who bungee jump are fools.” This translation might not seem as natural as the
originals, but, since the translation has the A form, it explicitly shows that this
claim has the logical properties shared by other A propositions.
With some stretching, it is also possible to translate statements about indi-
viduals into categorical form. The standard method is to translate “Socrates is
a man” as the A proposition “All things that are Socrates are men.” Similarly,
186
CHAPTER ■ 7 Categor ical Log ic
“The cannon is about to go off” in a typical context must not be translated as
the I proposition “Some cannon is about to go off,” because the original state-
ment is about a particular cannon. Instead, the original statement should be
translated as the A proposition “All things that are that cannon are about to go
off.” These translations might seem stilted, but they are necessary in order to
apply syllogistic logic to everyday forms of expression.
Similar difficulties arise with the other basic propositional forms. If a
woman says, “I am looking for a man who is not attached,” and a friend
responds, “All of the men in my church are not attached,” then this response
should probably be translated as “No men in my church are attached,” which
is an E proposition. In contrast, “All ocean dwellers are not fish” should us-
ually be translated not as the E proposition “No ocean dwellers are fish” but
rather as “Not all ocean dwellers are fish.” This means “Some ocean dwellers
are not fish,” which is an O proposition. Thus, some statements with the form
“All S are not P” should be translated as E propositions, but others should be
translated as O propositions. (This ambiguity in the form “All S are not P”
explains why it is standard to give E propositions in the less ambiguous form
“No S is P.”) Other sentences should also be translated as E propositions even
though they do not explicitly contain the word “no.” “Underground cables
are not easy to repair” and “If a cable is underground, it is not easy to repair”
and “There aren’t any underground cables that are easy to repair” can all be
translated as the E proposition “No underground cables are easy to repair.”
Similar complications also arise for I and O propositions. We already saw
that “Whales are found in the North Atlantic” should be translated as the I
proposition “Some whales are found in the North Atlantic.” In addition,
some common forms of expression can be translated as O propositions even
though they do not contain either the word “not” or the word “some.” For
example, “There are desserts without chocolate” can be translated as “Some
desserts are not chocolate,” which is an O proposition.
Because of such complications, there is no mechanical procedure for
translating common English sentences into A, E, I, and O propositions. To
find the correct translation, you need to think carefully about the sentence
and its context.
Translate each of the following sentences into an A, E, I, or O proposition. Be
sure that the subjects and predicates in your translations use nouns that refer
to classes of things (rather than adjectives or verbs). If the sentence can be
translated into different forms in different contexts, give each translation and
specify a context in which it seems natural.
1. Real men eat ants. 3. The hippo is charging.
2. Bats are not birds. 4. The hippo is a noble beast.
Exercise II
187
Categor ical Propos it ions
CONTRADICTORIES
Once we understand A, E, I, and O propositions by themselves, the next step
is to ask how they are related to each other. From their diagrams, some rela-
tionships are immediately evident. Consider the Venn diagrams for the E
and I propositional forms:
S P S P
E: No S is P. I: Some S is P.
*
The first diagram has shading in the very same region that contains an as-
terisk in the second diagram. This makes it obvious that an E proposition
and the corresponding I proposition (that is, the I proposition that has the
same subject and predicate terms as the E proposition) cannot both be true.
For an E proposition to be true, there must be nothing in the central region.
5. Not all crabs live in water.
6. All crabs do not live in water.
7. Movie stars are all rich.
8. If anybody hits me, I will
hate them.
9. If anything is broken, it does
not work.
10. Somebody loves you.
11. Somebody does not love you.
12. Nobody loves me but my mother.
13. Anybody who is Mormon
believes in God.
14. My friends are the only ones
who care.
15. Only seniors may take this
course.
16. Our pit bull is a good pet.
17. Everything that is cheap is
no good.
18. Some things that are expensive
are no good.
19. Some things that are cheap are
good.
20. Some things that are not cheap
are good.
21. Some things that are cheap are
not good.
22. Some things that are not cheap
are not good.
23. Not all cars have four wheels.
24. There are couples without
children.
25. There are no people who hate
chocolate.
26. There are people who hate
chocolate.
27. Nothing that is purple is an apple.
28. Nothing that is not white is snow.
29. There aren’t any runners who are
slow.
30. Flamingos aren’t friendly.
188
CHAPTER ■ 7 Categor ical Log ic
But for the corresponding I proposition to be true, there must be something
in the central region. Thus, they cannot both be true. They also cannot both
be false. The only way for an E proposition to be false is for there to be
something in the central region, but then the corresponding I proposition is
not false but true. The only way for the I proposition to be false is if there is
nothing in the central region, and then the E proposition is not false but
true. Thus, they cannot both be true, and they cannot both be false. In other
words, they always have opposite truth values. This relation is described
by saying that these propositions are contradictories.
More generally, we can produce a diagram for the denial of a proposition
by a simple procedure. The only information given in a Venn diagram is
represented either by shading out some region, thereby indicating that
nothing exists in it, or by putting an asterisk in a region, thereby indicating
that something does exist in it. We are given no information about regions
that are unmarked. To represent the denial of a proposition, we simply re-
verse the information in the diagram. That is, where there is an asterisk, we
put in shading; where there is shading, we put in an asterisk. Everything
else is left unchanged. Thus, we can see at once that corresponding E and I
propositions are denials of one another, so they must always have opposite
truth values. This makes them contradictories.
The same relation exists between an A proposition and its corresponding
O proposition. Consider their forms:
S P S P
A: All S is P. O: Some S is not P.
*
The diagram for an A proposition has shading exactly where the correspon-
ding O proposition has an asterisk, and they contain no other information.
Consequently, corresponding A and O propositions cannot both be false and
cannot both be true, so they are contradictories.
1. Is an A proposition a contradictory of its corresponding E proposition?
Why or why not?
2. Is an I proposition a contradictory of its corresponding O proposition?
Why or why not?
3. If one proposition is the contradictory of another, is the latter always the
contradictory of the former? Why or why not?
Exercise III
189
Categor ical Propos it ions
EXISTENTIAL COMMITMENT
It might also seem that an A proposition (with the form “All S is P”) implies
the corresponding I proposition (with the form “Some S is P”). This, how-
ever, raises a difficult problem that logicians have not fully settled. Usually
when we make a statement, we are talking about certain specific things. If
someone claims that all whales are mammals, that person is talking about
whales and mammals and stating a relationship between them. In making
this statement, the person seems to be taking the existence of whales and
mammals for granted. The remark seems to involve what logicians call exis-
tential commitment to the things referred to in the subject and predicate
terms. In the same way, stating an E proposition often seems to commit the
speaker to the existence of things in the subject and predicate classes and,
thus, to imply an O proposition. For example, someone who says, “No
whales are fish” seems committed to “Some whales are not fish.”
In other contexts, however, we seem to use universal (A and E) proposi-
tions without committing ourselves to the existence of the things referred to
in the subject and predicate terms. For example, if we say, “All trespassers
will be fined,” we are not committing ourselves to the existence of any tres-
passers or to any actual fines for trespassing; we are only saying, “If there are
trespassers, then they will be fined.” Similarly, if we tell a sleepy child, “No
ghosts are under your bed,” we are not committing ourselves to the existence
of ghosts or anything under the bed. Finally, when Newton said, “All bodies
that are acted on by no forces are at rest,” he did not commit himself to the
existence of bodies that are acted on by no forces. Given these examples of A
and E propositions that carry no commitment to the things referred to, it is
easy to think of many others.
The question then arises whether we should include existential commitment
in our treatment of universal propositions or not. Once more, we must make a
decision. (Remember that we had to make decisions concerning the truth-table
definitions of both disjunction and conditionals in Chapter 6.) Classical logic
was developed on the assumption that universal (A and E) propositions carry
existential commitment. Modern logic makes the opposite decision, treating the
claim “All men are mortal” as equivalent to “If someone is a man, then that
person is mortal,” and the claim “No men are islands” as equivalent to
“If someone is a man, then that person is not an island.” This way of speaking
carries no commitment to the existence of any men.
Which approach should we adopt? The modern approach is simpler and
has proved more powerful in the long run. For these reasons, we will adopt
the modern approach and not assign existential commitment to universal
(A and E) propositions, so these propositions do not imply particular (I and
O) propositions. All the same, there is something beautiful about the classi-
cal approach, and it does seem appropriate in some contexts to some people,
so it is worth exploring in its own right. The Appendix to this chapter will
show how to develop the classical theory by adding existential commitment
to the modern theory.
190
CHAPTER ■ 7 Categor ical Log ic
Give two new examples of contexts in which:
1. Stating an A proposition does not seem to commit the speaker to the
existence of the things to which the subject term refers.
2. Stating an A proposition does not seem to commit the speaker to the
existence of the things to which the predicate term refers.
3. Stating an E proposition does not seem to commit the speaker to the
existence of the things to which the subject term refers.
4. Stating an E proposition does not seem to commit the speaker to the
existence of the things to which the predicate term refers.
Exercise IV
VALIDITY FOR CATEGORICAL ARGUMENTS
We have introduced Venn diagrams because they provide an efficient and
illuminating way to test the validity of arguments made up of categorical
(A, E, I, and O) propositions. The basic idea is simple: An argument made
up of categorical propositions is valid if all the information contained
in the Venn diagram for the conclusion is already contained in the Venn
diagram for the premises. There are only two ways to put information into
a Venn diagram: We can either shade out an area or put an asterisk in an
area. Hence, to test the validity of an argument made up of categorical
propositions, we need only examine the diagram of the conclusion for its
information (its shading or asterisks) and then check to see if the diagram
for the premises contains this information (the same shading or asterisks).
The following simple example will give a general idea of how this works:
Diagrams
whales mammals
mammals whales
Argument
*
*
Some whales are mammals.
Some mammals are whales.
Notice that the only information contained in the diagram for the conclusion is
the asterisk in the overlap between the two circles, and that information is
already included in the diagram for the premise. Thus, the argument is valid.
191
Val id i ty for Categor ical Arguments
This argument form is valid, because all the information contained in the
Venn diagram for the conclusion is contained in the Venn diagram for the
premise. And any argument that is a substitution instance of a valid argu-
ment form is valid.
Notice that we did not say that an argument is invalid if it fails these tests—
that is, if some of the information in the Venn diagram for the conclusion (or its
form) is not contained in the Venn diagram for the premises (or their forms).
As with truth tables in propositional logic (see Chapter 6), Venn diagrams test
whether arguments are valid by virtue of a certain form, but some arguments
will be valid on a different basis, even though they are not valid by virtue of
their categorical form. Here is one example:
Diagrams
S P
P S
Argument Form
Some S are P.
Some P are S.
*
*
The same method can be used to test argument forms for validity. The form
of the previous argument and the corresponding diagrams look like this:
Diagrams
fathers male parents
male parents fathers
Argument
All fathers are male parents.
All male parents are fathers.
The Venn diagram for the conclusion includes shading in the circle for male
parents, whereas the Venn diagram for the premise includes shading in the
circle for fathers, so the premise does not contain the information for the
192
CHAPTER ■ 7 Categor ical Log ic
conclusion. Thus, this form of argument is not valid, and some arguments of
this form are not valid. Nonetheless, this particular argument is clearly
valid, since it is not possible for the premise to be true when the conclusion
is false, for the simple reason that the conclusion cannot be false. Because of
such cases, Venn diagrams can show us that an argument is valid, but they
cannot prove that an argument is invalid.
Despite this limitation, the method of Venn diagrams can be used to test
many different kinds of arguments and argument forms for validity. We will
show how this method works for two main kinds of argument: immediate
inferences and syllogisms.
CATEGORICAL IMMEDIATE INFERENCES
A categorical immediate inference is an argument with the following features:
1. It has a single premise. (That is why the inference is called immediate.)
2. It is constructed from A, E, I, and O propositions. (That is why the
inference is called categorical.)
These arguments deserve attention because they occur quite often in everyday
reasoning.
We will focus on the simplest kind of immediate inference, which is con-
version. We convert a proposition (and produce its converse) simply by revers-
ing the subject term and the predicate term. By the subject term, we mean the
term that occurs as the grammatical subject; by the predicate term, we mean
the term that occurs as the grammatical predicate. In the A proposition “All
spies are aliens,” “spies” is the subject term and “aliens” is the predicate
term; the converse is “All aliens are spies.”
In this case, identifying the predicate term is straightforward because
the grammatical predicate is a noun—a predicate nominative. Often, how-
ever, we have to change the grammatical predicate from an adjective to a
noun phrase in order to get a noun that refers to a class of things. “All
spies are dangerous” becomes “All spies are dangerous things.” Here
“spies” is the subject term and “dangerous things” is the predicate term.
Although this change is a bit artificial, it is necessary because, when we
convert a proposition (that is, reverse its subject and predicate terms), we
need a noun phrase to take the place of the grammatical subject. In English
we cannot say, “All dangerous are spies,” but we can say, “All dangerous
things are spies.”
Having explained what conversion is, we now want to know when this
operation yields a valid immediate inference. To answer this question, we
use Venn diagrams to examine the relationship between each of the four
basic categorical propositional forms and its converse. The immediate infer-
ence is valid if the information contained in the conclusion is also contained
in the premise—that is, if any region that is shaded in the conclusion is
shaded in the premise, and if any region that contains an asterisk in the con-
clusion contains an asterisk in the premise.
193
Val id i ty for Categor ical Arguments
Two cases are obvious: Both I and E propositions validly convert. From
an I proposition with the form “Some S is P,” we may validly infer its con-
verse, which has the form “Some P is S.”
S P*
I: Some S is P. Converse of I: Some P is S.
P S*
From an E proposition with the form “No S is P,” we may validly infer its
converse, which has the form “No P is S.”
S P
E: No S is P. Converse of E: No P is S.
P S
Notice that in both these cases, the information (the asterisk or shading) is
in the center of the original diagram, and the diagram for the converse flips
the original diagram. Thus, the two diagrams contain the same information,
since the diagram for the converse has exactly the same markings in the
same areas as does the diagram for the original propositional form. This
shows that E and I propositions not only logically imply their converses but
are also logically implied by them. Because the implication runs both ways,
these propositions are said to be logically equivalent to their converses, and
they always have the same truth values as their converses.
The use of a Venn diagram also shows that an O proposition cannot always
be converted validly. From a proposition with the form “Some S is not P,” we
may not always infer its converse, which has the form “Some P is not S.”
S P*
O: Some S is not P. Converse of O: Some P is not S.
P S*
Notice that in this case the information is not in the center but is instead off
to one side. As a result, the information changes when the diagram is
flipped. The asterisk is in a different circle—it is in the circle for S in the dia-
gram for an O proposition, but it is in the circle for P in the diagram for the
converse of the O proposition. That shows that an argument from an O
proposition to its converse is not always valid.1
194
CHAPTER ■ 7 Categor ical Log ic
Finally, we can see that A propositions also do not always validly convert.
From a proposition with the form “All S is P,” we may not always infer its
converse, which has the form “All P is S.”
S P P S
A: All S is P. Converse of A: All P is S.
Since the diagram is not symmetrical, the information changes when the
diagram is flipped; the shading ends up in a different circle. That shows
why this form of argument is not always valid.
Traditionally, other immediate inferences have also been studied, but we
will not run through them all here. The single example of conversion is
enough to illustrate how Venn diagrams can be used to test some arguments
for validity.
Use Venn diagrams to determine whether the following immediate inferences
are valid:
1. All dinosaurs are animals. Therefore, all animals are dinosaurs.
2. Some pterodactyls can fly. Therefore, some flying things are pterodactyls.
3. Some eryopses are not meat eaters. Therefore, some things that eat meat
are not eryopses.
4. No tyrannosaurus is a king. Therefore, no king is a tyrannosaurus.
5. Some dinosaurs are reptiles. Therefore, all dinosaurs are reptiles.
6. Some dinosaurs are not alive today. Therefore, no dinosaurs are alive today.
7. All dimetrodons eat meat. Therefore, some dimetrodons eat meat.
8. No dinosaurs are warm-blooded. Therefore, some dinosaurs are not
warm-blooded.
Exercise V
THE THEORY OF THE SYLLOGISM
In an immediate inference, we draw a conclusion directly from a single A,
E, I, or O proposition. Moreover, when two categorical propositions are
contradictories, the falsity of one can be validly inferred from the truth of
the other, and the truth of one can be validly inferred from the falsity of the
other. All these forms of argument contain only one premise. The next step
in understanding categorical propositions is to consider arguments with
two premises.
An important group of such arguments is called categorical syllogisms. The
basic idea behind these arguments is commonsensical. Suppose you wish to
195
Val id i ty for Categor ical Arguments
prove that all squares have four sides. A proof should present some link or con-
nection between squares and four-sided figures. This link can be provided by
some intermediate class, such as rectangles. You can then argue that, because
the set of squares is a subset of the set of rectangles and rectangles are a subset
of four-sided figures, squares must also be a subset of four-sided figures.
Of course, there are many other ways to link two terms by means of a third
term. All such arguments with categorical propositions are called categorical
syllogisms. More precisely, a categorical syllogism is any argument such that:
1. The argument has exactly two premises and one conclusion;
2. The argument contains only basic A, E, I, and O propositions;
3. Exactly one premise contains the predicate term;
4. Exactly one premise contains the subject term; and
5. Each premise contains the middle term.
The predicate term is simply the term in the predicate of the conclusion. It is
also called the major term, and the premise that contains the predicate term
is called the major premise. The subject term is the term in the subject of the
conclusion. It is called the minor term, and the premise that contains the sub-
ject term is called the minor premise. It is traditional to state the major prem-
ise first, the minor premise second.
Our first example of a categorical syllogism then looks like this:
All rectangles are things with four sides. (Major premise)
All squares are rectangles. (Minor premise)
All squares are things with four sides. (Conclusion)
Subject term = “Squares”
Predicate term = “Things with four sides”
Middle term = “Rectangles”
To get the form of this syllogism, we replace the terms with variables:
All M is P.
All S is M.
All S is P.
Of course, many other arguments fit the definition of a categorical syllogism.
Here is one with a negative premise:
No ellipses are things with sides.
All circles are ellipses.
No circles are things with sides.
The next categorical syllogism has a particular premise:
All squares are things with equal sides.
Some squares are rectangles.
Some rectangles are things with equal sides.
196
CHAPTER ■ 7 Categor ical Log ic
In each of the last two syllogisms, what is the subject term? The predicate
term? The middle term? The major premise? The minor premise? The form of
the syllogism (using S, P, and M)? Is the syllogism valid? Why or why not?
Exercise VI
Given the restrictions in the definition of a categorical syllogism, there are exactly
256 possible forms of categorical syllogism. Explain why.
Honors Exercise
VENN DIAGRAMS FOR SYLLOGISMS. In a previous section, we used Venn di-
agrams to test the validity of immediate inferences. Immediate inferences
contain only two terms or classes, so the corresponding Venn diagrams need
only two overlapping circles. Categorical syllogisms contain three terms or
classes. To reflect this, we will use diagrams with three overlapping circles.
If we use a bar over a letter to indicate that things in the area are not in the
class (so that S
_
indicates what is not in S), then our diagram looks like this:
S P
SPM
_ _
SPM
_
SPM
_
SPM
M
_ _ _
SPM
_ _
SPM
_ _ _
SPM SPM
This diagram has eight different areas, which can be listed in an order that
resembles a truth table:
S P M
S P M
-–
S
–
P M
S
–
P M
-–
–
S P M
–
S P M
-–
–
S
–
P M
–
S
–
P M
-–
Notice that, if something is neither an S nor a P nor an M, then it falls com-
pletely outside the system of overlapping circles. In every other case, a
thing is assigned to one of the seven compartments within the system of
overlapping circles.
197
Val id i ty for Categor ical Arguments
TESTING SYLLOGISMS FOR VALIDITY. To test the validity of a syllogism us-
ing a Venn diagram, we first fill in the diagram to indicate the information
contained in the premises. Remember that the only information contained
in a Venn diagram is indicated either by shading out an area or by putting
an asterisk in it. The argument is valid if the information expressed by the
conclusion is already contained in the diagram for the premises.2 To see
this, consider the diagrams for examples that we have already given:
All rectangles have four sides.
All squares are rectangles.
All squares have four sides.
Here’s the diagram for the premises:
Squares
Rectangles
Things having
four sides
This diagram for the conclusion contains only the information that nothing is
in the circle for squares that is not also in the circle for things having four
sides. In the diagram for the premises, all the things that are squares are cor-
ralled into the region of things that have four sides. Thus, the diagram for the
premises contains all of the information in the diagram for the conclusion.
That shows that this syllogism is valid.
Next, let’s try a syllogism with a negative premise:
No ellipses have sides.
All circles are ellipses.
No circles have sides.
Here’s the diagram for the premises:
Squares
Things having
four sides
Here’s the diagram for the conclusion:
Circles
Ellipses
Things having sides
The asterisk in the middle area of this diagram says that something is in
both circles, and that information already appears in the diagram for the
premises, so this argument is valid.
So far we have looked only at valid syllogisms. Let’s see how this method
applies to invalid syllogisms. Here is one:
All pediatricians are doctors.
All pediatricians like children.
All doctors like children.
We can diagram the premises at the left and the conclusion at the right:
198
CHAPTER ■ 7 Categor ical Log ic
We diagram the conclusion “No circles have sides” as follows:
That information is clearly already contained in the Venn diagram for the
premises, so this syllogism is also valid.
Let’s try a syllogism with a particular premise:
All squares have equal sides.
Some squares are rectangles.
Some rectangles have equal sides.
It is a good strategy to diagram a universal premise before diagramming a
particular premise. The diagram for the above argument then looks like this:
Circles Things having sides
Here’s the diagram for the conclusion—that there is something that is a
rectangle that has equal sides:
Rectangles
Squares
Things having
equal sides*
Rectangles
Things having
equal sides*
199
Val id i ty for Categor ical Arguments
It is evident that the information in the diagram for the conclusion is not
already contained in the diagram for the premises. The arrow shows dif-
ferences in informational content. Thus, this form of syllogism is not valid.
Notice that the difference between these diagrams not only tells us that this
form of syllogism is invalid; it also tells us why it is invalid. In the diagram for
the premises, there is no shading in the upper left area, which includes people
who are doctors but are not pediatricians and do not like children. This shows
that the premises do not rule out the possibility that some people are doctors
without being pediatricians or liking children. But if anyone is a doctor and
not a person who likes children, then it is not true that all doctors like chil-
dren. Because this is the conclusion of the syllogism, the premises do not rule
out all of the ways in which the conclusion might be false. As a result, this con-
clusion does not follow by virtue of categorical form.3
Here is an example of an invalid syllogism with particular premises:
Some doctors are golfers.
Some fathers are doctors.
Some fathers are golfers.
Doctors
Pediatricians
People who
like children
Doctors
Premises:
All pediatricians are doctors.
All pediatricians like children.
Conclusion:
All doctors like children.
People who
like children
Fathers Golfers
*
Fathers
Premises: Conclusion:
Golfers
* *
Doctors
Examine this diagram closely. Notice that in diagramming “Some doctors
are golfers,” we had to put an asterisk on the boundary of the circle for
200
CHAPTER ■ 7 Categor ical Log ic
fathers, because we were not given information saying whether anything
falls into the category of fathers or not. For the same reason, we had to put
an asterisk on the boundary of the circle for golfers when diagramming
“Some fathers are doctors.” The upshot was that we did not indicate that
anything exists in the region of overlap between fathers and golfers. But this
is what the conclusion demands, so the form of this syllogism is not valid.
Here is an invalid syllogism with negative premises:
No babies are golfers.
No fathers are babies.
No fathers are golfers.
Again, we see that the form of this syllogism is not valid, because the entire
area of overlap between the circles is shaded in the diagram for the conclu-
sion, but part of that area is not shaded in the diagram for the premises.
The method of Venn diagrams is adequate for deciding the validity or in-
validity of all possible forms of categorical syllogism. To master this method,
all you need is a little practice.
Fathers Golfers Fathers
Premises: Conclusion:
Golfers
Babies
Using Venn diagrams, test the following syllogistic forms for validity:
1. All M is P. 4. All P is M.
All M is S. Some M is S.
All S is P. Some S is P.
2. All P is M. 5. All P is M.
All M is S. Some S is M.
All S is P. Some S is P.
3. All M is P. 6. All P is M.
Some M is S. Some S is not M.
Some S is P. Some S is not P.
Exercise VII
201
Val id i ty for Categor ical Arguments
7. All M is P. 14. No P is M.
Some S is not M. No M is S.
Some S is not P. No S is P.
8. All M is P. 15. No P is M.
Some M is not S. All M is S.
Some S is not P. No S is P.
9. No M is P. 16. No P is M.
Some S is M. All S is M.
Some S is not P. No S is P.
10. No P is M. 17. All P is M.
Some S is M. No S is M.
Some S is not P. No S is P.
11. No P is M. 18. All M is P.
Some S is not M. No S is M.
Some S is not P. No S is P.
12. No M is P. 19. Some M is P.
Some S is not M. Some M is not S.
Some S is not P Some S is not P.
13. No P is M. 20. Some P is M.
Some M is not S. Some S is not M.
Some S is not P. Some S is P.
Explain why it is a good strategy to diagram a universal premise before dia-
gramming a particular premise in a syllogism with both.
Exercise VIII
PROBLEMS IN APPLYING THE THEORY OF THE SYLLOGISM. After mastering
the techniques for evaluating syllogisms, students naturally turn to argu-
ments that arise in daily life and attempt to use these newly acquired skills.
They are often disappointed with the results. The formal theory of the syllo-
gism seems to bear little relationship to everyday arguments, and there does
not seem to be any easy way to bridge the gap.
202
CHAPTER ■ 7 Categor ical Log ic
This gap between formal theory and its application occurs for a num-
ber of reasons. First, as we saw in Chapters 2 and 5, our everyday dis-
course leaves much unstated. Many things are conversationally implied
rather than explicitly asserted. We do not feel called on to say many
things that are matters of common agreement. Before we can apply the
theory of the syllogism to everyday arguments, these things that are sim-
ply understood must be made explicit. This is often illuminating, and
sometimes boring, but it usually involves a great deal of work. Second,
the theory of the syllogism applies to statements only in a highly stylized
form. Before we apply the theory of the syllogism to an argument, we
must cast its premises and conclusion into the basic A, E, I, and O forms.
As we saw earlier in this chapter, the needed translation is not always
simple or obvious. It may not always be possible. For these and related
reasons, modern logicians have largely abandoned the project of reducing
all reasoning to syllogisms.
Why study the theory of the syllogism at all, if it is hard to apply in
some circumstances and perhaps impossible to apply in others? The an-
swer to this question was given at the beginning of Chapter 6. The study
of formal logic is important because it deepens our insight into a central
notion of logic: validity. Furthermore, the argument forms we have stud-
ied do underlie much of our everyday reasoning, but so much else is go-
ing on in a normal conversational setting that this dimension is often
hidden. By examining arguments in idealized forms, we can study their
validity in isolation from all the other factors at work in a rich conversa-
tional setting.
There is a difference, then, between the techniques developed in Chapters
1–5 and the techniques developed in Chapters 6–7. The first five chapters
presented methods of informal analysis that may be applied directly to the
rich and complex arguments that arise in everyday life. These methods of
analysis are not wholly rigorous, but they do provide practical guides for
the analysis and evaluation of actual arguments. The chapters concerning
formal logic have the opposite tendency. In comparison with the first five
chapters, the level of rigor is very high, but the range of application is corre-
spondingly smaller. In general, the more rigor and precision you insist on,
the less you can talk about.
1. What are the chief differences between the logical procedures developed in
this chapter and those developed in the chapter on propositional logic?
2. If we evaluate arguments as they occur in everyday life by using the exact
standards developed in Chapters 6 and 7, we discover that our everyday
arguments rarely satisfy these standards, at least explicitly. Does this show
that most of our ordinary arguments are illogical? What else might it show?
Discussion Questions
203
Append ix : The Class ical Theory
APPENDIX: THE CLASSICAL THEORY
The difference between classical and modern logic is simply that the classi-
cal approach adds one more assumption—namely, that every categorical
proposition is about something. More technically, the assumption is that A,
E, I, and O propositions all carry commitment to the existence of something
in the subject class and something in the predicate class. To draw Venn dia-
grams for categorical propositions on the classical interpretation, then, all
we need to do is add existential commitment to the diagrams for their mod-
ern interpretations, which were discussed above.
But how should we add existential commitment to Venn diagrams? The
answer might seem easy: Just put an asterisk wherever there is existential
commitment. The story cannot be quite so simple, however, for the follow-
ing reason. The Venn diagram for the E propositional form on the modern
interpretation is this:
S P
Modern E: No S is P.
The classical interpretation adds existential commitment in both the subject
and the predicate, so if we represent existential commitment with an asterisk,
we get this diagram:
S P
Classical E: No S is P. (???)
* *
Although this diagram might seem to work, it breaks down when we perform
operations on it. We are supposed to be able to diagram the contradictory of a
proposition simply by substituting shading for asterisks and asterisks for shad-
ing. If we perform this operation on the previous diagram, we get this:
*S P
204
CHAPTER ■ 7 Categor ical Log ic
This diagram is very different from the Venn diagram for the I proposi-
tional form, which is the same on both classial and modern interpretations:
An I proposition, however, is supposed to be the contradictory of the corre-
sponding E proposition even on the classical interpretation. So something has
gone wrong. This problem shows that existential commitment cannot be
treated exactly like explicit existential assertion, as in I and O propositions. As
a result, we cannot use the same asterisk to represent existential commitment
in Venn diagrams. Instead, we will use a plus sign: “+.“ With this new sym-
bol, we can diagram the E propositional form on the classical interpretation
this way:
S P*
I: Some S is P.
The plus sign indicates that an E proposition carries commitment to the ex-
istence of something in each class, even though it does not explicitly assert
that something exists in either class.
From this new diagram, we can get the contradictory of a proposition by
substituting shading for asterisks and asterisks for shading, as long as we
also add plus signs to ensure that no class is empty, and drop plus signs that
are no longer needed to indicate this existential commitment. When this pro-
cedure is applied to the previous diagram, the shading becomes an asterisk
in the central area, and we can then drop the plus signs in the side areas be-
cause the central asterisk already assures us that something exists in both
circles. Thus, we get the (modern and classical) diagram for the I proposi-
tional form. Moreover, when this procedure is applied to the diagram for the
I propositional form, it yields the above diagram for the E propositional
form on the classical interpretation.
It might not be so clear, however, that E and I propositions are contra-
dictories on their classical interpretations; let us see why this is so. Two
S P
Classical E: No S is P.
+ +
205
Append ix : The Class ical Theory
propositions are contradictories if and only if they cannot both be true
and also cannot both be false. The diagram for an E proposition has
shading in the same area in which the diagram for its corresponding I
proposition has an asterisk, so they cannot both be true. It is harder to see
why these propositions cannot both be false on the classical interpreta-
tion, but this can be shown by the following argument. Suppose that an I
proposition is false. Then there is nothing in the central area, so that area
should be shaded. The classical interpretation insists that the subject and
predicate classes are not empty, so if there is nothing in the central area,
there must be something in each side area, which is indicated by a plus
sign in each side area. That gives us the diagram for the corresponding E
proposition, so that proposition is true. Thus, if an I proposition is false,
its corresponding E proposition is true. That means that they cannot both
be false. We already saw that they cannot both be true. So they are
contradictories.
The same procedure yields a classical O proposition when it is applied to
a classical A proposition, and a classical A proposition when it is applied to
a classical O proposition:
S P S P
Classical A: All S is P. Classical O: Some S is not P.
*+ +
(The plus sign on the line indicates that the commitment is to something in
the right-hand circle, but not to anything in either specific part of that circle.)
Propositions of these forms on their classical interpretations cannot both be
true and cannot both be false, so they are contradictories. Thus, this method
of diagramming seems to capture the classical interpretation of the basic
propositions.
Explain why an A proposition and its corresponding O proposition are contra-
dictories on their classical interpretations, using the diagrams above.
Exercise IX
THE CLASSICAL SQUARE OF OPPOSITION
In addition to the contradictories, there is a more extensive and elegant
set of logical relationships among categorical propositions on the classical
206
CHAPTER ■ 7 Categor ical Log ic
A: All S is P.
O: Some S is not P.
**
I: Some S is P.
E: No S is P.
+ +
+
+
Contraries
Subcontraries
Su
ba
lte
rn
s Subalterns
Contra dictoriesContra dicto
ries
S P PS
S P S P
The lines in this diagram show the logical relationships that each proposition
has to the other three. These relationships are explained below. Throughout
the discussion, it is important to remember that all of the basic propositions
are interpreted as carrying existential commitment in both their subjects and
their predicates.
CONTRADICTORIES. Two propositions are contradictories of each other (and
they contradict each other) when they are related in the following way:
1. They cannot both be true, and
2. They cannot both be false.
More simply, contradictory pairs of propositions always have opposite truth
values. We have already seen that the E and I propositions are contradicto-
ries of one another, as are the A and O propositions. This relationship holds
on both the modern interpretation and the classical interpretation.
CONTRARIES. Two propositions are said to be contraries of one another if
they are related in this way:
1. They cannot both be true, but
2. They can both be false.
interpretation. This system of relationships produces what has been called
the square of opposition.
207
Append ix : The Class ical Theory
On the classical interpretation (but not the modern interpretation), A and E
propositions with the same subject and predicate are contraries of one another.
In common life, the relationship between such corresponding A and E
propositions is captured by the notion that one claim is the complete opposite of
another. The complete opposite of “Everyone is here” is “No one is here.”
Clearly, such complete opposites cannot both be true at once. We see this read-
ily if we look at the diagrams for A and E propositions on the classical interpre-
tation. The middle region of the diagram for an A proposition shows the
existence of something that is both S and P, whereas the middle region of the
diagram for the corresponding E proposition is shaded, showing that nothing
is both S and P. It should also be clear that these A and E propositions can both
be false. Suppose that there is some S that is P and also some S that is not P:
S P**
Going from left to right, the first asterisk shows that the A proposition of the
form “All S is P” is false; the second asterisk shows that the corresponding E
proposition of the form “No S is P” is also false. Thus, these propositions can
both be false, but they cannot both be true. This makes them contraries.
SUBCONTRARIES. Propositions are subcontraries of one another when
1. They can both be true, and
2. They cannot both be false.
On the classical approach (but not the modern approach), corresponding I
and O propositions are subcontraries. To see how this works, compare the
diagrams for I and O propositions:
S P S P
I: Some S is P. O: Some S is not P.
** +
It should be clear that corresponding propositions with these forms can both
be true, since there can be some S that is P and another S that is not P. But
why can’t they both be false? Consider the left side of the following diagram:
S SP
208
CHAPTER ■ 7 Categor ical Log ic
S P S P
S P*
A: All S is P. E: No S is P.
I: Some S is P. O: Some S is not P.
S P*
Premise
Conclusion
+
+ + +
We know that, on the classical approach, there must be something in this
circle somewhere. If there is something in the overlapping region SP, then
the I proposition is true. If there is something in the nonoverlapping region
of S, then there must be something else in the circle for P, because P cannot
be empty on the classical approach; therefore, the O proposition is true.
Thus, either the I proposition or the corresponding O proposition must be
true, so they cannot both be false. We already saw that they can both be true.
Consequently, corresponding I and O propositions are subcontraries.
SUBALTERNS. Subalternation is the relationship that holds down the sides of
the classical square of opposition. Quite simply, an A proposition implies the
corresponding I proposition, and an E proposition implies the corresponding
O proposition. This relationship depends on the existential commitment found
on the classical approach and does not hold on the modern approach.
The validity of subalternation is illustrated by the following diagrams:
An A proposition includes a plus sign where the corresponding I proposi-
tion includes an asterisk, but both symbols indicate that something lies in
the middle area. Thus, the information for the I proposition is already
included in the diagram for the A proposition, which means that the A
proposition implies the I proposition.
The same point applies to the implication on the right-hand side, because
the diagram for an E proposition includes a plus sign in the same area as the
asterisk in the diagram for the corresponding O proposition. The diagram
for the E proposition also has a plus sign in the rightmost area. If something
exists in that area, then something exists in either that area or the middle
area, which is what is meant by the plus sign on the line in the diagram for
the O proposition. Thus, an E proposition implies its corresponding O
proposition.
The information given by the classical square of opposition can now be
summarized in two charts. We shall ask two questions. First, for each propo-
sitional form, if we assume that a proposition with that form is true, what
209
Append ix : The Class ical Theory
consequences follow for the truth or falsity of a corresponding proposition
with a different form?
A E I O
A T F T F
Assumed true E F T F T
I ? F T ?
O F ? ? T
“T” indicates that the corresponding proposition in that column is true, “F”
indicates that it is false, and “?” indicates that it might have either truth
value, because neither consequence follows.
Second, for each propositional form, if we assume that a proposition with
that form is false, what consequences follow for the truth or falsity of a corre-
sponding proposition with a different form?
A E I O
A F ? ? T
Assumed false E ? F T ?
I F T F T
O T F T F
THE CLASSICAL THEORY OF IMMEDIATE
INFERENCE
The difference between the modern and classical approaches is simply that
the classical approach assigns more information—specifically, existential
commitment—to the basic propositions than the modern interpretation
does. Because of this additional information, certain immediate inferences
hold on the classical approach that do not hold on the modern approach. In
particular, though conversion of an A proposition fails on both approaches,
what is known as conversion by limitation holds on the classical approach but
not on the modern approach. That is, from a proposition with the form “All
S is P,” we may not validly infer the proposition with the form “All P is S,”
but on the classical approach, we may validly infer “Some P is S.” The rea-
son is simple: From a proposition with the form “All S is P” on the classical
interpretation, we may infer a proposition with the form “Some S is P,” and
then we may convert this to get a proposition with the form “Some P is S.”
Using the Venn diagrams for the classical interpretation of the A propositional
form given above, show that conversion by limitation is classically valid for an
A proposition.
Exercise X
210
CHAPTER ■ 7 Categor ical Log ic
THE CLASSICAL THEORY OF SYLLOGISMS
As in the case of immediate inferences, the premises of syllogisms will con-
tain more information—specifically, existential commitment—on the classi-
cal interpretation than they do on the modern interpretation. This will make
some syllogisms valid on the classical approach that were not valid on the
modern approach.
We begin our study of this matter with an example that has had a curious
history:
All rectangles are four-sided.
All squares are rectangles.
Some squares are four-sided.
The argument is peculiar because its conclusion is weaker than it needs to
be. We could, after all, conclude that all squares are four-sided. The argu-
ment thus violates the conversational rule of Quantity. Perhaps for this rea-
son, this syllogism was often not included in traditional lists of valid
syllogisms. Yet the argument is valid on the classical interpretation of exis-
tential commitment, and our diagram should show this.
Squares
Rectangles
Four-sided things
Step I: Diagram the first premise
+
Squares
Rectangles
Four-sided things
Step II: Add the second premise
+ +
Notice that the plus sign is placed on the outer edge of the circle for squares,
because we are not in a position to put it either inside or outside that circle.
We now add the information for the second premise:
211
Append ix : The Class ical Theory
As expected, the conclusion that some squares are four-sided is already dia-
grammed, so the argument is valid—provided that we take A propositions
to have existential commitment.
Because classical logicians tended to ignore the previous argument, their
writings did not bring out the importance of existential commitment in eval-
uating it. There is, however, an argument that did appear on the classical
lists that makes clear the demand for existential commitment. These are syl-
logisms with the following form:
All M is P.
All M is S.
Some S is P.
This form of syllogism is diagrammed as follows:
S
M
P
+
Step II: Add the second premise
+S
M
P
Step I: Diagram the first premise
+
Again, we see that the conclusion follows, but only if we diagram A propo-
sitions to indicate existential commitment. This, then, is an argument that
was declared valid on the classical approach, but invalid on the modern
approach.
Use Venn diagrams to test the following syllogism forms for validity on the
classical approach:
1. All M is P. 3. No M is P.
No M is S. All S is M.
Some S is not P. Some S is not P.
2. No M is P. 4. No M is P.
All M is S. No M is S.
Some S is not P. Some S is not P.
Exercise XI
May not be copied, scanned, or duplicated, in whole or in part.
212
CHAPTER ■ 7 Categor ical Log ic
NOTES
1 We say “not always” rather than simply “not,” because there are some strange cases—
logicians call them “degenerate cases“—for which inferences of this pattern are valid. For ex-
ample, from “Some men are not men,” we may validly infer “Some men are not men.” Here, by
making the subject term and the predicate term the same, we trivialize conversion. Keeping
cases of this kind in mind, we must say that the inference from an O proposition to its converse
is usually, but not always, invalid. In contrast, the set of valid arguments holds in all cases, in-
cluding degenerate cases.
2 We cannot say “only if” here because of degenerate cases of categorical syllogisms that are
valid, but not by virtue of their syllogistic form. Here is one example: “All numbers divisible by
two are even. No prime number other than two is divisible by two. Therefore, no prime num-
ber other than two is even.” This syllogism is valid because it is not possible that its premises
are true and its conclusion is false, but other syllogisms with this same form are not valid.
3 We need to add “by virtue of its categorical form,” because, as we saw above, it still might be
valid on some other basis. In this particular example, however, nothing else makes this argu-
ment valid.
Are the following claims true or false? Explain your answers.
1. Every syllogism that is valid on the modern approach is also valid on the
classical approach.
2. Every syllogism that is valid on the classical approach but not on the
modern approach has a particular conclusion that starts with “Some.”
Exercise XII
May not be copied, scanned, or duplicated, in whole or in part.
How to Evaluate
Arguments: Inductive
Standards
Previous chapters have been concerned primarily with deductive arguments that
aim at validity. Many arguments encountered in daily life, however, are not
intended to meet this standard of validity. They are only supposed to provide rea-
sons (perhaps very strong reasons) for their conclusions. Such arguments are called
inductive and will be the focus of Part III. This part begins with a discussion of the
nature of inductive standards and arguments followed by a survey of five forms of
inductive argument: statistical generalizations, statistical applications, causal rea-
soning, inference to the best explanation, and arguments from analogy. The next
topic is probability, because, as we will see, the inductive standard of strength can
be understood in terms of probability. Part III will close by discussing how probabil-
ities get deployed in decision making.
III
213
May not be copied, scanned, or duplicated, in whole or in part.
This page intentionally left blank
ARGUMENTS TO AND FROM
GENERALIZATIONS
This chapter begins our investigation of inductive arguments by distinguishing the
inductive standard of strength from the deductive standard of validity. Inductive
arguments are defined as arguments that are intended to be strong rather than valid.
Two common examples of inductive arguments are discussed next. In statistical
generalizations, a claim is made about a population on the basis of features of a sam-
ple of that population. In statistical applications, a claim is made about members of
a population on the basis of features of the population. Statistical generalizations
take us up from samples to general claims, and statistical applications then take us
back down to individual cases.
INDUCTION VERSUS DEDUCTION
The distinction between deductive arguments and inductive arguments can
be drawn in a variety of ways, but the fundamental difference concerns the
relationship that is claimed to hold between the premises and the conclusion
for each type of argument. An argument is deductive insofar as it is intended
or claimed to be valid. As we know from Chapter 3, an argument is valid if
and only if it is impossible for the conclusion to be false when its premises
are true. The following is a valid deductive argument:
All ravens are black.
If there is a raven on top of Pikes Peak, then it is black.
Because the premise lays down a universal principle governing all ravens, if it’s
true, then it must be true of all ravens (if any) on top of Pikes Peak. This same re-
lationship does not hold for invalid arguments. Nonetheless, arguments that are
not valid can still be deductive if they are intended or claimed to be valid.
In contrast, inductive arguments are not intended to be valid, so they
should not be criticized for being invalid. The following is an example of an
inductive argument:
All ravens that we have observed so far are black.
All ravens are black.
8
215
216
CHAPTER 8 ■ Arguments To and From General izat ions
Here we have drawn an inductive inference from the characteristics of ob-
served ravens to the characteristics of all ravens, most of which we have not
observed. Of course, the premise of this argument could be true, yet the con-
clusion turn out to be false. A raven that has not yet been observed might
be albino. The obviousness of this possibility suggests that someone who
gives this argument does not put it forth as valid, so it is not a deductive ar-
gument. Instead, the premise is put forth as a reason or support for the con-
clusion. When an argument is not claimed to be valid but is intended only
to provide a reason for the conclusion, the argument is inductive.
Because inductive arguments are supposed to provide reasons, and rea-
sons vary in strength, inductive arguments can be evaluated as strong or
weak, depending on the strength of the reasons that they provide for their
conclusions. If we have seen only ten ravens, and all of them were in our
backyard, then the above argument gives at most a very weak reason to
believe that all ravens are black. But, if we have traveled around the world
and seen over half the ravens that exist, then the above argument gives a
strong reason to believe that all ravens are black. Inductive arguments are
usually intended to provide strong support for their conclusions, in which
case they can be criticized if the support they provide is not strong enough
for the purposes at hand.
The most basic distinction, then, is not between two kinds of argument
but is instead between two standards for evaluating arguments. The deduc-
tive standard is validity. The inductive standard is strength. Arguments
themselves are classified as either deductive or inductive in accordance with
the standard that they are intended or claimed to meet.
There are several important differences between deductive and inductive
standards. One fundamental feature of the deductive standard of validity is
that adding premises to a valid argument cannot make it invalid. The defi-
nition of validity guarantees this: In a valid argument, it is not possible for
the premises to be true without the conclusion being true as well. If any fur-
ther premises could change this, then it would be possible for this relation-
ship not to hold, so the argument would not be valid after all. Additional
information might, of course, lead us to question the truth of one of the
premises, but that is another matter.
The situation is strikingly different when we deal with inductive argu-
ments. To cite a famous example, before the time of Captain Cook’s voyage
to Australia, Europeans had observed a great many swans, and every one of
them was white. Thus, up to that time Europeans had very strong inductive
evidence to support the claim that all swans are white. Then Captain Cook
discovered black swans in Australia. What happens if we add this new piece
of information to the premises of the original inductive argument? Provided
that we accept Cook’s report, we now produce a sound deductive argument
in behalf of the opposite claim that not all swans are white; for, if some
swans are black, then not all of them are white. This, then, is a feature of the
inductive standard of strength: No matter how strong an inductive argu-
ment is, the possibility remains open that further information can undercut,
217
Induct ion versus Deduct ion
perhaps completely, the strength of the argument and the support that the
premises give to the conclusion. Because inductive strength and inductive argu-
ments can always be defeated in this way, they are described as defeasible. Valid
deductive arguments do not face a similar peril, so they are called indefeasible.
A second important difference between inductive and deductive standards
is that inductive strength comes in degrees, but deductive validity does not. An
argument is either valid or invalid. There is no question of how much validity
an argument has. In contrast, inductive arguments can be more or less strong.
The more varied ravens or swans we observe, the stronger the inductive argu-
ments above. Some inductive arguments are extremely strong and put their
conclusions beyond any reasonable doubt. Other inductive arguments are
much weaker, even though they still have some force.
Because of the necessary relationship between the premises and the conclu-
sion of a valid deductive argument, it is often said that the premises of valid
deductive arguments (if true) provide conclusive support for their conclusions,
whereas true premises of strong inductive arguments provide only partial sup-
port for their conclusions. There is something to this. Because the premises of a
valid deductive argument necessitate the truth of the conclusion, if those prem-
ises are definitely known to be true, then they do supply conclusive reasons for
the conclusion. The same cannot be said for inductive arguments.
It would be altogether misleading, however, to conclude from this that
inductive arguments are inherently inferior to deductive arguments in sup-
plying a justification or ground for a conclusion. In the first place, inductive
arguments often place matters beyond any reasonable doubt. It is possible
that the next pot of water will not boil at any temperature, however high,
but this is not something we worry about. We do not take precautions
against it, and we shouldn’t.
More important, deductive arguments normally enjoy no advantages
over their inductive counterparts. We can see this by comparing the two fol-
lowing arguments:
DEDUCTIVE INDUCTIVE
All ravens are black. All observed ravens are black.
If there is a raven on top If there is a raven on top of
of Pikes Peak, it is black. Pikes Peak, it is black.
Of course, it is true for the deductive argument (and not true for the induc-
tive argument) that if the premise is true, then the conclusion must be true.
This may seem to give an advantage to the deductive argument over the in-
ductive argument. But before we can decide how much support a deductive
argument gives its conclusion, we must ask if its premises are, after all, true.
That is not something we can just take for granted. If we examine the prem-
ises of these two arguments, we see that it is easier to establish the truth of
the premise of the inductive argument than it is to establish the truth of the
premise of the deductive argument. If we have observed carefully and kept
good records, then we might be fully confident that all observed ravens have
218
CHAPTER 8 ■ Arguments To and From General izat ions
been black. On the other hand, how can we show that all ravens (observed
and unobserved—past, present, and future) are black? The most obvious way
(though there may be other ways) would be to observe ravens to see if they
are black or not. This, of course, involves producing an inductive argument
(called a statistical generalization) for the premise of the deductive argument.
Here our confidence in the truth of the premise of the deductive argument
should be no greater than our confidence in the strength of the inference in
the statistical generalization. In this case—and it is not unusual—the deduc-
tive argument provides no stronger grounds in support of its conclusion than
does its inductive counterpart, because any reservations we might have
about the strength of the inductive inference will be paralleled by doubts con-
cerning the truth of the premise of the deductive argument.
We will also avoid the common mistake of saying that deductive argu-
ments always move from the general to the particular, whereas inductive
arguments always move from the particular to the general. In fact, both sorts
of arguments can move in either direction. There are inductive arguments
intended to establish particular matters of fact, and there are deductive
arguments that involve generalizations from particulars. For example, when
scientists assemble empirical evidence to determine whether the extinction
of the dinosaurs was caused by the impact of a meteor, their discussions are
models of inductive reasoning. Yet they are not trying to establish a general-
ization or a scientific law. Instead, they are trying to determine whether a
particular event occurred some 65 million years ago. Inductive reasoning
concerning particular matters of fact occurs constantly in everyday life as
well, for example, when we check to see whether our television reception is
being messed up by someone using a hair dryer. Deductive arguments from
the particular to the general also exist, though they tend to be trivial, and
hence boring. Here’s one:
Benjamin Franklin was the first postmaster general; therefore, anyone who is
identical with Benjamin Franklin was the first postmaster general.
Of course, many deductive arguments do move from the general to the par-
ticular, and many inductive arguments do move from particular premises to
a general conclusion. It is important to remember, however, that this is not
the definitive difference between these two kinds of arguments. What makes
deductive arguments deductive is precisely that they are intended to meet
the deductive standard of validity, and what makes inductive arguments
inductive is just that they are not intended to be deductively valid but are,
instead, intended to be inductively strong.
Assuming a standard context, label each of the following arguments as deduc-
tive or inductive. Explain what it is about the words or form of argument that
indicates whether or not each argument is intended or claimed to be valid. If it
is not clear whether the argument is inductive or deductive, say why.
Exercise I
219
Stat ist ical General izat ions
STATISTICAL GENERALIZATIONS
One classic example of an inductive argument is an opinion poll. Suppose a
candidate wants to know how popular she is with voters. Because it would
be practically impossible to survey all voters, she takes a sample of voting
opinion and then infers that the opinions of those sampled indicate the over-
all opinion of voters. Thus, if 60 percent of the voters sampled say that they
will vote for her, she concludes that she will get around 60 percent of the
vote in the actual election. As we shall see later, inferences of this kind often
1. The sun is coming out, so the rain will probably stop soon.
2. It’s going to rain tomorrow, so it will either rain or be clear tomorrow.
3. No woman has ever been elected president. Therefore, no woman will
ever be elected president.
4. Diet cola never keeps me awake at night. I know because I drank it just
last night without any problems.
5. The house is a mess, so Jeff must be home from college.
6. If Harold were innocent, he would not go into hiding. Since he is hiding,
he must not be innocent.
7. Nobody in Paris seems to understand me, so either my French is rotten or
Parisians are unfriendly.
8. Because both of our yards are near rivers in Tennessee, and my yard has
lots of mosquitoes, there must also be lots of mosquitoes in your yard.
9. Most likely, her new husband speaks English with an accent, because he
comes from Germany, and most Germans speak English with an accent.
10. There is no even number smaller than 2, so 1 is not an even number.
1. The following arguments are not clearly inductive and also not clearly
deductive. Explain why.
a. All humans are mortal, and Socrates is a human, so Socrates is likely to
be mortal also.
b. We checked every continent there is, and every raven in every continent
was observed to be black, so every raven is black.
c. If there’s radon in your basement, this monitor will go off. The monitor
is going off, so there must be radon in your basement. (Said by an engi-
neer while running the monitor in your basement.)
2. In mathematics, proofs are sometimes employed using the method of
mathematical induction. If you are familiar with this procedure, determine
whether these proofs are deductive or inductive in character. Explain why.
Discussion Questions
220
CHAPTER 8 ■ Arguments To and From General izat ions
go wrong, even when made by experts, but the general pattern of this rea-
soning is quite clear: Statistical features of a sample are used to make statis-
tical claims about the population as a whole.
Basically the same form of reasoning can be used to reach a universal con-
clusion. An example is the inductive inference discussed at the start of this
chapter: All observed ravens are black, so all ravens are black. Again, we
sample part of a population to draw a conclusion about the whole. Argu-
ments of this form, whether the conclusion is universal or partial (as when it
cites a particular percentage), are called statistical generalizations.
How do we assess such inferences? To begin to answer this question, we can
consider a simple example of a statistical generalization. On various occasions,
Harold has tried to use Canadian quarters in American payphones and found
that they have not worked. From this he draws the conclusion that Canadian
quarters do not work in American payphones. Harold’s inductive reasoning
looks like this:
In the past, when I tried to use Canadian quarters in American
payphones, they did not work.
Canadian quarters do not work in American payphones.
The force of the conclusion is that Canadian quarters never work in Ameri-
can payphones.
In evaluating this argument, what questions should we ask? We can start
with a question that we should ask of any argument.
SHOULD WE ACCEPT THE PREMISES?
Perhaps Harold has a bad memory, has kept bad records, or is a poor ob-
server. For some obscure reason, he may even be lying. It is important to ask
this question explicitly, because fairly often the premises, when challenged,
will not stand up to scrutiny.
If we decide that the premises are acceptable (that is, true and justified),
then we can shift our attention to the relationship between the premises and
the conclusion and ask how much support the premises give to the conclu-
sion. One commonsense question is this: “How many times has Harold tried
to use Canadian quarters in American payphones?” If the answer is “Once,”
then our confidence in his argument should drop to almost nothing. So, for
statistical generalizations, it is always appropriate to ask about the size of the
sample.
IS THE SAMPLE LARGE ENOUGH?
One reason we should be suspicious of small samples is that they can be af-
fected by runs of luck. Suppose Harold flips a Canadian quarter four times
and it comes up heads each time. From this, he can hardly conclude that
Canadian quarters always come up heads when flipped. He could not even
reasonably conclude that this Canadian quarter would always come up
221
Stat ist ical General izat ions
heads when flipped. The reason for this is obvious enough: If you spend a
lot of time flipping coins, runs of four heads in a row are not all that unlikely
(the probability is actually one in sixteen), and therefore samples of this size
can easily be distorted by chance. On the other hand, if Harold flipped the
coin twenty times and it continued to come up heads, he would have strong
grounds for saying that this coin, at least, will always come up heads. In fact,
he would have strong grounds for thinking that he has a two-headed coin.
Because an overly small sample can lead to erroneous conclusions, we need
to make sure that our sample includes enough trials.
How many is enough? On the assumption, for the moment, that our
sampling has been fair in all other respects, how many samples do we need
to provide the basis for a strong inductive argument? This is not always an
easy question to answer, and sometimes answering it demands subtle
mathematical techniques. Suppose your company is selling 10 million com-
puter chips to the Department of Defense, and you have guaranteed that no
more than 0.2 percent of them will be defective. It would be prohibitively
expensive to test all the chips, and testing only a dozen would hardly be
enough to reasonably guarantee that the total shipment of chips meets the
required specifications. Because testing chips is expensive, you want to test
as few as possible; but because meeting the specifications is crucial, you
want to test enough to guarantee that you have done so. Answering ques-
tions of this kind demands sophisticated statistical techniques beyond the
scope of this text.
Sometimes, then, it is difficult to decide how many instances are needed
to give reasonable support to inductive generalizations; yet many times it
is obvious, without going into technical details, that the sample is too small.
Drawing an inductive conclusion from a sample that is too small can lead
to the fallacy of hasty generalization. It is surprising how common this fal-
lacy is. We see a person two or three times and find him cheerful, and we
immediately leap to the conclusion that he is a cheerful person. That is,
from a few instances of cheerful behavior, we draw a general conclusion
about his personality. When we meet him later and find him sad, morose,
or grouchy, we then conclude that he has changed—thus swapping one
hasty generalization for another.
This tendency toward hasty generalization was discussed over 200 years
ago by the philosopher David Hume, who saw that we have a strong ten-
dency to “follow general rules which we rashly form to ourselves, and
which are the source of what we properly call prejudice.“1 More recently,
this tendency toward hasty generalization has been the subject of extensive
psychological investigation. The cognitive psychologists Amos Tversky and
Daniel Kahneman put the matter this way:
We submit that people view a sample randomly drawn from a population as
highly representative, that is, similar to the population in all essential characteristics.
Consequently, they expect any two samples drawn from a particular population to
be more similar to one another and to the population than sampling theory predicts,
at least for small samples.2
222
CHAPTER 8 ■ Arguments To and From General izat ions
To return to a previous example, we make our judgments of someone’s
personality on the basis of a very small sample of his or her behavior and ex-
pect this person to behave in similar ways in the future when we encounter
further samples of behavior. We are surprised, and sometimes indignant,
when the future behavior does not match our expectations.
By making our samples sufficiently large, we can guard against distor-
tions due to “runs of luck,” but even very large samples can give us a poor
basis for a statistical generalization. Suppose that Harold has tried hun-
dreds of times to use a Canadian quarter in an American payphone, and it
has never worked. This will increase our confidence in his generalization,
but size of sample alone is not a sufficient ground for a strong inductive ar-
gument. Suppose that Harold has tried the same coin in hundreds of differ-
ent payphones, or tried a hundred different Canadian coins in the same
payphone. In the first case, there might be something wrong with this par-
ticular coin; in the second case, there might be something wrong with this
particular payphone. In neither case would he have good grounds for mak-
ing the general claim that no Canadian quarters work in any American pay-
phones. This leads us to the third question we should ask of any statistical
generalization.
IS THE SAMPLE BIASED?
When the sample, however large, is not representative of the population,
then it is said to be unfair or biased. Here we can speak of the fallacy of
biased sampling.
One of the most famous errors of biased sampling was committed by a
magazine named the Literary Digest. Before the presidential election of 1936,
this magazine sent out 10 million questionnaires asking which candidate the
recipient would vote for: Franklin Roosevelt or Alf Landon. It received
2.5 million returns, and on the basis of the results, confidently predicted that
Landon would win by a landslide: 56 percent for Landon to only 44 percent
for Roosevelt. When the election results came in, Roosevelt had won by an
even larger landslide in the opposite direction: 62 percent for Roosevelt to a
mere 38 percent for Landon.
What went wrong? The sample was certainly large enough; in fact, by
contemporary standards it was much larger than needed. It was the way the
sample was selected, not its size, that caused the problem: The sample was
randomly drawn from names in telephone books and from club member-
ship lists. In 1936 there were only 11 million payphones in the United States,
and many of the poor—especially the rural poor—did not have payphones.
During the Great Depression there were more than 9 million unemployed
in America; they were almost all poor and thus underrepresented on club
membership lists. Finally, a large percentage of these underrepresented
groups voted for Roosevelt, the Democratic candidate. As a result of these
biases in its sampling, along with some others, the Literary Digest underesti-
mated Roosevelt’s percentage of the vote by a whopping 18 percent.
223
Stat ist ical General izat ions
Looking back, it may be hard to believe that intelligent observers could
have done such a ridiculously bad job of sampling opinion, but the story re-
peats itself, though rarely on the grand scale of the Literary Digest fiasco. In
1948, for example, the Gallup poll, which had correctly predicted Roo-
sevelt’s victory in 1936, predicted, as did other major polls, a clear victory
for Thomas Dewey over Harry Truman. Confidence was so high in this
prediction that the Chicago Tribune published a banner headline declaring
that Dewey had won the election before the votes were actually counted.
What went wrong this time? The answer here is more subtle. The Gallup
pollsters (and others) went to great pains to make sure that their sample
was representative of the voting population. The interviewers were told to
poll a certain number of people from particular social groups—rural poor,
suburban middle class, urban middle class, ethnic minorities, and so on—
so that the proportions of those interviewed matched, as closely as pos-
sible, the proportions of those likely to vote. (The Literary Digest went
bankrupt after its incorrect prediction, so the pollsters were taking no
chances.) Yet somehow bias crept into the sampling; the question was,
“How?” One speculation was that a large percentage of those sampled did
not tell the truth when they were interviewed; another was that a large
number of people changed their minds at the last minute. So perhaps the
data collected were not reliable. The explanation generally accepted was
more subtle. Although Gallup’s workers were told to interview specific
numbers of people from particular classes (so many from the suburbs, for
example), they were not instructed to choose people randomly from within
each group. Without seriously thinking about it, they tended to go to
“nicer” neighborhoods and interview “nicer” people. Because of this, they
biased the sample in the direction of their own (largely) middle-class pref-
erences and, as a result, under-represented constituencies that would give
Truman his unexpected victory.
IS THE RESULT BIASED IN SOME OTHER WAY?
Because professionals using modern techniques can make bad statistical gen-
eralizations through biased sampling, it is not surprising that our everyday, in-
formal inductive generalizations are often inaccurate. Sometimes we go astray
because of small samples and biased samples. This happens, for example,
when we form opinions about what people think or what people are like by
asking only our friends. But bias can affect our reasoning in other ways as well.
One of the main sources of bias in everyday life is prejudice. Even if we
sample a wide enough range of cases, we often reinterpret what we hear or
see in light of some preconception. People who are prejudiced will find very
little good and a great deal bad in those they despise, no matter how these
people actually behave. In fact, most people are a mixture of good and bad
qualities. By ignoring the former and dwelling on the latter, it is easy enough
for a prejudiced person to confirm negative opinions. Similarly, stereo-
types, which can be either positive or negative, often persist in the face of
224
CHAPTER 8 ■ Arguments To and From General izat ions
overwhelming counterevidence. Criticizing the beliefs common in Britain in
his own day, David Hume remarked:
An Irishman cannot have wit, and a Frenchman cannot have solidity; for which
reason, though the conversation of the former in any instance be very agreeable,
and of the latter very judicious, we have entertained such a prejudice against
them, that they must be dunces and fops in spite of sense and reason.3
Although common stereotypes have changed somewhat since Hume’s day,
prejudice continues to distort the beliefs of many people in our own time.
Another common source of bias in sampling arises from phrasing ques-
tions in ways that encourage certain answers while discouraging others.
Even if a fair sample is asked a question, it is well known that the way a
question is phrased can exert a significant influence on how people will an-
swer it. Questions like the following are not intended to elicit information,
but instead to push people’s answers in one direction rather than another.
Which do you favor: (a) preserving a citizen’s constitutional right to bear
arms or (b) leaving honest citizens defenseless against armed criminals?
Which do you favor: (a) restricting the sale of assault weapons or (b)
knuckling under to the demands of the well-financed gun lobby?
In both cases, one alternative is made to sound attractive, the other unattrac-
tive. When questions of this sort are used, it is not surprising that different
pollsters can come up with wildly different results.
Now we can summarize and restate our questions. Confronted with induc-
tive generalizations, there are four questions that we should routinely ask:
1. Are the premises acceptable?
2. Is the sample too small?
3. Is the sample biased?
4. Are the results affected by other sources of bias?
By asking the preceding questions, specify what, if anything, is wrong with
the following statistical generalizations:
1. This philosophy class is about logic, so most philosophy classes are
probably about logic.
2. Most college students like to ski, because I asked a lot of students at
several colleges in the Rocky Mountains, and most of them like to ski.
3. K-Mart asked all of their customers throughout the country whether they
prefer K-Mart to Walmart, and 90 percent said they did, so 90 percent of
all shoppers in the country prefer K-Mart.
4. A Swede stole my bicycle, so most Swedes are thieves.
Exercise II
225
Stat ist ical Appl icat ions
STATISTICAL APPLICATIONS
In a statistical generalization, we draw inferences concerning a population
from information concerning a sample of that population. If 60 percent of
the population sampled said that they would vote for candidate X, we
might draw the conclusion that roughly 60 percent of the population will
vote for candidate X. With a statistical application (sometimes called a statisti-
cal syllogism), we reason in the reverse direction: From information concern-
ing a population, we draw a conclusion concerning a member or subset of
that population. Here is an example:
Ninety-seven percent of the Republicans in California voted for McCain.
Marvin is a Republican from California.
Marvin voted for McCain.
Such arguments have the following general form:
X percent of Fs have the feature G.
a is an F.
a has the feature G.4
5. I’ve never tried it before, but I just put a kiwi fruit in a tub of water. It
floated. So most kiwi fruits float in water.
6. I have lots of friends. Most of them think that I would make a great
president. So most Americans would probably agree.
7. In exit polls after people had just voted, most people told our candidate
that they voted for her, so probably most people did vote for her.
8. Mary told me that all of her older children are geniuses, so her baby will
probably be a genius, too.
9. When asked whether they would prefer a tax break or a bloated budget,
almost everyone said that they wanted a tax break. So a tax break is over-
whelmingly popular with the people.
10. When hundreds of convicted murderers in states without the death
penalty were asked whether they would have committed the murder if
the state had a death penalty, most of them said that they would not have
done it. So most murders can be deterred by the death penalty.
It is often easy to see that a sample is biased, but how can you tell that a sam-
ple is not biased? How can you determine whether a sample is big enough?
Discussion Question
226
CHAPTER 8 ■ Arguments To and From General izat ions
Obviously, when we evaluate the strength of a statistical application, the
percentage of Fs that have the feature G will be important. As the figure ap-
proaches 100 percent, the argument gains strength. Thus, our original argu-
ment concerning Marvin is quite strong. We can also get strong statistical
applications when the figure approaches 0 percent. The following is a strong
inductive argument:
Three percent of the socialists from California voted for McCain.
Maureen is a socialist from California.
Maureen did not vote for McCain.
Statistical applications of the kind considered here are strong only if the fig-
ures are close to 100 percent or 0 percent. When the percentages are in the
middle of this range, such statistical applications are weak.
A more interesting problem in evaluating the strength of a statistical
application concerns the relevance of the premises to the conclusion. In the
above schematic representation, F stands for what is called the reference class.
In our first example, being a Republican from California is the reference
class; in our second example, being a socialist from California is the refer-
ence class. A striking feature of statistical applications is that using different
reference classes can yield incompatible results. To see this, consider the fol-
lowing example:
Three percent of Obama’s relatives voted for McCain.
Marvin is a relative of Obama.
Marvin did not vote for McCain.
We now have a statistical application that gives us strong support for the
claim that Marvin did not vote for McCain. This is incompatible with our
first statistical application, which gave strong support to the claim that he
did. To overlook this conflict between arguments based on different refer-
ence classes would be a kind of fallacy. Which statistical application, if ei-
ther, should we trust? This will depend on which of the reference classes we
take to be more relevant. Which counts more, political affiliation or family
ties? That might be hard to say.
One way of dealing with competing statistical applications is to combine
the reference classes. We could ask, for example, what percentage of Repub-
licans from California who are relatives of Obama voted for McCain? The re-
sult might come out this way:
Forty-two percent of Republicans from California who were relatives
of Obama voted for McCain.
Marvin is a Republican from California who is a relative of Obama.
Marvin voted for McCain.
This statistical application provides very weak support for its conclusion. In-
deed, it supplies some weak support for the denial of its conclusion—that is,
for the claim that Marvin did not vote for McCain.
227
Stat ist ical Appl icat ions
This situation can be diagrammed with ellipses of varying sizes to repre-
sent the percentages of Californians and relatives of Obama who do or do
not vote for McCain. First, we draw an ellipse to represent Republicans from
California and place a vertical line so that it cuts off roughly (very
roughly!) 97 percent of the area of that ellipse to represent the premise that
97 percent of the Republicans from California voted for McCain:
Voted for McCain
Republicans from California
Did not Vote for McCain
Next, we add a second ellipse to represent Obama’s relatives:
Voted for McCain
Republicans from California
Obama’s
relatives
Did not Vote for McCain
Only about 3 percent of the small ellipse is left of the line to represent the
premise that 3 percent of Obama’s relatives voted for McCain. The area that
lies within both ellipses represents the people who are both Republicans
from California and also relatives of Obama. About 42 percent of that area is
left of the line to represent the premise that 42 percent of Republicans from
California who were relatives of Obama voted for McCain. The whole dia-
gram now shows how all of these premises can be true, even though they
lead to conflicting conclusions.
This series of arguments illustrates in a clear way what we earlier called
the defeasibility of inductive inferences: A strong inductive argument can be
made weak by adding further information to the premises. Given that Mar-
vin is a Republican from California, we seemed to have good reason to think
that he voted for McCain. But when we added to this the additional piece of
information that he was a relative of Obama, the original argument lost most
of its force. And new information could produce another reversal. Suppose
we discover that Marvin, though a relative of Obama, actively campaigned
for McCain. Just about everyone who actively campaigns for a candidate
votes for that candidate, so it seems that we again have good reason for
thinking that Marvin voted for McCain.
It is clear, then, that the way we select our reference classes will affect
the strength of a statistical application. The general idea is that we should
define our reference classes in a way that brings all relevant evidence to
228
CHAPTER 8 ■ Arguments To and From General izat ions
bear on the subject. But this raises difficulties. It is not always obvious
which factors are relevant and which are not. In our example, party affilia-
tion is relevant to how people voted in the 2008 election; shoe size presum-
ably is not. Whether gender is significant, and, if so, how significant, is a
matter for further statistical research.
These difficulties concerning the proper way to fix reference classes re-
flect a feature of all inductive reasoning: To be successful, such reasoning
must take place within a broader framework that helps determine which
features are significant and which features are not. Without this framework,
there would be no reason not to consider shoe size when trying to decide
how someone will vote. This shows how statistical applications, like all of
the other inductive arguments that we will study, cannot work properly
without appropriate background assumptions.
Carry the story of Marvin two steps further, producing two more reversals in
the strength of the statistical application with the conclusion that Marvin
voted for McCain.
Exercise III
For each of the following statistical applications, identify the reference class,
and then evaluate the strength of the argument in terms of the percentages or
proportions cited and the relevance of the reference class.
1. Less than 1 percent of the people in the world voted for McCain.
Michelle is a person in the world.
Michelle did not vote for McCain.
2. Very few teams repeat as Super Bowl champions.
New England was the last Super Bowl champion.
New England will not repeat as Super Bowl champion.
3. A very high percentage of people in the Senate are men.
Hillary Clinton is in the Senate.
Hillary Clinton is a man.
4. Three percent of socialists with blue eyes voted for McCain.
Maureen is a socialist with blue eyes.
Maureen did not vote for McCain.
Exercise IV
229
Stat ist ical Appl icat ions
5. Ninety-eight percent of what John says is true.
John said that his father is also named John.
John’s father is named John.
6. Ninety-eight percent of what John says is true.
John said that the Giants are going to win.
The Giants are going to win.
7. Half the time he doesn’t know what he is doing.
He is eating lunch.
He does not know that he is eating lunch.
8. Most people do not understand quantum mechanics.
My physics professor is a person.
My physics professor probably does not understand quantum mechanics.
9. Almost all birds can fly.
This penguin is a bird.
This penguin can fly.
10. Most people who claim to be psychic are frauds.
Mary claims to be psychic.
Mary is a fraud.
Although both in science and in daily life, we rely heavily on the methods of
inductive reasoning, this kind of reasoning raises a number of perplexing
problems. The most famous problem concerning the legitimacy of induction
was formulated by the eighteenth-century philosopher David Hume, first in
his Treatise of Human Nature and then later in his Enquiry Concerning Human
Understanding. A simplified version of Hume’s skeptical argument goes as fol-
lows: Our inductive generalizations seem to rest on the assumption that unob-
served cases will follow the patterns that we discovered in observed cases. That
is, our inductive generalizations seem to presuppose that nature operates uni-
formly: The way things are observed to behave here and now are accurate
indicators of how things behave anywhere and at any time. But by what right
can we assume that nature is uniform? Because this claim itself asserts a con-
tingent matter of fact, it could only be established by inductive reasoning. But
because all inductive reasoning presupposes the principle that nature is uni-
form, any inductive justification of this principle would seem to be circular. It
seems, then, that we have no ultimate justification for our inductive reasoning
at all. Is this a good argument or a bad one? Why?
Discussion Question
230
CHAPTER 8 ■ Arguments To and From General izat ions
NOTES
1 David Hume, A Treatise of Human Nature, 2nd ed. (1739; Oxford: Oxford University Press,
1978), 146.
2 Amos Tversky and Daniel Kahneman, “Belief in the Law of Small Numbers,” Psychological
Bulletin 76, no. 2 (1971), 105.
3 Hume, A Treatise of Human Nature, 146–47.
4 We can also have a probabilistic version of the statistical syllogism:
Ninety-seven percent of the Republicans from California voted for McCain.
Marvin is a Republican from California.
There is a 97 percent chance that Marvin voted for McCain.
We will discuss arguments concerning probability in Chapter 11.
Causal Reasoning
Statistical generalization can tell us that all ravens are black and that most Texans
will vote for McCain, but these generalizations alone cannot tell us what makes
ravens black or what makes most Texans vote for McCain. To determine what
causes such phenomena, we need to engage in a new kind of inductive reasoning—
causal reasoning—which is the topic of this chapter. We will show how causal
reasoning is often based on negative and positive tests for necessary conditions
and for sufficient conditions. After developing these tests and applying them to a
concrete example, we will discuss concomitant variation as a method of drawing
causal conclusions from imperfect correlations. Our goal throughout this chapter
is to improve our ability to identify causes so that we can better understand why
certain effects happened and also make better predictions about whether similar
events will happen in the future.
REASONING ABOUT CAUSES
If our car goes dead in the middle of rush-hour traffic just after its 20,000-mile
checkup, we assume that there must be some reason why this happened.
Cars just don’t stop for no reason at all. So we ask, “What caused our car to
stop?” The answer might be that it ran out of gas. If we find, in fact, that it
did run out of gas, then that will usually be the end of the matter. We will
think that we have discovered why this particular car stopped running. This
reasoning is about a particular car on a particular occasion, but it rests on cer-
tain generalizations: We are confident that our car stopped running when it ran
out of gas, because we believe that all cars stop running when they run out of
gas. We probably did not think about this, but our causal reasoning in this
particular case appealed to a commonly accepted causal generalization: Lack
of fuel causes cars to stop running. Many explanations depend on causal
generalizations.
Causal generalizations are also used to predict the consequences of partic-
ular actions or events. A race car driver might wonder, for example, what
would happen if he added just a bit of nitroglycerin to his fuel mixture.
Would it give him better acceleration, blow him up, do very little, or what?
9
231
232
CHAPTER 9 ■ Causal Reason ing
In fact, the driver may not be in a position to answer this question straight
off, but his thinking will be guided by the causal generalization that igniting
nitroglycerin can cause a dangerous explosion.
So a similar pattern arises for both causal explanation and causal prediction.
These inferences contain two essential elements:
1. The facts in the particular case. (For example, the car stopped and the
gas gauge reads empty; or I just put a pint of nitroglycerin in the gas
tank of my Maserati, and I am about to turn the ignition key.)
2. Certain causal generalizations. (For example, cars do not run without
gas, or nitroglycerin explodes when ignited.)
The basic idea is that causal inferences bring particular facts under causal
generalizations.
This shows why causal generalizations are important, but what exactly
are they? Although this issue remains controversial, here we will treat them
as a kind of general conditional. A general conditional has the following form:
For all x, if x has the feature F, then x has the feature G.
We will say that, according to this conditional, x’s having the feature F is a
sufficient condition for its having the feature G; and x’s having the feature G is
a necessary condition for its having the feature F.
Some general conditionals are not causal. Neither of these two general
conditionals expresses a causal relationship:
If something is a square, then it is a rectangle.
If you are eighteen years old, then you are eligible to vote.
The first conditional tells us that being a square is sufficient for being a rec-
tangle, but this is a mathematical (or a priori) relationship, not a causal one.
The second conditional tells us that being eighteen years old is a sufficient
condition for being eligible to vote. The relationship here is legal, not causal.
Although many general conditionals are not causal, all causal conditionals
are general, in our view. Consequently, if we are able to show that a causal
conditional is false just by virtue of its being a general conditional, we will
have refuted it. This will serve our purposes well, for in what follows we will
be largely concerned with finding reasons for rejecting causal generalizations.
It is important to weed out false causal generalizations, because they can
create lots of trouble. Doctors used to think that bloodletting would cure dis-
ease. They killed many people in the process of trying to heal them. Thus,
although we need causal generalizations for getting along in the world, we
also need to get them right. We will be more likely to succeed if we have
proper principles for testing and applying such generalizations.
In the past, very elaborate procedures have been developed for this pur-
pose. The most famous set of such procedures was developed by John Stuart
Mill and has come to be known as Mill’s methods.1 Though inspired by
Mill’s methods, the procedures introduced here involve some fundamental
233
Suff ic ient Cond it ions and Necessary Cond it ions
simplifications; whereas Mill introduced five methods, we will introduce
only three primary rules.
The first two rules are the sufficient condition test (SCT) and the necessary
condition test (NCT). We will introduce these tests first at an abstract level.
One advantage of formulating these tests abstractly is so that they can be
applied to other kinds of sufficient and necessary conditions, for example,
those that arise in legal and moral reasoning, the topics of Chapters 18 and 19.
Once it is clear how these tests work in general, we will apply them specifi-
cally to causal reasoning.
SUFFICIENT CONDITIONS AND NECESSARY
CONDITIONS
To keep our discussion as general as possible, we will adopt the following
definitions of sufficient conditions and necessary conditions:
Feature F is a sufficient condition for feature G if and only if anything that
has feature F also has feature G.
Feature F is a necessary condition for feature G if and only if anything that
lacks feature F also lacks feature G.
These definitions are equivalent to those in the previous section, because, if
anything that lacks feature F also lacks feature G, then anything that has feature
G must also have feature F; and if anything that has feature G must also have
feature F, then anything that lacks feature F also lacks feature G. It follows that
feature F is a sufficient condition for feature G if and only if feature G is a nec-
essary condition for feature F.
When F is sufficient for G, the relation between these features can be dia-
grammed like this:
The inside circle represents the sufficient condition, because anything inside
that inside circle must also be inside the outside circle. The outside circle repre-
sents the necessary condition, for anything outside the outside circle must also
be outside the inside circle.
These diagrams, along with the preceding definitions, should make it
clear that something can be a sufficient condition for a feature without
being a necessary condition for that feature, and vice versa. For example,
234
CHAPTER 9 ■ Causal Reason ing
being the element mercury is a sufficient condition for being a metal, but it
is not a necessary condition for being a metal, since there are other metals.
Similarly, being a metal is a necessary condition for being mercury, but it is
not a sufficient condition for being mercury. Of course, some necessary
conditions are also sufficient conditions. Being mercury is both necessary
and sufficient for being a metallic element that is liquid at twenty degrees
Centigrade. Nonetheless, many necessary conditions are not sufficient
conditions, and vice versa, so we need to be careful not to confuse the two
kinds of conditions.
This distinction becomes complicated when conditions get complex. Our
definitions and tests hold for all features, whether positive or negative (such
as not having hair) and whether simple or conjunctive (such as having both
a beard and a mustache) or disjunctive (such as having either a beard or a
mustache). Thus, not having any hair (anywhere) on your head is a suffi-
cient condition of not having a beard, so not having a beard is a necessary
condition of not having any hair on your head. But not having any hair on
your head is not necessary for not having a beard, because you can have
some hair on the top of your head without having a beard. Negation can cre-
ate confusion, so we need to think carefully about what is being claimed to
be necessary or sufficient for what.
Even in simple cases without negation, conjunction, or disjunction, there
is a widespread tendency to confuse necessary conditions with sufficient
conditions. It is important to keep these concepts straight, for, as we will see,
the tests concerning them are fundamentally different.
Which of the following claims are true? Which are false?
1. Being a car is a sufficient condition for being a vehicle.
2. Being a car is a necessary condition for being a vehicle.
3. Being a vehicle is a sufficient condition for being a car.
4. Being a vehicle is a necessary condition for being a car.
5. Being an integer is a sufficient condition for being an even number.
6. Being an integer is a necessary condition for being an even number.
7. Being an integer is a sufficient condition for being either an even number
or an odd number.
8. Being an integer is a necessary condition for being either an even number
or an odd number.
9. Not being an integer is a sufficient condition for not being an odd number.
10. Not being an integer is a sufficient condition for not being an even
number.
11. Being both an integer and divisible by 2 without remainder is a sufficient
condition for being an even number.
Exercise I
235
Suff ic ient Cond it ions and Necessary Cond it ions
12. Being both an integer and divisible by 2 without remainder is a necessary
condition for being an even number.
13. Being an integer divisible by 2 without remainder is a necessary condition
for being an even number.
14. Driving seventy-five miles per hour (for fun) is a sufficient condition for
violating a legal speed limit of sixty-five miles per hour.
15. Driving seventy-five miles per hour (for fun) is a necessary condition for
violating a legal speed limit of sixty-five miles per hour.
16. Cutting off Joe’s head is a sufficient condition for killing him.
17. Cutting off Joe’s head is a necessary condition for killing him.
18. Cutting off Joe’s head and then holding his head under water for ten min-
utes is a sufficient condition for killing him.
Indicate whether the following principles are true or false and why.
1. If having feature F is a sufficient condition for having feature G, then
having feature G is a necessary condition for having feature F.
2. If having feature F is a sufficient condition for having feature G, then
lacking feature F is a necessary condition for lacking feature G.
3. If lacking feature F is a sufficient condition for having feature G, then
having feature F is a necessary condition for lacking feature G.
4. If lacking feature F is a sufficient condition for having feature G, then
lacking feature F is a necessary condition for having feature G.
5. If having either feature F or feature G is a sufficient condition for having
feature H, then having feature F is a sufficient condition for having
feature H.
6. If having either feature F or feature G is a sufficient condition for having
feature H, then having feature G is a sufficient condition for having
feature H.
7. If having either feature F or feature G is a sufficient condition for having
feature H, then not having feature F is a necessary condition for not having
feature H.
8. If having both feature F and feature G is a necessary condition for having
feature H, then lacking feature F is a sufficient condition for lacking feature H.
9. If not having both feature F and feature G is a sufficient condition for
having feature H, then lacking feature F is a sufficient condition for having
feature H.
10. If having either feature F or feature G is a sufficient condition for having
feature H, then having both feature F and feature G is a sufficient condition
for having feature H.
Exercise II
236
CHAPTER 9 ■ Causal Reason ing
THE SUFFICIENT CONDITION TEST
We can now formulate tests to determine when something meets our defini-
tions of sufficient conditions and necessary conditions. It will simplify mat-
ters if we first state these tests formally using letters. We will also begin with
a simple case where we consider only four candidates—A, B, C, and D—for
sufficient conditions for a target feature, G. A will indicate that the feature is
present; ~A will indicate that this feature is absent. Using these conventions,
suppose that we are trying to decide whether any of the four features—A, B,
C, or D—could be a sufficient condition for G. To this end, we collect data of
the following kind:
TABLE 1
Case 1: A B C D G
Case 2: ~A B C ~D ~G
Case 3: A ~B ~C ~D ~G
We know by definition that, for something to be a sufficient condition of
something else, when the former is present, the latter must be present as
well. Thus, to test whether a candidate really is a sufficient condition of G,
we only have to examine cases in which the target feature, G, is absent, and
then check to see whether any of the candidate features are present. The suf-
ficient condition test (SCT) can be stated as follows:
SCT: Any candidate that is present when G is absent is eliminated as a
possible sufficient condition of G.
The test applies to Table 1 as follows: Case 1 need not be examined because G
is present, so there can be no violation of SCT in Case 1. Case 2 eliminates two
of the candidates, B and C, for both are present in a situation in which G is ab-
sent. Finally, Case 3 eliminates A for the same reason. We are thus left with D
as our only remaining candidate for a sufficient condition for G.
Now let’s consider feature D. Having survived the application of the SCT,
does it follow that D is a sufficient condition for G? No! On the basis of what
we have been told so far, it remains entirely possible that the discovery of a
further case will reveal an instance where D is present and G absent, thus
showing that D is also not a sufficient condition for G.
Case 4: ~A B C D ~G
In this way, it is always possible for new cases to refute any inference from a
limited group of cases to the conclusion that a certain candidate is a suffi-
cient condition. In contrast, no further case can change the fact that A, B, and
C are not sufficient conditions, because they fail the SCT.
This observation shows that, when we apply the SCT to rule out a candidate
as a sufficient condition, our argument is deductive. We simply find a counter-
example to the universal claim that a certain feature is sufficient. (See Chapter 17
on counterexamples.) However, when a candidate is not ruled out and we draw
the positive conclusion that that candidate is a sufficient condition, then our
237
Suff ic ient Cond it ions and Necessary Cond it ions
argument is inductive. Inductive inferences, however well confirmed, are always
defeasible. (Recall Captain Cook’s discovery of black swans at the start of Chap-
ter 8.) That is why our inductive inference to the conclusion that D is a sufficient
condition could be refuted by the new data in Case 4.
THE NECESSARY CONDITION TEST
The necessary condition test (NCT) is like the SCT, but it works in the reverse
fashion. With SCT we eliminated a candidate F from being the sufficient con-
dition for G, if F was ever present when G was absent. With the necessary
condition test, we eliminate a candidate F from being a necessary condition
for G if we can find a case where G is present, but F is not. This makes sense,
because if G can be present when F is not, then F cannot be necessary for the
occurrence of G. Thus, in applying the necessary condition test, we only have
to examine cases in which the target feature, G, is present, and then check to
see whether any of the candidate features are absent.
NCT: Any candidate that is absent when G is present is eliminated as a
possible necessary condition of G.
The following table gives an example of an application of this test:
TABLE 2
Case 1: A B C D ~G
Case 2: ~A B C D G
Case 3: A ~B C ~D G
Because Case 1 does not provide an instance where G is present, it cannot
eliminate any candidate as a necessary condition of G. Case 2 eliminates A
as a necessary condition of G, since it shows that G can be present without A
being present. Case 3 then eliminates both B and D, leaving C as the only
possible candidate for being a necessary condition for G.
From this, of course, it does not follow that C is a necessary condition for G,
for, as always, new cases might eliminate it as well. The situation is the same as
with the SCT. An argument for a negative conclusion that a candidate is not a
necessary condition, because that candidate fails the NCT, is a deductive argu-
ment that cannot be overturned by any further cases. In contrast, an argument
for a positive conclusion that a candidate is a necessary condition, because that
candidate passes the NCT, is an inductive argument that can be overturned
by a further case where this candidate fails the NCT. For example, suppose
we find:
Case 4: ~A ~B ~C ~D G
The information in this new Case 4 is enough to show that C cannot be a neces-
sary condition of the target feature G, regardless of what we found in Cases 1–3.
In applying both the SCT and the NCT, it is crucial to specify the target
feature. Case 4 shows that candidate C is not a necessary condition for target
feature G. Nonetheless, candidate C still might be necessary for the opposite
238
CHAPTER 9 ■ Causal Reason ing
target feature, ~G. It also might be necessary for features A, B, and D. Nothing
in Cases 1–4 rules out these possibilities. Thus, even after Case 4, we cannot
say simply that C is not a necessary condition. Case 4 shows that candidate
feature C is not a necessary conditions for target feature G, but C still might be
necessary for something else. The same point applies to sufficient conditions
as well. In Table 1, Case 2 ruled out the possibility that candidate feature B
is sufficient for target feature G, but none of the cases in Table 1 show that B
is not sufficient for target feature C. To avoid confusion, then, it is always
important to specify the target feature when talking about what is or is not a
necessary or sufficient condition.
THE JOINT TEST
It is also possible to apply these rules simultaneously in the search for possible
conditions that are both sufficient and necessary. Any candidate cannot be both
sufficient and necessary if it fails either the SCT or the NCT. In Table 2, C is the
only possible necessary condition for G, and it is not also a possible sufficient
condition for G, since C fails the SCT in Case 1, where C is present and G is
absent. In Table 1, however, D is a possible sufficient condition of G, because D
is never present when G is absent; and D might also be a necessary condition
for G, since G is never present when D is absent. Thus, none of Cases 1–3 in
Table 1 eliminates D as a candidate for a condition that is both sufficient and
necessary for G. As before, this possibility still might be refuted by Case 4, so
any inference to a positive conclusion that some candidate is a necessary
and sufficient condition must be defeasible and, hence, inductive.
For each of the following tables determine
a. Which, if any, of the candidates—A, B, C, or D—is not eliminated by the
sufficient condition test as a sufficient condition for target feature G?
b. Which, if any, of the candidates—A, B, C, or D—is not eliminated by the
necessary condition test as a necessary condition for target feature G?
c. Which, if any, of the candidates—A, B, C, or D—is not eliminated by
either test?
EXAMPLE: Case 1: A B �C D �G
Case 2: �A B C D G
Case 3: A �B C D G
a. Only C passes the SCT.
b. Only C and D pass the NCT.
c. Only C passes both tests.
Exercise III
239
Suff ic ient Cond it ions and Necessary Cond it ions
1. Case 1: A B C D G
Case 2: �A B �C D �G
Case 3: A �B C �D G
2. Case 1: A B C �D G
Case 2: �A B C D G
Case 3: A �B C �D G
3. Case 1: A B C D �G
Case 2: �A B C D G
Case 3: A �B C �D G
4. Case 1: A B �C D G
Case 2: �A �B C D G
Case 3: A B �C �D �G
5. Case 1: A �B C D �G
Case 2: �A B C �D �G
Case 3: A �B �C D G
6. Case 1: A B �C D G
Case 2: �A �B C D �G
Case 3: A �B C �D �G
7. Case 1: A B �C D �G
Case 2: �A B �C D �G
Case 3: A B �C �D �G
8. Case 1: A B C D �G
Case 2: �A �B C D G
Case 3: A �B �C �D �G
Imagine that your desktop computer system won’t work, and you want to find
out why. After checking to make sure that it is plugged in, you experiment
with a new central processing unit (CPU), a new monitor (MON), and new
system software (SSW) in the combinations on the table below. The candidates
for necessary conditions and sufficient conditions of failure are the plug posi-
tion (in or out), the CPU (old or new), the monitor (old or new), and the soft-
ware (old or new). For each candidate, say (1) which cases, if any, eliminate it
as a sufficient condition of your computer’s failure and (2) which cases, if any,
eliminate it as a necessary condition of your computer’s failure. Which candi-
dates, if any, are not eliminated as a sufficient condition of failure? As a neces-
sary condition of failure? Does it follow that these candidates are necessary
conditions or sufficient conditions of failure? Why or why not?
Exercise IV
(continued)
240
CHAPTER 9 ■ Causal Reason ing
Plug CPU Monitor Software Result
Case 1 ln Old CPU Old MO Old SW Works
Case 2 ln Old CPU Old MO New SW Works
Case 3 ln Old CPU New MO Old SW Fails
Case 4 ln Old CPU New MO New SW Works
Case 5 ln Old CPU Old MO Old SW Works
Case 6 ln Old CPU Old MO New SW Works
Case 7 ln Old CPU New MO Old SW Fails
Case 8 ln Old CPU New MO New SW Works
Case 9 ln New CPU Old MO Old SW Fails
Case 10 ln New CPU Old MO New SW Works
Case 11 ln New CPU New MO Old SW Fails
Case 12 ln New CPU New MO New SW Works
After a banquet, several diners get sick and die. You suspect that something they
ate or drank caused their deaths. The following table records their meals and
fates. The target feature is death. The candidates for necessary conditions and suf-
ficient conditions of death are the soup, entrée, wine, and dessert. For each candi-
date, say (1) which cases, if any, eliminate it as a sufficient condition of death and
(2) which cases, if any, eliminate it as a necessary condition of death. Which can-
didates, if any, are not eliminated as a sufficient condition of death? Which candi-
dates, if any, are not eliminated as a necessary condition of death? Does it follow
that these candidates are necessary conditions or sufficient conditions of death?
Why or why not?
Diners Soup Entrée Wine Dessert Result
Ann Tomato Chicken White Pie Alive
Barney Tomato Fish Red Cake Dead
Cathy Tomato Beef Red Ice Cream Alive
Doug Tomato Beef Red Cake Alive
Emily Tomato Fish Red Pie Dead
Fred Tomato Fish Red Cake Dead
Gertrude Leek Fish White Pie Alive
Harold Tomato Beef White Cake Alive
Irma Leek Fish Red Pie Dead
Jack Leek Beef Red Ice Cream Alive
Ken Leek Chicken Red Ice Cream Alive
Leslie Tomato Chicken White Cake Alive
Exercise V
RIGOROUS TESTING
Going back to Table 1, it is easy to see that candidates A, B, C, and D are not
eliminated by the NCT as necessary conditions of target G, as G is present in
only one case (Case 1) and A, B, C, and D are present there as well. So far, so
241
Suff ic ient Cond it ions and Necessary Cond it ions
good. But if we wanted to test these features more rigorously, it would be
important to find more cases in which target G was present and see whether
these candidates are also present and thus continue to survive the NCT.
The following table gives a more extreme example of nonrigorous testing:
TABLE 3
Case 1: A ~B C D G
Case 2: A ~B ~C ~D ~G
Case 3: A ~B C ~D ~G
Case 4: A ~B ~C D G
Here candidate feature A is eliminated by SCT (in Cases 2 and 3) but is
not eliminated by NCT, so it is a possible necessary condition but not a
possible sufficient condition for target feature G. B is not eliminated by
SCT but is eliminated by NCT (in Cases 1 and 4), so it is a possible suffi-
cient condition but not a possible necessary condition for target feature G.
C is eliminated by both rules (in Cases 3 and 4). Only D is not eliminated
by either test, so it is the only candidate for being both a necessary and a
sufficient condition for G.
The peculiarity of this example is that candidate A is always present
whether target G is present or not, and candidate B is always absent whether
target G is absent or not. Now if something is always present, as A is, then it
cannot possibly fail the NCT; for there cannot be a case where the target is
present and the candidate is absent if the candidate is always present. If we
want to test candidate A rigorously under the NCT, then we should try to find
cases in which A is absent and then check to see whether G is absent as well.
In reverse fashion, but for similar reasons, if we want to test candidate B
rigorously under the SCT, then we should try to find cases in which B is
present and then check to see if G is present as well. If we restrict our atten-
tion to cases where B is always absent, as in Table 3, then B cannot possibly
fail the SCT, but passing that test will be trivial for B and so will not even
begin to show that B is a sufficient condition for G.
Now consider two more sets of data just like Table 2, except with regard
to the target feature, G:
TABLE 4
Case 1: A B C D G
Case 2: ~A B C D G
Case 3: A ~B C ~D G
TABLE 5
Case 1: A B C D ~G
Case 2: ~A B C D ~G
Case 3: A ~B C ~D ~G
Because G is present in all of the cases in Table 4, no candidate can be
eliminated by the SCT as a sufficient condition for target feature G. This
242
CHAPTER 9 ■ Causal Reason ing
result is trivial, however. Table 4 does not provide rigorous testing for a
sufficient condition of G, because our attention is restricted to a range of
cases that is too narrow. Nothing could possibly be eliminated as a suffi-
cient condition of G as long as G is always present.
Similarly, G is absent in all of the cases in Table 5, so no candidate can be
eliminated by the NCT as a necessary condition of target feature G. Still, be-
cause this data is so limited, its failure to eliminate candidates does not even
begin to show that anything is a necessary condition of G.
For both rules, then, rigorous testing involves seeking out cases in which
failing the test is a live possibility. For the SCT, this requires looking both at
cases in which the candidates are present and also at cases in which the target
is absent. For the NCT, rigorous testing requires looking both at cases in which
the candidates are absent and also at cases in which the target is present. With-
out cases like these, passing the tests is rather like a person bragging that he has
never struck out when, in fact, he has never come up to bat.
REACHING POSITIVE CONCLUSIONS
Suppose that we performed rigorous testing on candidate C, and it passed
the SCT with flying colors. Can we now draw the positive conclusion that C
is a sufficient condition for the target G? That depends on which kinds of
candidates and cases have been considered. Since rigorous testing was
passed, these three conditions are met:
1. We have tested some cases in which the candidate, C, is present.
2. We have tested some cases in which the target, G, is absent.
3. We have not found any case in which the candidate, C, is present and
the target, G, is absent.
For it to be reasonable to reach a positive conclusion that C is sufficient for
G, this further condition must also be met:
4. We have tested enough cases of the various kinds that are likely to
include a case in which C is present and G is absent if there is any
such case.
This new condition cannot be applied in the mechanical way that condi-
tions 1–3 could be applied. To determine whether condition 4 is met, we
need to rely on background information about how many cases are “enough”
and about which kinds of cases “are likely to include a case in which C is
present and G is absent, if there is any such case.” For example, if we are
trying to figure out whether our new software is causing our computer to
crash, we do not need to try the same kind of computer in different colors.
What we need to try are different kinds of CPUs, monitors, software, and
so on, because we know that these are the kinds of factors that can affect
performance. Background information like this is what tells us when we
have tested enough cases of the right kinds.
Of course, our background assumptions might turn out to be wrong. Even
if we have tested many variations of every feature that we think might be
243
Apply ing These Methods to F ind Causes
relevant, we still might be surprised and find a further case in which C and
~G are present. All that shows, however, is that our inference is defeasible,
like all inductive arguments. Despite the possibility that future discoveries
might undermine it, our inductive inference can still be strong if our back-
ground beliefs are justified and if we have looked long and hard without
finding any case in which C is present and G is absent.
Similar rules apply in reverse to positive conclusions about necessary
conditions. We have good reason to suppose that candidate C is a necessary
condition for target G, if the following conditions are met:
1. We have tested some cases in which the candidate, C, is absent.
2. We have tested some cases in which the target, G, is present.
3. We have not found any case in which the candidate, C, is absent and
the target, G, is present.
4. We have tested enough cases of the various kinds that are likely to
include a case in which C is absent and G is present, if there is any
such case.
This argument again depends on background assumptions in determining
whether condition 4 is met. This argument is also defeasible, as before.
Nonetheless, if our background assumptions are justified, the fact that
conditions 1–4 are all met can still provide a strong reason for the positive
conclusion that candidate C is a necessary condition for target G.
The SCT and NCT themselves are still negative and deductive; but that does
not make them better than the positive tests encapsulated in conditions 1–4.
The negative SCT and NCT are of no use when we need to argue that some
condition is sufficient or is necessary. Such positive conclusions can be reached
only by applying something like condition 4, which will require background
information. These inductive arguments might not be as clear-cut or secure
as the negative ones, but they can still be inductively strong under the right
circumstances. That is all they claim to be.
APPLYING THESE METHODS TO FIND CAUSES
In stating the SCT and NCT and applying these tests to abstract patterns of
conditions to eliminate candidates, our procedure was fairly mechanical. We
cannot be so mechanical when we try to reach positive conclusions that certain
conditions are necessary, sufficient, or both. Applying these rules to actual con-
crete situations introduces a number of further complications, especially when
using our tests to determine causes.
NORMALITY
First, it is important to keep in mind that, in our ordinary understanding of
causal conditions, we usually take it for granted that the setting is normal. It
is part of common knowledge that if you strike a match, then it will light.
244
CHAPTER 9 ■ Causal Reason ing
Thus, we consider striking a match sufficient to make it light. But if someone
has filled the room with carbon dioxide, then the match will not light, no
matter how it is struck. Here one may be inclined to say that, after all, strik-
ing a match is not sufficient to light it. We might try to be more careful and
say that if a match is struck and the room is not filled with carbon dioxide,
then it will light. But this new conditional overlooks other possibilities—for
example, that the room has been filled with nitrogen, that the match has been
fireproofed, that the wrong end of the match was struck, that the match has
already been lit, and so forth. It now seems that the antecedent of our condi-
tional will have to be endlessly long in order to specify a true or genuine suf-
ficient condition. In fact, however, we usually feel quite happy with saying
that if you strike a match, then it will light. We simply do not worry about the
possibility that the room has been filled with carbon dioxide, the match has
been fireproofed, and so on. Normally we think that things are normal, and
give up this assumption only when some good reason appears for doing so.
These reflections suggest the following contextualized restatement of our
original definitions of sufficient conditions and necessary conditions:
F is a sufficient condition for G if and only if, whenever F is present in a
normal context, G is present there as well.
F is a necessary condition for G if and only if, whenever F is absent from
a normal context, G is absent from it as well.
What will count as a normal context will vary with the type and the aim of
an investigation, but all investigations into causally sufficient conditions
and causally necessary conditions take place against the background of
many factors that are taken as fixed.
BACKGROUND ASSUMPTIONS
If we are going to subject a causal hypothesis to rigorous testing with the SCT
and the NCT, we have to seek out a wide range of cases that might refute that
hypothesis. In general, the wider the range of possible refuters the better. Still,
some limit must be put on this activity or else testing will get hopelessly
bogged down. If we are testing a drug to see whether it will cure a disease, we
should try it on a variety of people of various ages, medical histories, body
types, and so on, but we will not check to see whether it works on people
named Edmund or check to see whether it works on people who drive Volvos.
Such factors, we want to say, are plainly irrelevant. But what makes them
irrelevant? How do we distinguish relevant from irrelevant considerations?
The answer to this question is that our reasoning about causes occurs within
a framework of beliefs that we take to be established as true. This framework
contains a great deal of what is called common knowledge—knowledge we
expect almost every sane adult to possess. We all know, for example, that
human beings cannot breathe underwater, cannot walk through walls, cannot
be in two places at once, and so on. The stock of these commonplace beliefs is
almost endless. Because they are commonplace beliefs, they tend not to be
245
Apply ing These Methods to F ind Causes
mentioned; yet they play an important role in distinguishing relevant factors
from irrelevant ones.
Specialized knowledge also contains its own principles that are largely
taken for granted by experts. Doctors, for example, know a great deal about
the detailed structure of the human body, and this background knowledge
constantly guides their thought in dealing with specific illnesses. Even if
someone claimed to discover that blood does not circulate, no doctor would
take the time to refute that claim.
It might seem close-minded to refuse to consider a possibility that some-
one else suggests. However, giving up our basic beliefs can be very costly. A
doctor who took seriously the suggestion that blood does not circulate, for
example, would have to abandon our whole way of viewing humans and
other animals, along with the rest of biology and science. It is not clear how
this doctor could go on practicing medicine. Moreover, there is usually no
practical alternative in real life. When faced with time pressure and limited
information, we have no way to judge new ideas without taking some back-
ground assumptions for granted.
A DETAILED EXAMPLE
To get a clearer idea of the complex interplay between our tests and the re-
liance on background information, it will be helpful to look in some detail at
actual applications of these tests. For this purpose, we will examine an at-
tempt to find the cause of a particular phenomenon, an outbreak of what
came to be known as Legionnaires’ disease. The example not only shows
how causal reasoning relies on background assumptions, it has another
interesting feature as well: In the process of discovering the cause of Legion-
naires’ disease, the investigators were forced to abandon what was previ-
ously taken to be a well-established causal generalization. In fact, until it
was discarded, this false background principle gave them no end of trouble.
The story began at an otherwise boring convention:
The 58th convention of the American Legion’s Pennsylvania Department was
held at the Bellevue-Stratford Hotel in Philadelphia from July 21 through 24,
1976. . . . Between July 22 and August 3, 149 of the conventioneers developed
what appeared to be the same puzzling illness, characterized by fever, coughing
and pneumonia. This, however, was an unusual, explosive outbreak of pneumo-
nia with no apparent cause. . . . Legionnaires’ disease, as the illness was quickly
named by the press, was to prove a formidable challenge to epidemiologists
and laboratory investigators alike.2
Notice that at this stage the researchers begin with the assumption that they
are dealing with a single illness and not a collection of similar but different
illnesses. That assumption could turn out to be wrong; but, if the symptoms
of the various patients are sufficiently similar, this is a natural starting as-
sumption. Another reasonable starting assumption is that this illness had
a single causative agent. This assumption, too, could turn out to be false,
though it did not. The assumption that they were dealing with a single
246
CHAPTER 9 ■ Causal Reason ing
disease with a single cause was at least a good simplifying assumption, one
to be held onto until there was good reason to give it up. In any case, we
now have a clear specification of our target feature, G: the occurrence of a
carefully described illness that came to be known as Legionnaires’ disease.
The situation concerning it was puzzling because people had contracted a
disease with symptoms much like those of pneumonia, yet they had not
tested positive for any of the known agents that cause such diseases.
The narrative continues as follows:
The initial step in the investigation of any epidemic is to determine the character
of the illness, who has become ill and just where and when. The next step is to
find out what was unique about the people who became ill: where they were and
what they did that was different from other people who stayed well. Knowing
such things may indicate how the disease agent was spread and thereby suggest
the identity of the agent and where it came from.
Part of this procedure involves a straightforward application of the NCT:
Was there any interesting feature that was always present in the history of
people who came down with the illness? Progress was made almost at once
on this front:
We quickly learned that the illness was not confined to Legionnaires. An
additional 72 cases were discovered among people who had not been directly
associated with the convention. They had one thing in common with the sick
conventioneers: for one reason or another they had been in or near the
Bellevue-Stratford Hotel.
Strictly speaking, of course, all these people who had contracted the dis-
ease had more than one thing in common. They were, for example, all
alive at the time they were in Philadelphia, and being alive is, in fact, a
necessary condition for getting Legionnaires’ disease. But the researchers
were not interested in this necessary condition because it is a normal
background condition for the contraction of any disease. Furthermore, it
did not provide a condition that distinguished those who contracted the
disease from those who did not. The overwhelming majority of people
who were alive at the time did not contract Legionnaires’ disease. Thus,
the researchers were not interested in this necessary condition because it
would fail so badly when tested by the SCT as a sufficient condition. On
the basis of common knowledge and specialized medical knowledge, a
great many other conditions were also kept off the candidate list.
One prime candidate on the list was presence at the Bellevue-Stratford
Hotel. The application of the NCT to this candidate was straightforward.
Everyone who had contracted the disease had spent time in or near that
hotel. Thus, presence at the Bellevue-Stratford could not be eliminated as a
necessary condition of Legionnaires’ disease.
The application of the SCT was more complicated, because not everyone
who stayed at the Bellevue-Stratford contracted the disease. Other factors
made a difference: “Older conventioneers had been affected at a higher rate
247
Apply ing These Methods to F ind Causes
than younger ones, men at three times the rate for women.” Since some
young women (among others) who were present at the Bellevue-Stratford
did not get Legionnaires’ disease, presence at that hotel could be eliminated
as a sufficient condition of Legionnaires’ disease. Nonetheless, it is part of
medical background knowledge that susceptibility to disease often varies
with age and gender. Given these differences, some people who spent time
at the Bellevue-Stratford were at higher risk for contracting the disease than
others. The investigation so far suggested that, for some people, being at
the Bellevue-Stratford was connected with a sufficient condition for con-
tracting Legionnaires’ disease. Indeed, the conjunction of spending time at
the Bellevue-Stratford and being susceptible to the disease could not be
ruled out by the SCT as a sufficient condition of getting the disease.
As soon as spending time at the Bellevue-Stratford became the focus of
attention, other hypotheses naturally suggested themselves. Food poisoning
was a reasonable suggestion, since it is part of medical knowledge that diseases
are sometimes spread by food. It was put on the list of possible candidates, but
failed. Investigators checked each local restaurant and each function where
food and drink were served. Some of the people who ate in each place did not
get Legionnaires’ disease, so the food at these locations was eliminated by the
SCT as a sufficient condition of Legionnaires’ disease. These candidates were
also eliminated by the NCT as necessary conditions because some people who
did get Legionnaires’ disease did not eat at each of these restaurants and func-
tions. Thus, the food and drink could not be the cause.
Further investigation turned up another important clue to the cause of
the illness.
Certain observations suggested that the disease might have been spread through
the air. Legionnaires who became ill had spent on the average about 60 percent
more time in the lobby of the Bellevue-Stratford than those who remained well;
the sick Legionnaires had also spent more time on the sidewalk in front of the
hotel than their unaffected fellow conventioneers. . . . It appeared, therefore, that
the most likely mode of transmission was airborne.
Merely breathing air in the lobby of the Bellevue-Stratford Hotel still could not
be a necessary or sufficient condition, but the investigators reasoned that
something in the lobby air probably caused Legionnaires’ disease, since the
rate of the disease varied up or down in proportion to the time spent in the
lobby (or near it on the sidewalk in front). This is an application of the method
of concomitant variation, which will be discussed soon.
Now that the focus was on the lobby air, the next step was to pinpoint a
specific cause in that air. Again appealing to background medical knowl-
edge, there seemed to be three main candidates for the airborne agents
that could have caused the illness: “heavy metals, toxic organic substances,
and infectious organisms.” Examination of tissues taken from patients who
had died from the disease revealed “no unusual levels of metallic or toxic
organic substances that might be related to the epidemic,” so this left an
infectious organism as the remaining candidate. Once more we have an
248
CHAPTER 9 ■ Causal Reason ing
application of NCT. If the disease had been caused by heavy metals or toxic
organic substances, then there would have been unusually high levels of
these substances in the tissues of those who had contracted the disease. Be-
cause this was not always so, these candidates were eliminated as necessary
conditions of the disease.
Appealing to background knowledge once more, it seemed that a bac-
terium would be the most likely source of an airborne disease with the
symptoms of Legionnaires’ disease. But researchers had already made a rou-
tine check for bacteria that cause pneumonia-like diseases, and they had
found none. For this reason, attention was directed to the possibility that
some previously unknown organism had been responsible but had some-
how escaped detection.
It turned out that an undetected and previously unknown bacterium had
caused the illness, but it took more than four months to find this out. The
difficulties encountered in this effort show another important fact about the
reliance on a background assumption: Sometimes it turns out to be false. To
simplify, the standard way to test for the presence of bacteria is to try to
grow them in culture dishes—flat dishes containing nutrients that bacteria
can live on. If, after a reasonable number of tries, a colony of a particular kind
of bacterium does not appear, then it is concluded that the bacterium is not
present. As it turned out, the bacterium that caused Legionnaires’ disease
would not grow in the cultures commonly used to detect the presence of
bacteria. Thus, an important background assumption turned out to be false.
After a great deal of work, a suspicious bacterium was detected using a
live-tissue culture rather than the standard synthetic culture. The task,
then, was to show that this particular bacterium in fact caused the disease.
Again to simplify, when people are infected by a particular organism, they
often develop antibodies that are specifically aimed at this organism. In
the case of Legionnaires’ disease, these antibodies were easier to detect
than the bacterium itself. They also remained in the patients’ bodies after
the infection had run its course. We thus have another chance to apply the
NCT: If Legionnaires’ disease was caused by this particular bacterium,
then whenever the disease was present, this antibody should be present as
well. The suspicious bacterium passed this test with flying colors and was
named, appropriately enough, Legionella pneumophila. Because the investi-
gators had worked so hard to test such a wide variety of candidates, they
assumed that the disease must have some cause among the candidates that
they checked. So, since only one candidate remained, they felt justified in
reaching a positive conclusion that the bacterium was a necessary condi-
tion of Legionnaires’ disease.
The story of the search for the cause of Legionnaires’ disease brings out
two important features of the use of inductive methods in the sciences. First,
it involves a complicated interplay between what is already established and
what is being tested. Confronted with a new problem, established principles
can be used to suggest theoretically significant hypotheses to be tested. The
tests then eliminate some hypotheses and leave others. If, at the end of the
249
Apply ing These Methods to F ind Causes
investigation, a survivor remains that fits in well with our previously estab-
lished principles, then the stock of established principles is increased. The
second thing that this example shows is that the inductive method is fallible.
Without the background of established principles, the application of induc-
tive principles like the NCT and the SCT would be undirected; yet sometimes
these established principles let us down, for they can turn out to be false. The
discovery of the false background principle that hindered the search for
the cause of Legionnaires’ disease led to important revisions in laboratory
techniques. The discovery that certain fundamental background principles
are false can lead to revolutionary changes in science. (See Chapter 20.)
CALLING THINGS CAUSES
After their research was finally completed, with the bacterium identified,
described, and named, it was then said that Legionella pneumophila was the
cause of Legionnaires’ disease. What was meant by this? To simplify a bit,
suppose L. pneumophila (as it is abbreviated) entered the bodies of all those
who contracted the disease: Whenever the disease was present, L. pneumophila
was present. Thus, L. pneumophila passes the NCT for the disease. We will
further suppose, as is common in bacterial infections, that some people’s
immune systems were successful in combating L. pneumophila, and they
never actually developed the disease. Thus, the presence of L. pneumophila
would not pass the SCT for the disease. This suggests that we sometimes call
something a cause of an effect if it passes the NCT for that effect, even if it
does not pass the SCT for that effect.
But even if we sometimes consider necessary conditions to be causes, we
certainly do not consider all necessary conditions to be causes. We have already
noted that to get Legionnaires’ disease, one has to be alive, yet no one thinks
that being alive is the cause of Legionnaires’ disease. To cite another example,
this time one that is not silly, it might be that another necessary condition for de-
veloping Legionnaires’ disease is that the person be in a run-down condition—
healthy people might always be able to resist L. pneumophila. Do we then want
to say that being in a run-down condition is the cause of Legionnaires’ disease?
As we have described the situation, almost certainly not, but we might want to
say that it is an important causal factor or causally relevant factor.
Although the matter is far from clear, what we call the cause rather than sim-
ply a causal factor or causally relevant factor seems to depend on a number of
considerations. We tend to reserve the expression “the cause” for changes that
occur prior to the effect, and describe permanent or standing features of the con-
text as “causal factors” instead. That is how we speak about Legionnaires’ dis-
ease. Being exposed to L. pneumophila, which was a specific event that occurred
before the onset of the disease, caused it. Being in a run-down condition, which
was a feature that patients possessed for some time before they contracted the
disease, was not called the cause, but instead called a causal factor.
It is not clear, however, that we always draw the distinction between
what we call the cause and what we call a causal factor based on whether
250
CHAPTER 9 ■ Causal Reason ing
something is a prior event or a standing condition. For example, if we are
trying to explain why certain people who came in contact with L. pneu-
mophila contracted the disease whereas others did not, then we might say
that the former group contracted the disease because they were in a run-
down condition. Thus, by limiting our investigation only to those who
came in contact with L. pneumophila, our perspective has changed. We
want to know why some within that group contracted the disease and
others did not. Citing the run-down condition of those who contracted the
disease as the cause now seems entirely natural. These examples suggest
that we call something the cause when it plays a particularly important
role relative to the purposes of our investigation. Usually this will be an
event or change taking place against the background of fixed necessary
conditions; sometimes it will not.
Sometimes we call sufficient conditions causes. We say that short circuits
cause fires because in many normal contexts, a short circuit is sufficient to
cause a fire. Of course, short circuits are not necessary to cause a fire, because,
in the same normal contexts, fires can be caused by a great many other
things. With sufficient conditions, as with necessary conditions, we often
draw a distinction between what we call the cause as opposed to what we
call a causal factor, and we seem to draw it along similar lines. Speaking
loosely, we might say that we sometimes call the key components of sufficient
conditions causes. Then, holding background conditions fixed, we can use the
SCT to evaluate such causal claims.
In sum, we can use the NCT to eliminate proposed necessary causal con-
ditions. We can use the SCT to eliminate proposed sufficient causal condi-
tions. Those candidates that survive these tests may be called causal
conditions or causal factors if they fit in well with our system of other
causal generalizations. Finally, some of these causal conditions or causal
factors will be called causes if they play a key role in our causal investiga-
tions. Typically, though not always, we call something the cause of an event
if it is a prior event or change that stands out against the background of
fixed conditions.
Reread the passage on what killed the dinosaurs in the Discussion Question at
the end of Chapter 1. Where do the authors use the NCT? Where do they use
the SCT? Where do they rely on background assumptions?
Discussion Question
CONCOMITANT VARIATION
The use of the sufficient condition test and the necessary condition test de-
pends on certain features of the world being sometimes present and some-
times absent. Some features of the world, however, are always present to
251
Concomitant Var iat ion
some degree. Because they are always present, the NCT will never elimi-
nate them as possible necessary conditions of any event, and the SCT will
never eliminate anything as a sufficient condition for them. Yet the extent
or degree to which a feature exists in the world is often a significant phe-
nomenon that demands causal explanation.
An example should make this clear. In recent decades, a controversy has
raged over the impact of acid rain on the environment of the northeastern
United States and Canada. Part of the controversy involves the proper inter-
pretation of the data that have been collected. The controversy has arisen for
the following reason: The atmosphere always contains a certain amount of
acid, much of it from natural sources. It is also known that an excess of acid in
the environment can have severe effects on both plants and animals. Lakes are
particularly vulnerable to the effects of acid rain. Finally, it is also acknowl-
edged that industries, mostly in the Midwest, discharge large quantities of
sulfur dioxide (SO2) into the air, and this increases the acidity of water in the
atmosphere. The question—and here the controversy begins—is whether the
contribution of acid from these industries is the cause of the environmental
damage downwind of them.
How can we settle such a dispute? The two rules we have introduced pro-
vide no immediate help, for, as we have seen, they provide a rigorous test of a
causal hypothesis only when we can find contrasting cases with the presence
or the absence of a given feature. The NCT provides a rigorous test for a nec-
essary condition only if we can find cases in which the feature does not occur
and then check to make sure that the target feature does not occur either. The
SCT provides a rigorous test for a sufficient condition only when we can find
cases in which the target phenomenon is absent and then check whether the
candidate sufficient condition is absent as well. In this case, however, neither
check applies, for there is always a certain amount of acid in the atmosphere,
so it is not possible to check what happens when atmospheric acid is com-
pletely absent. Similarly, environmental damage, which is the target phenom-
enon under investigation, is so widespread in our modern industrial society
that it is also hard to find a case in which it is completely absent.
So, if there is always acid in the atmosphere, and environmental damage
always exists at least to some extent, how can we determine whether the SO2
released into the atmosphere is significantly responsible for the environmental
damage in the affected areas? Here we use what John Stuart Mill called the
Method of Concomitant Variation. We ask whether the amount of environmental
damage varies directly in proportion to the amount of SO2 released into the
environment. If environmental damage increases with the amount of SO2
released into the environment and drops when the amount of SO2 is lowered,
this means that the level of SO2 in the atmosphere is positively correlated with
environmental damage. We would then have good reason to believe that low-
ering SO2 emissions would lower the level of environmental damage, at least
to some extent.
Arguments relying on the method of concomitant variation are difficult
to evaluate, especially when there is no generally accepted background
252
CHAPTER 9 ■ Causal Reason ing
theory that makes sense of the concomitant variation. Some such variations
are well understood. For example, most people know that the faster you
drive, the more gasoline you consume. (Gasoline consumption varies di-
rectly with speed.) Why? There is a good theory here: It takes more energy to
drive at a high speed than at a low speed, and this energy is derived from
the gasoline consumed in the car’s engine. Other correlations are less well
understood. There seems to be a correlation between how much a woman
smokes during pregnancy and how happy her children are when they reach
age thirty. The correlation here is not nearly as good as the correlation be-
tween gasoline consumption and speed, for many people are very happy at
age thirty even though their mothers smoked a lot during pregnancy, and
many others are very unhappy at age thirty even though their mothers
never smoked. Furthermore, no generally accepted background theory has
been found to explain the correlation that does exist.
This reference to background theory is important, because two sets of
phenomena can be correlated to a very high degree, even with no direct
causal relationship between them. A favorite example that appears in many
statistics texts is the discovered positive correlation in boys between foot
size and quality of handwriting. It is hard to imagine a causal relation hold-
ing in either direction. Having big feet should not make you write better
and, just as obviously, writing well should not give you big feet. The correct
explanation is that both foot size and handwriting ability are positively
correlated with age. Here a noncausal correlation between two phenomena
(foot size and handwriting ability) is explained by a third common correla-
tion (maturation) that is causal.
At times, it is possible to get causal correlations backward. For example, a
few years ago, sports statisticians discovered a negative correlation between
forward passes thrown and winning in football. That is, the more forward
passes a team threw, the less chance it had of winning. This suggested that
passing is not a good strategy, since the more you do it, the more likely you
are to lose. Closer examination showed, however, that the causal relation-
ship, in fact, went in the other direction. Toward the end of a game, losing
teams tend to throw a great many passes in an effort to catch up. In other
words, teams throw a lot of passes because they are losing, rather than the
other way around.
Finally, some correlations seem inexplicable. For example, a strong posi-
tive correlation reportedly holds between the birth rate in Holland and the
number of storks nesting in chimneys. There is, of course, a background the-
ory that would explain this—storks bring babies—but that theory is not fa-
vored by modern science. For the lack of any better background theory, the
phenomenon just seems weird.
So, given a strong correlation between phenomena of types A and B, four
possibilities exist:
1. A is the cause of B.
2. B is the cause of A.
253
Concomitant Var iat ion
3. Some third thing is the cause of both.
4. The correlation is simply accidental.
Before we accept any one of these possibilities, we must have good reasons
for preferring it over the other three.
One way to produce such a reason is to manipulate A or B. If we vary fac-
tor A up and down, but B does not change at all, this finding provides some
reason against possibility 1, since B would normally change along with A if
A did cause B. Similarly, if we manipulate B up and down, but A does not
vary at all, this result provides some reason against alternative 2 and for the
hypothesis that that B does not cause A. Together these manipulations can
reduce the live options to items 3 and 4.
Many scientific experiments work this way. When scientists first discov-
ered the correlation between smoking and lung cancer, some cigarette man-
ufacturers responded that lung cancer might cause the desire to smoke or
there might be a third cause of both smoking and lung cancer that explains
the correlation. Possibly, it was suggested, smoking relieves discomfort due
to early lung cancer or due to a third factor that itself causes lung cancer. To
test these hypotheses, scientists manipulated the amount of smoking by lab
animals. When all other factors were held as constant as possible, but smok-
ing was increased, lung cancer increased; and when smoking went down,
lung cancer went down. These results would not have occurred if some
third factor had caused both smoking and lung cancer but remained stable
as smoking was manipulated. The findings would also have been different
if incipient lung cancer caused smoking, but had remained constant as sci-
entists manipulated smoking levels. Such experiments can, thus, help us
rule out at least some of the options 1–4.
Direct manipulation like this is not always possible or ethically permis-
sible. The data would probably be more reliable if the test subjects were
human beings rather than lab animals, but that is not an ethical option.
Perhaps more complicated statistical methods could produce more reliable
results, but they often require large amounts and special kinds of data.
Such data is, unfortunately, often unavailable.
In each of the following examples a strong correlation, either negative or posi-
tive, holds between two sets of phenomena, A and B. Try to decide whether A
is the cause of B, B is the cause of A, both are caused by some third factor, C, or
the correlation is simply accidental. Explain your choice.
1. For a particular United States president, there is a negative correlation
between the number of hairs on his head (A) and the population of
China (B).
Exercise VI
(continued)
254
CHAPTER 9 ■ Causal Reason ing
1. After it became beyond doubt that smoking is dangerous to people’s health,
a new debate arose concerning the possible health hazards of secondhand
smoke on nonsmokers. Collect statements pro and con on this issue and
evaluate the strength of the inductive arguments on each side.
2. The high positive correlation between CO2 concentrations in the atmosphere
and the Earth’s mean surface temperatures is often cited as evidence that
increases in atmospheric CO2 cause global warming. This argument is
illustrated by the famous “hockey stick” diagram in Al Gore’s An Incon-
venient Truth. Is this argument persuasive? How could skeptics about global
warming respond?
3. In Twilight of the Idols, Nietzsche claims that the following examples
illustrate “the error of mistaking cause for consequence.” Do you agree?
Why or why not?
Everyone knows the book of the celebrated Cornaro in which he recommends
his meager diet as a recipe for a long and happy life—a virtuous one, too. . . . I
do not doubt that hardly any book (the Bible rightly excepted) has done so much
harm, has shortened so many lives, as this curiosity, which was so well meant.
Discussion Questions
2. My son’s height (A) increases along with the height of the tree outside my
front door (B).
3. It has been claimed that there is a strong positive correlation between
those students who take sex education courses (A) and those who contract
venereal disease (B).
4. At one time there was a strong negative correlation between the number
of mules in a state (A) and the salaries paid to professors at the state
university (B). In other words, the more mules, the lower professional
salaries.3
5. There is a high positive correlation between the number of fire engines in
a particular borough in New York City (A) and the number of fires that
occur there (B).4
6. “Washington (UPI)—Rural Americans with locked doors, watchdogs or
guns may face as much risk of burglary as neighbors who leave doors
unlocked, a federally financed study says. The study, financed in part by a
three-year $170,000 grant from the Law Enforcement Assistance Adminis-
tration, was based on a survey of nearly 900 families in rural Ohio. Sixty
percent of the rural residents surveyed regularly locked doors [A], but
were burglarized more often than residents who left doors unlocked [B].“5
7. The speed of a car (A) is exactly the same as the speed of its shadow (B).
8. The length of a runner’s ring finger minus the length of the runner’s
index finger (A) is correlated with the runner’s speed in the one-hundred-
yard dash. (B)
255
Concomitant Var iat ion
NOTES
1 Mill’s “methods of experimental inquiry” are found in book 3, chap. 8 of his A System of Logic
(London: John W. Parker, 1843). Mill’s method of difference, method of agreement, and joint
method parallel our SCT, NCT, and Joint Test, respectively. Our simplification of Mill’s methods
derives from Brian Skyrms, Choice and Chance, 3rd ed. (Belmont, CA: Wadsworth, 1986), chap. 4.
2 These excerpts are drawn from David W. Fraser and Joseph E. McDade, “Legionellosis,” Scientific
American, October 1979, 82–99.
3 From Gregory A. Kimble, How to Use (and Misuse) Statistics (Englewood Cliffs, NJ: Prentice-
Hall, 1978), 182.
4 From Kimble, How to Use (and Misuse) Statistics, 182.
5 “Locked Doors No Bar to Crime, Study Says,” Santa Barbara [California] Newspress, Wednesday,
February 16, 1977. This title suggests that locking your doors will not increase safety. Is that a
reasonable lesson to draw from this study?
6 Friedrich Nietzsche, Twilight of the Idols and The Anti-Christ, trans. R. J. Hollingdale (1889;
Harmondsworth: Penguin, 1968), 47–48.
The reason: mistaking the consequence for the cause. The worthy Italian saw in
his diet the cause of his long life; while the prerequisite of long life, an extraordi-
narily slow metabolism, a small consumption, was the cause of his meager diet.
He was not free to eat much or little as he chose, his frugality was not an act of
“free will“: he became ill when he ate more. But if one is not a bony fellow of this
sort one does not merely do well, one positively needs to eat properly. A scholar
of our day, with his rapid consumption of nervous energy, would kill himself
with Cornaro’s regimen. . . .
Long life, a plentiful posterity is not the reward of virtue, virtue itself is rather
just that slowing down of the metabolism which also has, among other things, a
long life, a plentiful posterity, in short Cornarism, as its outcome.—The Church
and morality say: “A race, a people perishes through vice and luxury.” My re-
stored reason says: when a people is perishing, degenerating physiologically,
vice and luxury (that is to say the necessity for stronger and stronger and more
and more frequent stimulants, such as every exhausted nature is acquainted
with) follow therefrom. A young man grows prematurely pale and faded. His
friends say: this and that illness is to blame. I say: that he became ill, that he failed
to resist the illness, was already the consequence of an impoverished life, an
hereditary exhaustion. The newspaper reader says: this party will ruin itself if
it makes errors like this. My higher politics says: a party which makes errors
like this is already finished—it is no longer secure in its instincts. Every error,
of whatever kind, is a consequence of degeneration of instinct, degeneration of
will: one has thereby virtually defined the bad. Everything good is instinct—and
consequently easy, necessary, free. Effort is an objection, the god is typically
distinguished from the hero (in my language: light feet are the first attribute of
divinity).6
This page intentionally left blank
Inference to the Best
Explanation and from Analogy
Once we know the cause of a phenomenon, we can cite this cause in a premise of an
argument whose purpose is to explain the phenomenon (as we saw in Chapter 1).
Explanation and causation are also related in a different way, for explanations can be
used to pick out the cause from among various conditions correlated with the phe-
nomenon (a problem faced at the end of Chapter 9). The general strategy is then to
cite the explanatory value of a causal hypothesis as evidence for that hypothesis. This
form of argument, which is described as an inference to the best explanation, is
the first topic in this chapter. It requires us to determine which explanation is best, so
we will investigate common standards for assessing explanations, including falsifia-
bility, conservativeness, modesty, simplicity, power, and depth. After explaining
these standards, this chapter will turn to a related form of argument called an
argument from analogy, in which the fact that two things have certain features
in common is taken as evidence that they have further features in common. The
chapter ends by suggesting that many, or maybe even all, arguments from analogy
are ultimately based on implicit inferences to the best explanation.
INFERENCES TO THE BEST EXPLANATION
One of the most common forms of inductive argument is inference to the best
explanation.1 The general idea behind such inferences is that a hypothesis
gains inductive support if, when added to our stock of previously accepted
beliefs, it enables us to explain something that we observe or believe, and no
competing explanation works nearly as well.
To see how inferences to the best explanation work, suppose you return to
your home and discover that the lock on your front door is broken and some
valuables are missing. In all likelihood, you will immediately conclude that
you have been burglarized. Of course, other things could have produced the
mess. Perhaps the police mistakenly busted into your house looking for drugs
and took your valuables as evidence. Perhaps your friends are playing a
strange joke on you. Perhaps a meteorite struck the door and then vaporized
10
257
258
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
your valuables. In fact, all of these things could have happened (even the last),
and further investigation could show that one of them did. Why, then, do we
so quickly accept the burglary hypothesis without even considering these
competing possibilities? The reason is that the hypothesis that your home was
robbed is not highly improbable; and this hypothesis, together with other
things we believe, provides the best—the strongest and the most natural—
explanation of the phenomenon. The possibility that a meteorite struck your
door is so wildly remote that it is not worth taking seriously. The possibility
that your house was raided by mistake or that your friends are playing a
strange practical joke on you is not wildly remote, but neither fits the overall
facts very well. If it was a police raid, then you would expect to find a police
officer there or at least a note. If it is a joke, then it is hard to see the point of it.
By contrast, burglaries are not very unusual, and that hypothesis fits the facts
extremely well. Logically, the situation looks like this:
(1) OBSERVATION: Your lock is broken, and your valuables are missing.
(2) EXPLANATION: The hypothesis that your house has been burglarized,
combined with previously accepted facts and principles, provides a
suitably strong explanation of observation 1.
(3) COMPARISON: No other hypothesis provides an explanation nearly
as good as that in 2.
(4) CONCLUSION: Your house was burglarized.
The explanatory power of the conclusion gives us reason to believe it be-
cause doing so increases our ability to understand our observations and to
make reliable predictions. Explanation is important because it makes sense
out of things—makes them more intelligible—and we want to understand
the world around us. Prediction is important because it tests our theories
with new data and sometimes allows us to anticipate or even control future
events. Inference to the best explanation enables us to achieve such goals.
Here it might help to compare inferences to the best explanation with
other forms of argument. Prior to any belief about burglars, you were al-
ready justified in believing that your lock was broken and your valuables
were missing. You could see that much. What you could not see was why
your lock was broken. That question is what the explanation answers. Ex-
planations help us understand why things happen, when we are already jus-
tified in believing those things did happen. (Recall Chapter 1.)
Explanations often take the form of arguments. In our example, we could
argue:
(1) Your house was burglarized.
(2) When houses are burglarized, valuables are missing.
(3) Your valuables are missing.
This explanatory argument starts with the hypothesis that was the conclusion
of the inference to the best explanation, and it ends with the observation that
was the first premise in that inference to the best explanation. The difference
259
Inferences to the Best Explanat ion
is that this new argument explains why its conclusion is true—why the valu-
ables are missing—whereas the inference to the best explanation justified be-
lief in its conclusion that your house was burglarized.
More generally, in an explanatory use of argument, we try to make sense
of something by deriving it (sometimes deductively) from premises that are
themselves well established. With an inference to the best explanation, we
reason in the opposite direction: Instead of deriving an observation from its
explanation, we derive the explanation from the observation. That a hypoth-
esis provides the best explanation of something whose truth is already
known provides evidence for the truth of that hypothesis.
Once we grasp the notion of an inference to the best explanation, we can
see this pattern of reasoning everywhere. If you see your friend kick the wall,
you infer that he must be angry, because there is no other explanation of why
he would kick the wall. Then if he turns away when you say, “Hello,” you
might think that he is angry at you, if you cannot imagine any other reason
why he would not respond. Similarly, when your car goes dead right after a
checkup (as at the start of Chapter 9), you may conclude that it is out of fuel, if
that is the best explanation of why your car stopped. Psychologists infer that
people care what others think about them, even when they deny it, because
that explains why people behave differently in front of others than when they
are alone. Linguists argue that the original Indo-European language arose mil-
lennia ago in an area that was not next to the sea but did have lakes and rivers,
because that is the best explanation of why Indo-European languages have no
common word for seas but do share a common root “nav-” that connotes
boats or ships. Astronomers believe that our Universe began with a Big Bang,
because that hypothesis best explains the background microwave radiation
and spreading of galaxies. All of these arguments and many more are basi-
cally inferences to the best explanation.
Solutions to murder mysteries almost always have the form of an infer-
ence to the best explanation. The facts of the case are laid out and then the
clever detective argues that, given these facts, only one person could possi-
bly have committed the crime. In the story “Silver Blaze,” Sherlock Holmes
concludes that the trainer must have been the dastardly fellow who stole
Silver Blaze, the horse favored to win the Wessex Cup, which was to be run
the following day. Holmes’s reasoning, as usual, was very complex, but the
key part of his argument was that the dog kept in the stable did not bark
loudly when someone came and took away the horse.
I had grasped the significance of the silence of the dog, for one true inference
invariably suggests others. [I knew that] a dog was kept in the stables, and yet,
though someone had been in and fetched out a horse, he had not barked enough
to arouse the two lads in the loft. Obviously the midnight visitor was someone
whom the dog knew well.2
Together with other facts, this was enough to identify the trainer, Straker, as
the person who stole Silver Blaze. In this case, it is the fact that something
didn’t occur that provides the basis for an inference to the best explanation.
260
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
Of course, Holmes’s inference is not absolutely airtight. It is possible that
Straker is innocent and Martians with hypnotic powers over dogs commit-
ted the crime. But that only goes to show that this inference is neither valid
nor deductive in our sense. It does not show anything wrong with Holmes’s
inference. Since his inference is inductive, it is enough for it to be strong.
Inferences to the best explanation are also defeasible. No matter how
strong such an inference might be, it can always be overturned by future ex-
perience. Holmes might later find traces of a sedative in the dog’s blood or
someone else might confess or provide Straker with an alibi. Alternatively,
Holmes (or you) might think up some better explanation. Still, unless and un-
til such new evidence or hypothesis comes along, we have adequate reason
to believe that Straker stole the horse, because that hypothesis provides the
best available explanation of the information that we have now. The fact that
future evidence or hypotheses always might defeat inferences to the best ex-
planation does not show that such inferences are all bad. If it did show this,
then science and everyday life would be in trouble, because so much of sci-
ence and our commonsense view of the world depends on inferences to the
best explanation.
To assess such inferences, we still need some standards for determining
which explanation is the best. There is, unfortunately, no simple rule for
deciding this, but we can list some factors that go into the evaluation of
an explanation.3
First, the hypothesis should really explain the observations. A good expla-
nation makes sense out of that which it is intended to explain. In our origi-
nal example, the broken lock can be explained by a burglary but not by the
hypothesis that a friend came to see you (unless you have strange friends).
Moreover, the hypothesis needs to explain all of the relevant observations.
The hypothesis of a mistaken police raid might explain the broken lock but
not the missing valuables or the lack of any note or police officers when
you return home.
The explanation should also be deep. An explanation is not deep but shal-
low when the explanation itself needs to be explained. It does not help to ex-
plain something that is obscure by citing something just as obscure. Why
did the police raid your house? Because they suspected you. That explana-
tion is shallow if it immediately leads to another question: Why did they
suspect you? Because they had the wrong address. If they did not have the
wrong address, then we would wonder why they suspected you. Without
an explanation of their suspicions, the police raid hypothesis could not ade-
quately explain even the broken lock.
Third, the explanation should be powerful. It is a mark of excellence in
an explanation that the same kind of explanation can be used successfully
over a wide range of cases. Many broken locks can be explained by burglar-
ies. Explanatory range is especially important in science. One of the main
reasons why Einstein’s theory of relativity replaced Newtonian physics is
that Einstein could explain a wider range of phenomena, including very
small particles at very high speeds.
261
Inferences to the Best Explanat ion
Explanations go too far, however, when they could explain any possible
event. Consider the hypothesis that each particle of matter has its own indi-
vidual spirit that makes it do exactly what it does. This hypothesis might
seem to explain some phenomena that even Einstein’s theory cannot ex-
plain. But the spirit hypothesis really explains nothing, because it does not
explain why any particle behaves one way as opposed to another. Either be-
havior is compatible with the hypothesis, so neither is explained. To suc-
ceed, therefore, explanations need to be incompatible with some possible
outcome. In short, they need to be falsifiable. (See Chapter 16 on self-sealers.)
Moreover, explanations should be modest in the sense that they should not
claim too much—indeed, any more than is needed to explain the observa-
tions. When you find your lock broken and valuables gone, you should not
jump to the conclusion that there is a conspiracy against you or that gangs
have taken over your neighborhood. Without further information, there is
no need to specify that there was more than one burglar in order to explain
what you see. There is also no need to hypothesize that there was only one
burglar. For this reason, the most modest explanation would not specify any
number of burglars, so no inference to the best explanation could justify any
claim about the number of burglars, at least until more evidence comes
along.
Modesty is related to simplicity. One kind of simplicity is captured by the
celebrated principle known as Occam’s razor, which tells us not to multiply
entities beyond necessity. Physicists, for example, should not postulate new
kinds of subatomic particles or forces unless there is no other way to explain
their experimental results. Similar standards apply in everyday life. We
should not believe in ghosts unless they really are necessary to explain the
noises in our attic or some other phenomenon. Simplicity is not always a
matter of new kinds of entities. In comparison with earlier views, the theory
that gases are composed of particles too small to see was simpler insofar as
the particle theory allowed gas laws to be explained by the standard physi-
cal principles governing the motions of larger particles without having to
add any new laws. Simplicity is a mark of excellence in an explanation
partly because simple explanations are easier to understand and apply, but
considerations of plausibility and aesthetics are also at work in judgments
of which explanation is simplest.
The tests of modesty and simplicity might seem to be in tension with the
test of power. This tension can be resolved only by finding the right balance.
The best explanation will not claim any more than is necessary (so it will be
modest), but it will claim enough to cover a wide range of phenomena (so it
will be powerful). This is tricky, but the best explanations succeed in recon-
ciling and incorporating these conflicting virtues as much as possible.
Finally, an explanation should be conservative. Explanations are better when
they force us to give up fewer well-established beliefs. We have strong reasons
to believe that cats cannot break metal locks. This rules out the hypothesis that
your neighbor’s cat broke your front-door lock. Explanations should also not
contain claims that are themselves too unlikely to be true. A meteorite would
262
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
be strong enough to break your lock, but it is very unlikely that a meteorite
struck your lock. That makes the burglary hypothesis better, at least until we
find other evidence (such as meteorite fragments) that cannot be explained
except by a meteorite.
In sum, a hypothesis provides the best explanation when it is more ex-
planatory, powerful, falsifiable, modest, simple, and conservative than any
competing hypothesis. Each of these standards can be met to varying de-
grees, and they can conflict. As we saw, the desire for simplicity might have
to be sacrificed to gain a more powerful explanation. Conservatism also
might have to give way to explain some unexpected observations, and so
on. These standards are not always easy to apply, but they can often be used
to determine whether a particular explanation is better than its competitors.
Once we determine that one explanation is the best, we still cannot yet in-
fer that it is true. It might turn out that the best explanation out of a group of
weak explanations isn’t good enough. For centuries people were baffled by
the floods that occurred in the Nile River each spring. The Nile, as far as any-
one knew, flowed from an endless desert. Where, then, did the flood waters
come from? Various wild explanations were suggested—mostly about deities
of one kind or another—but none was any good. Looking for the best expla-
nation among these weak explanations would be a waste of time. It was only
after it was discovered that central Africa contains a high mountain range cov-
ered with snow in the winter that a reasonable explanation became possible.
That, in fact, settled the matter. So it must be understood that the best expla-
nation must also be a good enough explanation.
Even when an explanation is both good and best, what it explains might
be illusory. Many people believe that shark cartilage prevents cancer, be-
cause the best explanation of why sharks do not get cancer lies in their carti-
lage. One serious problem for this inference is that sharks do get cancer. They
even get cancer in their cartilage. So this inference to the best explanation
fails.
When a particular explanation is both good and much better than any
competitor, and when the explained observation is accurate, then an infer-
ence to the best explanation will provide strong inductive support. At other
times, no clear winner or even reasonable contender emerges. In such cases,
an inference to the best explanation will be correspondingly weak.
Whether an inference to the best explanation is strong enough depends on
the context. As contexts shift, standards of rigor can change. Evidence that is
strong enough to justify my belief that my spouse took our car might not be
strong enough to convict our neighbor of stealing our car. Good judgment is
often required to determine whether a certain degree of strength is adequate
for the purposes at hand.
Context can also affect the rankings of various factors. Many explanations,
for example, depend on universal premises. In such cases, compatibility with
observation is usually the primary test. The universal principle should not be
refuted by counterexamples (see Chapter 17). But sometimes explanatory
263
Inferences to the Best Explanat ion
power will take precedence: If a principle has strong explanatory power, we
may accept it even in the face of clear disconfirming evidence. We do not give
up good explanations lightly—nor should we. To understand why, recall
(from Chapter 9) that we do not test single propositions in isolation from
other propositions in our system of beliefs. When faced with counterevidence
to our beliefs, we often have a choice between what to give up and what to
continue to hold on to. A simple example will illustrate this. Suppose that we
believe the following things:
(1) Either John or Joan committed the crime.
(2) Whoever committed the crime must have had a motive for doing so.
(3) Joan had no motive to commit the crime.
From these three premises we can validly infer that John committed the
crime. Suppose, however, that we discover that John could not have com-
mitted the crime. (Three bishops and two judges swear that John was some-
where else at the time.) Now, from the fact that John did not commit the
crime, we could not immediately conclude that Joan committed it, for that
would lead to an inconsistency. If she committed the crime, then, according
to premise 3, she would have committed a motiveless crime, but that con-
flicts with premise 2, which says that motiveless crimes do not occur. So the
discovery that John did not commit the crime entails that at least one of the
premises in the argument must be abandoned, but it does not tell us which
one or which ones.
This same phenomenon occurs when we are dealing with counterevidence
to a complex system of beliefs. Counterevidence shows that there must be
something wrong somewhere in the system, but it does not show exactly
where the problem lies. One possibility is that the supposed counterevidence
is itself in error. Imagine that a student carries out an experiment and gets the
result that one of the fundamental laws of physics is false. This will not shake
the scientific community even a little, for the best explanation of the student’s
result is that she messed things up. Given well-established principles, she
could not have gotten the result she did if she had run the experiment cor-
rectly. Of course, if a great many reputable scientists find difficulties with a
supposed law, then the situation is different. The hypothesis that all of these
scientists, like the student, simply messed up is itself highly unlikely. But it is
surprising how much contrary evidence will be tolerated when dealing with
a strong explanatory theory. Scientists often continue to employ a theory in
the face of counterevidence. Sometimes this perpetuates errors. For years, in-
struments reported that the levels of ozone above Antarctica were lower than
before, but scientists attributed these measurements to bad equipment, until
finally they announced an ozone hole there. Still, there is often good reason
to hold on to a useful theory despite counterevidence, as long as its defects
do not make serious trouble—that is, give bad results in areas that count.
Good judgment is required to determine when it is finally time to shift to a
different explanation.
264
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
Imagine that you offer an explanation, and a critic responds in the following
way. Which virtue (explanatoriness, depth, power, falsifiability, modesty, sim-
plicity, or conservativeness) is your critic claiming that your explanation lacks?
1. But that won’t explain anything other than this particular case.
2. But that conflicts with everything we know about biology.
3. But you don’t have to claim all of that in order to explain what we see.
4. But that just raises new questions that you need to answer.
5. But that explains only a small part of the story.
6. But that would apply whatever happened.
Exercise I
For each of the following explanations, specify which standard of a good ex-
planation, if any, it violates. The standards require that a good explanation be
explanatory, deep, powerful, falsifiable, modest, simple, and conservative. A
single explanation might violate more than one standard.
1. Although we usually have class at this time in this room, I don’t see any-
body in the classroom, because a wicked witch made them all invisible.
2. Although we usually have class at this time in this room, I don’t see any-
body in the classroom, because they all decided to skip class today.
3. Although we usually have class at this time in this room, I don’t see any-
body in the classroom, because it’s Columbus Day.
4. My house fell down, because it was painted red.
5. My house fell down, because of a powerful earthquake centered on my
property that did not affect anything or anybody else.
6. My house fell down, because its boards were struck by a new kind of sub-
atomic particle.
7. Although I fished here all day, I didn’t catch any fish, because there are
no fish in this whole river.
8. Although I fished here all day, I didn’t catch any fish, because the river
gods don’t like me.
9. Although I fished here all day, I didn’t catch any fish, because I was
unlucky today.
10. That light far up in the night sky is moving quickly, because it is the daily
United Airlines flight from Boston to Los Angeles.
11. That light far up in the night sky is moving quickly, because it is an alien
space ship.
12. That light far up in the night sky looks like it is moving quickly, because
there’s something wrong with my eyes right now.
Exercise II
265
Inferences to the Best Explanat ion
Give two competing hypotheses that might be offered to explain each of the
following phenomena. Which of these hypotheses is better? Why?
1. You follow a recipe carefully, but the bread never rises.
2. Your house begins to shake so violently that pictures fall off your walls.
3. Your key will not open the door of your house.
4. People start putting television cameras on your lawn, and a man with a
big smile comes walking up your driveway.
5. Virtually all of the food in markets has suddenly sold out.
6. You put on a shirt and notice that there is no pocket on the front like there
used to be.
7. A cave is found containing the bones of both prehistoric humans and
now-extinct predators.
8. A cave is found containing the bones of both prehistoric humans and
now-extinct herbivores.
9. After being visited by lobbyists for cigarette producers, your senator votes
in favor of tobacco price supports, although he opposed them before.
10. Large, mysterious patterns of flattened wheat appear in the fields of Britain.
(Some people attribute these patterns to visitors from another planet.)
11. A palm reader foretells that something wonderful will happen to you
soon, and it does.
12. A neighbor sprinkles purple powder on his lawn to keep away tigers,
and, sure enough, no tigers show up on his lawn.
Exercise III
1. Put the following inference to the best explanation in standard form, and
then evaluate it as carefully as you can, using the tests discussed above.
[During the Archean Era, which extended from about 3.8 to 2.5 million years before
the present,] the sun’s luminosity was perhaps 25% less than that of today. . . . This
faint young sun has led to a paradox. There is no evidence from the scant rock record
of the Archean that the planetary surface was frozen. However, if Earth had no
atmosphere or an atmosphere of composition like that of today, the amount of
radiant energy received by Earth from the sun would not be enough to keep it from
freezing. The way out of this dilemma is to have an atmosphere present during the
early Archean that was different in composition that that of today. . . . For a variety
of reasons, it has been concluded, although still debated, that the most likely gases
present in greater abundance in the Archean atmosphere were carbon dioxide, water
vapor (the most important greenhouse gas) and perhaps methane. The presence of
these greenhouse gases warmed the atmosphere and planetary surface and pre-
vented the early Archean Earth from being frozen.4
Discussion Questions
(continued)
266
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
2. Reread the passage on what killed the dinosaurs in the Discussion Questions
at the end of Chapter 1. Reconstruct the argument as an inference to the best
explanation. How well does that argument meet the standards for assessing
explanations?
3. Also in the Discussion Questions at the end of Chapter 1, Colin Powell
gives several arguments that in 2003 Saddam Hussein was still trying to
obtain fissile material for a nuclear weapons program. Which of his argu-
ments is an inference to the best explanation? How well do these
arguments meet the standards for this form of argument?
4. Find three inferences to the best explanation in the readings on scientific
reasoning in Chapter 20. This should be easy because scientists use this
form of argument often. Put those inferences in standard form, and then
evaluate them using the tests discussed above.
5. Sherlock Holmes was lauded for his ability to infer a great deal of informa-
tion about strangers from simple observations of their clothing and behav-
ior. He displays this ability in the following exchange with his brother,
Mycroft, in front of his friend Dr. Watson, who is the first-person narrator.
Reconstruct and evaluate the inferences to the best explanation by Holmes
and Mycroft in the closing four paragraphs.
“THE GREEK INTERPRETER”
from The Memoirs of Sherlock Holmes
by Sir Arthur Conan Doyle
. . . The two sat down together in the bow-window of the club. “To anyone who
wishes to study mankind this is the spot,” said Mycroft. “Look at the magnifi-
cent types! Look at these two men who are coming towards us for example.”
“The billiard-marker and the other?”
“Precisely. What do you make of the other?”
The two men had stopped opposite the window. Some chalk marks over
the waistcoat pocket were the only signs of billiards which I could see in one
of them. The other was a very small, dark fellow, with his hat pushed back and
several packages under his arm.
“An old soldier, I perceive,” said Sherlock.
“And very recently discharged,” remarked the brother.
“Served in India, I see.”
“And a non-commissioned officer.”
“Royal Artillery, I fancy” said Sherlock.
“And a widower.”
“But with a child.”
“Children, my dear boy, children.”
“Come,” said I, laughing, “this is a little too much.”
“Surely,” answered Holmes, “it is not hard to say that a man with that
bearing, expression of authority, and sun-baked skin, is a soldier, is more than
a private, and is not long from India.”
267
Arguments from Analogy
“That he has not left the service long is shown by his still wearing his am-
munition boots, as they are called,” observed Mycroft.
“He had not the cavalry stride, yet he wore his hat on one side, as is shown
by the lighter skin on that side of his brow. His weight is against his being a
sapper [a soldier who builds and repairs fortifications]. He is in the artillery.”
“Then, of course, his complete mourning shows that he has lost someone
very dear. The fact that he is doing his own shopping looks as though it were
his wife. He has been buying things for children, you perceive. There is a rat-
tle, which shows that one of them is very young. The wife probably died in
childbed. The fact that he has a picture-book under his arm shows that there is
another child to be thought of.”
ARGUMENTS FROM ANALOGY
Another very common kind of inductive argument moves from a premise
that two things are similar in some respects to a conclusion that they must
also be analogous in a further respect. Such arguments from analogy can be
found in many areas of everyday life. When we buy a new car, how can we
tell whether it is going to be reliable? Consumer Reports might help if it is an
old model; but if it is a brand-new model with no track record, then all we
can go on is its similarities to earlier models. Our reasoning then seems to be
that the new model is like the old model in various ways, and the old model
was reliable, so the new model is probably reliable, too.
The same form of argument is used in science. Here’s an example from
geology:
Meteorites composed predominantly of iron provide evidence that parts of other
bodies in the solar system, presumably similar in origin to Earth, were composed
of metallic iron. The evidence from meteorite compositions and origins lends
support to the conclusion that Earth’s core is metallic iron.5
The argument here is that Earth is analogous to certain meteors in their ori-
gins, and those meteors have a large percentage of iron, so the Earth as a
whole probably contains about the same percentage of iron. Because a
smaller amount of iron is present in the Earth’s crust, the rest must lie in the
Earth’s core.
Similarly, archaeologists might argue that a certain knife was used in rit-
ual sacrifices because it resembles other sacrificial knives in its size, shape,
materials, carvings, and so on. The analogy in this case is between the newly
discovered knife and the other knives. This analogy is supposed to support
a conclusion about the function of the newly discovered knife.
Although such arguments from analogy have diverse contents, they share
a common form that can be represented like this:
(1) Object A has properties P, Q, R, and so on.
(2) Objects B, C, D, and so on also have properties P, Q, R, and so on.
268
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
(3) Objects B, C, D, and so on have property X.
(4) Object A probably also has property X.
In the archaeological example, object A is the newly discovered knife, and
objects B, C, D, and so on are previously discovered knives that are known
to have been used in sacrifices. Properties P, Q, R, and so on are the size,
shape, materials, and carvings that make A analogous to B, C, D, and so on.
X is the property of being used as a sacrificial knife. Premise 3 says that the
previously discovered artifacts have this property. The conclusion, on line 4,
says that the newly discovered artifact probably also has this property.
Since arguments from analogy are inductive, they normally aren’t valid.
It is possible that, even though this knife is analogous to other sacrificial
knives, this knife was used to shave the king or just to cut bread. These ar-
guments are also defeasible. The argument about knives obviously loses all
of its strength if we find “Made in China” printed on the newly discovered
knife. Still, none of this shows that arguments from analogy are no good. De-
spite being invalid and defeasible, some arguments from analogy can still
provide reasons—even strong reasons—for their conclusions.
How can we tell whether an argument from analogy is strong or weak? One
obvious requirement is that the premises must be true. If the previously discov-
ered knives were not really used in sacrifices, or if they do not really have the
same carvings on their handles as the newly discovered knife, then this argu-
ment from analogy does not provide much, if any, support for its conclusion.
In addition, the cited similarities must be relevant. Suppose someone ar-
gues that his old car was red with a black interior and had four doors and a
sunroof, and his new car also has these properties, so his new car is proba-
bly going to be as reliable as his old car. This argument is very weak because
the cited similarities are obviously irrelevant to reliability. Such assessments
of relevance depend on background beliefs, such as that reliability depends
on the drive train and the engine rather than on the color or the sunroof.
The similarities must also be important. Similarities are usually more im-
portant the more specific they are. Lots of cars with four tires and a motor
are reliable, but this is not enough to infer that, because this particular car
also has four tires and a motor, it will be reliable, too. The reason is obvious:
There are also lots of unreliable cars with four tires and a motor. In general,
if many objects have properties P, Q, and R, and many of those lack property
X, then arguments from these analogies will be weak. In contrast, if a
smaller percentage of objects that have properties P, Q, and R lack property
X, then the argument from these analogies will be strong.
If we are not sure which respects are important, we still might have some
idea of which respects might be important. Then we can try to cite objects
that are analogous in as many as possible of those respects. By increasing the
number of potentially relevant respects for which the analogy holds, we can
increase the likelihood that the important respects will be on our list. That
shows why arguments from analogy are usually stronger when they cite
more and closer analogies between the objects.
269
Arguments from Analogy
Another factor that affects the strength of an argument from analogy is
the presence of relevant disanalogies. Because arguments from analogy are de-
feasible, as we saw, a strong argument from analogy can become weak if we
add a premise that states an important disanalogy. Suppose my new car is
like my old cars in many ways, but there is one difference: The new car has
an electric motor, whereas the old cars were powered by gasoline. This one
difference is enough to weaken any argument to the conclusion that the new
car will be reliable. Of course, other disanalogies, such as a different color,
won’t matter to reliability; and it will often require background knowledge
to determine how important a disanalogy is.
We need to be careful here. Some disanalogies that are relevant do not un-
dermine an argument from analogy. If a new engine design was introduced
by top engineers to increase reliability, then this disanalogy might not un-
dermine the argument from analogy. Differences that point to more reliabil-
ity rather than less might even make the argument from analogy stronger.
Other disanalogies can increase the strength of an argument from anal-
ogy in a different way. If the same markings are found on very different
kinds of sacrificial knives, then the presence of those markings on the newly
discovered knife is even stronger evidence that this knife was also used in
sacrifices. Differences among the cases cited only in the premises as analo-
gies (that is, B, C, D, and so on) can strengthen an argument from analogy.
Finally, the strength of an argument from analogy depends on its conclu-
sion. Analogies to other kinds of cars provide stronger evidence for a weak
conclusion (such as that the new model will probably be pretty reliable)
and weaker evidence for a strong conclusion (such as that the new model
will definitely be just as reliable as the old model). As with other forms of
argument, an argument from analogy becomes stronger as its conclusion
becomes weaker and vice versa.
These standards can be summarized by saying that an argument from
analogy is stronger when:
1. It cites more and closer analogies that are more important.
2. There are fewer or less important disanalogies between the object in
the conclusion and the other objects.
3. The objects cited only in the premises are more diverse.
4. The conclusion is weaker.
After learning about arguments from analogy, it is natural to wonder how
they are related to inferences to the best explanation. Although this is some-
times disputed, it seems to us that arguments from analogy are often—if not
always—implicit and incomplete inferences to the best explanation. As we
pointed out, analogies don’t support any conclusion unless they are rele-
vant, and whether they are relevant depends on how they fit into explana-
tions. The color of a car is irrelevant to its reliability, because color plays no
role in explaining its reliability. What explains its reliability is its drive train
design, materials, care in manufacturing, and so on. That is why analogies
270
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
in those respects can support a conclusion about reliability. Similarly, the
markings on an artifact are relevant to whether it is a sacrificial knife if the
best explanation of why it has those markings is that it was used in sacri-
fices. What makes that explanation best is that it also explains similar mark-
ings on other sacrificial knives. Thus, such arguments from analogy can be
seen as involving an inference to the best explanation of why objects B, C, D,
and so on have property X followed by an application of that explanation to
the newly discovered object A.
Sometimes the explanation runs in the other direction. Whereas the con-
clusion about the knife’s use (X) is supposed to explain its shared markings
(P, Q, R), sometimes it is the shared features (P, Q, R) that are supposed to
explain the feature claimed in the conclusion (X). Here is a classic example:
We may observe a very great [similarity] between this earth which we inhabit,
and the other planets, Saturn, Jupiter, Mars, Venus, and Mercury. They all revolve
around the sun, as the earth does, although at different distances and in different
periods. They borrow all their light from the sun, as the earth does. Several of
them are known to revolve around their axis like the earth, and, by that means,
must have a like succession of day and night. Some of them have moons that
serve to give them light in the absence of the sun, as our moon does to us. They
are all, in their motions, subject to the same law of gravitation, as the earth is.
From all this similarity it is not unreasonable to think that those planets may, like
our earth, be the habitation of various orders of living creatures. There is some
probability in this conclusion from analogy.6
The argument here seems to be that some other planet probably supports
life, because Earth does and other planets are similar to Earth in revolving
around the sun and around an axis, getting light from the sun, and so on.
What makes certain analogies relevant is not, of course, that the motion of
Earth is explained by the presence of life here. Rather, certain features of
Earth explain why Earth is habitable. The argument suggests that the best
explanation of why there is life on our planet is that certain conditions make
life possible. That generalization can then be used to support the conclusion
that other planets with the same conditions probably support life as well.
In one way or another, many (or maybe even all) arguments from analogy
can be seen as inferences to the best explanation. But they are usually incom-
plete explanations. The argument for life on other planets did not have to com-
mit itself to any particular theory about the origin of life or about which
conditions are needed to support life. Nor did the car argument specify exactly
what makes cars reliable. Such arguments from analogy merely list a number
of similarities so that the list will be likely to include whatever factors are
needed for life or for reliability. In this way, arguments from analogy can avoid
depending on any complete theory about what is and what is not relevant.
This incompleteness makes arguments from analogy useful in situations
where we do not yet know enough to formulate detailed theories or even to
complete an inference to the best explanation. Yet the incompleteness of ar-
guments from analogy also makes them more vulnerable to refutation, since
the analogies that they list might fail to include a crucial respect. This does
271
Arguments from Analogy
not mean that arguments from analogy are never any good. They can be
strong. However, it does suggest that their strength will increase as they ap-
proach or approximate more complete inferences to the best explanation.
For each of the following arguments, state whether the indicated changes
would make the argument weaker or stronger, and explain why. The strength
of the argument might not be affected at all. If so, say why it is not affected.
1. My friend and I have seen many movies together, and we have always
agreed on whether they are good or bad. My friend liked the movie tril-
ogy The Lord of the Rings. So I probably will like it as well.
Would this argument be weaker or stronger if:
a. The only movies that my friend and I have watched together are come-
dies, and The Lord of the Rings is not a comedy.
b. My friend and I have seen very many, very different movies together.
c. My friend and I always watched movies together on Wednesdays, but
my friend watched The Lord of the Rings on a weekend.
d. The conclusion claims that I definitely will like The Lord of the Rings a lot.
e. The conclusion claims that I probably won’t totally dislike The Lord of
the Rings.
2. All the students from Joe’s high school with high grades and high board
scores did well in college. Joe also had high grades and board scores. So
he will probably do well in college.
Would this argument be weaker or stronger if:
a. The other students worked hard, but Joe’s good grades came easily to
him, so he never learned to work hard.
b. Joe is going to a different college than the students with whom he is
being compared.
c. Joe plans to major in some easy subject, but the other students were
pre-med.
d. Joe recently started taking drugs on a regular basis.
e. Joe needs to work full-time to pay his college expenses, but the others
had their expenses paid by their parents.
3. A new drug cures a serious disease in rats. Rats are similar to humans in
many respects. Therefore, the drug will probably cure the same disease in
humans.
Would this argument be weaker or stronger if:
a. The disease affects the liver, and rat livers are very similar to human livers.
Exercise IV
(continued)
272
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
b. The drug does not cure this disease in cats.
c. The drug has to be injected into the rat’s tail to be effective (that is, it
does not work if it is injected anywhere else in the rat).
d. No drug of this general type has been used on humans before.
e. The effects of the drug are enhanced by eating cooked foods.
Using the criteria mentioned above, evaluate each of the following arguments
as strong or weak. Explain your answers. Be sure to specify the properties on
which the analogy is based, as well as any background beliefs on which your
evaluation depends.
1. This landscape by Cézanne is beautiful. He did another painting of a
similar scene around the same time. So it is probably beautiful, too.
2. My aunt had a Siamese cat that bit me, so this Siamese cat will probably
bite me, too.
3. The students I know who took this course last year got grades of A. I am a
lot like them, since I am also smart and hardworking; and the course this
year covers very similar material. So I will probably get an A.
4. This politician was caught cheating in his marriage, and he will have to
face similarly strong temptations in his public duties, so he will probably
cheat in political life as well.
5. A very high minimum wage led to increased unemployment in one coun-
try. That country’s economy is similar to the economy in a different coun-
try. So a very high minimum wage will probably lead to increased
unemployment in the other country as well.
6. I feel pain when someone hits me hard on the head with a baseball bat.
Your body is a lot like mine. So you would probably feel pain if I hit you
hard on the head with a baseball bat. (This is related to the so-called
“Problem of Other Minds.”)
7. It is immoral for a doctor to lie to a patient about a test result, even if the
doctor thinks that lying is in the patient’s best interest. We know this
because even doctors would agree that it would be morally wrong for a
financial adviser to lie to them about a potential investment, even if the
financial advisor thinks that this lie is in the doctor’s best interests.
8. Chrysler was held legally liable for damages due to defects in the
suspension of its Corvair. The defects in the Pinto gas tank caused injuries
that were just as serious. Thus, Ford should also be held legally liable for
damages due to those defects.
Exercise V
273
Arguments from Analogy
1. As we will see in Chapter 18 of Part V, legal reasoning often uses analogies
to legal precedents. In the case of Plessy v. Ferguson (1896), Plessy cites
similarities to a laundry ordinance in Yick Wo v. Hopkins to argue against
segregation in public transportation. Later, in Grutter and Gratz (excerpted
below), critics argue that affirmative action is unconstitutional because of
analogies to other forms of racial discrimination that were found unconsti-
tutional in precedents. Reconstruct these arguments from analogy and then
evaluate them by applying the criteria discussed above.
2. As we will see in Chapter 19, moral reasoning also often depends on argu-
ments from analogies. One famous example occurs when Judith Jarvis
Thomson defends the morality of abortion by means of an analogy to a kid-
napped violinist (in the excerpt below). Reconstruct Thomson’s argument
from analogy and then evaluate it by applying the criteria discussed above.
3. The following excerpt presents evidence that Neanderthals were cannibals.
Put the central argument from analogy, which is italicized here, into stan-
dard form. Then reconstruct the argument as an inference to the best expla-
nation. Which representation best captures the force of the argument, or
are they equally good?
“A GNAWING QUESTION IS ANSWERED”
from The Toronto Star, October 10, 1999
by Michael Downey
Tim White is worried that he may have helped to pin a bad rap on the Nean-
derthals, the prehistoric Europeans who died out 25,000 years ago. “There is a
danger that everyone will think that all Neanderthals were cannibals and that’s
not necessarily true,” he says. White was part of a French-American team of pa-
leoanthropologists who recently found conclusive evidence that at least some
Neanderthals ate others about 100,000 years ago. But that doesn’t mean they
were cannibalistic by nature, he stresses. Most people don’t realize that cannibal-
ism is widespread throughout nature, says White, a professor at the University
of California at Berkeley and the author of a book on prehistoric cannibalism.
The question of whether the Neanderthals were cannibals had long been a
hotly debated topic among anthropologists. No proof had ever been found. That
debate ended, however, with the recent analysis by the team of stone tools and
bones found in a cave at Moula-Guercy in southern France. The cave is about
the size of a living room, perched about 80 metres above the Rhone River. “This
one site has all of the evidence right together,” says White. “It’s as if somebody
put a yellow tape around the cave for 100,000 years and kept the scene intact.”
The bones of deer and other fauna show the clear markings of the nearby stone
tools, indicating the deer had been expertly butchered; they were skinned, their
body parts cut off and the meat and tendons sliced from the bone. Long bones
were bashed open “to get at the fatty marrow inside,” says White.
Discussion Questions
(continued)
274
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
So what does all this have to do with cannibalism? The bones of the six (so
far) humans in the same locations have precisely the same markings made by the same
tools. That means these fairly modern humans were skinned and eaten in the same
manner as the deer.
And if you are thinking they were eaten after they just happened to die,
they do represent all age groups. Two were children about 6 years old, two
were teenagers and two were adults.
But maybe they were eaten at a time when food was unusually scarce,
right? Not so. There is a large number of animal bones at the same dig, indi-
cating that there were options to eating other Neanderthals.
Human bones with similar cut marks have been found throughout Europe,
from Spain to Croatia, providing tantalizing hints of Neanderthal cannibalism
activity over tens of thousands of years. But finding such clear evidence of the
same preparation techniques being used on deer in the same cave site in France,
will “necessitate reassessment of earlier finds,” always attributed to ritual burial
practices or some other explanation, says White.
It was not clear whether the Neanderthals ate human flesh of their own
tribe or exclusively from an enemy tribe, White stresses. Nor was there any in-
dication the purpose of the cannibalism involved nourishment. Eating human
flesh could have had another purpose altogether, he says. Surprising to some,
cannibalism has been found in 75 mammal species and in 15 primate groups.
White says it has often been practiced for reasons not associated with normal
hunger. White quotes an archeological maxim: “Actions fossilize, intentions
don’t.” In other words, the reason for the cannibalism remains unknown. He
notes that the flesh of other humans has sometimes been eaten to stave off star-
vation, to show contempt for an enemy or as part of a ritual of affection for the
deceased. “Were the victims already dead or killed to be eaten?” he asks,
“Were they enemies of the tribe or family members?” Learning these details is
the “important and extremely difficult part.”
This excavation represents a breakthrough in archeological practice. In a
series of papers, White has long advocated the importance of treating a dig site
like a crime scene, leaving every piece of evidence in place. In many earlier
digs, animal bones were frequently pulled out and thrown away as being irrel-
evant and human bones were often coated with shellac. The human bones were
all tossed into the same bag with no regard to their juxtaposition to each other
or precise location. This is one of the first times that modern forensic techniques
have been utilized in an archeological excavation, White says, and conclusions
drawn have been much more precise than in previous digs that used cruder
methods. The project team, which is headed by Alban Defleur of the Universite
du Mediterrane at Marseilles, has been digging in the cave since 1991.
4. In the following passage, William Paley argues for the existence of God on
the basis of an analogy to a watch. Reconstruct this argument from analogy
and then evaluate it by applying the criteria discussed above. Could
Paley’s argument also be reconstructed as an inference to the best explana-
tion? If so, would that reconstruction better capture the force of the
argument?
275
Arguments from Analogy
“THE WATCH AND THE WATCHMAKER”
from Natural Theology (New York: Hopkins, 1836)
by William Paley
. . . In crossing a heath, suppose I pitched my foot against a stone and were
asked how the stone came to be there, I might possibly answer that for any-
thing I knew to the contrary it had lain there forever; nor would it, perhaps, be
very easy to show the absurdity of this answer. But suppose I had found a
watch upon the ground, and it should be inquired how the watch happened to
be in that place, I should hardly think of the answer which I had before given,
that for anything I knew the watch might have always been there. Yet why
should not this answer serve for the watch as well as for the stone? Why is it
not as admissible in the second case as in the first? For this reason, and for no
other, namely, that when we come to inspect the watch, we perceive—what we
could not discover in the stone—that its several parts are framed and put to-
gether for a purpose, e.g., that they are so formed and adjusted as to produce
motion, and that motion so regulated as to point out the hour of the day; that
if the different parts had been differently shaped from what they are, of a dif-
ferent size from what they are, or placed after any other manner or in any
other order than that in which they are placed, either no motion at all would
have been carried on in the machine, or none which would have answered the
use that is now served by it. . . . This mechanism being observed—it requires
indeed an examination of the instrument, and perhaps some previous knowl-
edge of the subject, to perceive and understand it; but being once, as we have
said, observed and understood—the inference we think is inevitable, that the
watch must have had a maker—that there must have existed, at some time and
at some place or other, an artificer or artificers who formed it for the purpose
which we find it actually to answer, who comprehended its construction and
designed its use. . . .
[E]very indication of contrivance, every manifestation of design, which
existed in the watch, exists in the works of nature; with the difference, on the
side of nature, of being greater and more, and that in a degree which exceeds
all computation. I mean that the contrivances of nature surpass the con-
trivances of art, in the complexity, subtlety, and curiosity of the mechanism;
and still more, if possible, do they go beyond them in number and variety; yet
in a multitude of cases, are not less evidently mechanical, not less evidently
contrivances, not less evidently accommodated to their end, or suited to their
office, than are the most perfect productions of human ingenuity.
I know no better method of introducing so large a subject, than that of
comparing a single thing with a single thing: an eye, for example, with a tele-
scope. As far as the examination of the instrument goes, there is precisely the
same proof that the eye was made for vision, as there is that the telescope was
made for assisting it. They are made upon the same principles; both being ad-
justed to the laws by which the transmission and refraction of rays of light are
regulated. I speak not of the origin of the laws themselves; but such laws be-
ing fixed, the construction in both cases is adapted to them. . . .
(continued)
NOTES
1 Gilbert Harman deserves much credit for calling attention to the importance of inferences to
the best explanation; see, for example, his Thought (Princeton, NJ: Princeton University Press,
1973). A similar form of argument called abduction was analyzed long ago by Charles Sanders
Peirce; see, for example, his Collected Papers of Charles Sanders Peirce (Cambridge, MA: Harvard
University Press, 1931), 5: 189. A wonderful recent discussion is Peter Lipton, Inference to the Best
Explanation (London: Routledge, 1991).
2 Sir Arthur Conan Doyle, “Silver Blaze,” The Complete Sherlock Holmes (Garden City, NY: Dou-
bleday, 1930), 1: 349. The stories describe Holmes as a master of deduction, but his arguments
are inductive as we define the terms.
3 This discussion in many ways parallels and is indebted to the fifth chapter of W. V. Quine and
J. S. Ullian, The Web of Belief, 2nd ed. (New York: Random House, 1978).
4 From Fred T. Mackenzie, Our Changing Planet (Upper Saddle River, NJ: Prentice-Hall, 1998), 192.
5 Mackenzie, Our Changing Planet, 42.
6 Thomas Reid, Essays on the Intellectual Powers of Man (Cambridge, MA: MIT Press, 1969), essay
I, section 4, 48.
276
CHAPTER 10 ■ Inference to the Best Explanat ion and from Analogy
To some it may appear a difference sufficient to destroy all similitude be-
tween the eye and the telescope, that the one is a perceiving organ, the other
an unperceiving instrument. The fact is that they are both instruments. And as
to the mechanism, at least as to mechanism being employed, and even as to
the kind of it, this circumstance varies not the analogy at all. . . . The end is the
same; the means are the same. The purpose in both is alike; the contrivance for
accomplishing that purpose is in both alike. The lenses of the telescopes, and
the humors of the eye, bear a complete resemblance to one another, in their fig-
ure, their position, and in their power over the rays of light, viz. in bringing
each pencil to a point at the right distance from the lens; namely, in the eye, at
the exact place where the membrane is spread to receive it. How is it possible,
under circumstances of such close affinity, and under the operation of equal
evidence, to exclude contrivance from the one; yet to acknowledge the proof
of contrivance having been employed, as the plainest and clearest of all propo-
sitions, in the other? . . .
Were there no example in the world of contrivance except that of the eye,
it would be alone sufficient to support the conclusion, which we draw from it,
as to the necessity of an intelligent Creator. . . . The proof is not a conclusion
that lies at the end of a chain of reasoning, of which chain each instance of con-
trivance is only a link, and of which, if one link fail, the whole fails; but it is an
argument separately supplied by every separate example. An error in stating
an example affects only that example. The argument is cumulative in the
fullest sense of that term. The eye proves it without the ear; the ear without the
eye. The proof in each example is complete; for when the design of the part
and the conduciveness of its structure to that design is shown, the mind may
set itself at rest; no further consideration can detract anything from the force
of the example.
Chances
The kinds of arguments discussed in the preceding three chapters are all inductive, so
they need not meet the deductive standard of validity. They are, instead, intended to
meet the inductive standard of strength. Whereas deductive validity hinges on what is
possible, inductive strength hinges on what is probable. Roughly, an argument is in-
ductively strong to the extent that its premises make its conclusion more likely or prob-
able. Hence, just as we can get a better theoretical understanding of deductive validity
by studying formal logic, as we did in Chapters 6–7, so we can get a better theoretical
understanding of inductive strength by studying probability, as we will do in this chap-
ter. To complete our survey of inductive arguments, this chapter offers an elementary
discussion of probability. It begins by illustrating several common mistakes about prob-
ability. To help avoid these fallacies, we need to approach probability more carefully, so
formal laws of probability are presented along with Bayes’s theorem.
SOME FALLACIES OF PROBABILITY
Probability is pervasive. We all assume or make probability judgments
throughout our lives. We do so whenever we form a belief about which we are
not certain, as in all of the kinds of inductive arguments studied in Chapters
8–10. Such arguments do not pretend to reach their conclusions with certainty,
even if their premises are true. They merely try to show that a conclusion is
likely or probable. Judgments about probability are, thus, assumed in assess-
ing such arguments and beliefs. Probability also plays a crucial role in our
most important decisions. Mistakes about probability can then lead to disas-
ters. Doctors lose patients’ lives, stockbrokers lose clients’ money, and coaches
lose games because they overestimate or underestimate probabilities. Such
mistakes are common and fall into several regular patterns. It is useful to un-
derstand these fallacies, so that we can learn to avoid them.
THE GAMBLER’S FALLACY
Casinos thrive partly because so many gamblers misunderstand probability.
One mistake is so common that it has been dubbed the gambler’s fallacy.
When people have a run of bad luck, they often increase their bets because
11
277
278
CHAPTER 11 ■ Chances
they assume that they are due for a run of good luck to even things out.
Gambling systems are sometimes based on this fallacious idea. People keep
track of the numbers that come up on a roulette wheel, trying to discover a
number that has not come up for a long time. They then pile their money on
that number on the assumption that it is due. They usually end up losing a
bundle.
These gamblers seem to assume, “In the long run, things will even out
(or average out).” Interpreted one way, this amounts to what mathemati-
cians call the law of large numbers, and it is perfectly correct. When flipping
a coin, we expect it to come up heads half the time, so it should come up
heads five times in ten flips. If we actually check this out, however, we
discover that the number of times it comes up heads in ten flips varies
significantly from this predicted value—sometimes coming up heads
more than five times, sometimes coming up fewer. What the law of large
numbers tells us is that the actual percentage of heads will tend to come
closer to the theoretically predicted percentage of heads the more trials
we make. If you flipped a coin a million times, it would be very surpris-
ing if the percentage of heads were more than 1 percent away from the
predicted 50 percent.
This law of large numbers is often misunderstood in a way that leads to
the gambler’s fallacy. Some people assume that each possible outcome will
occur the average number of times in each series of trials. To see that this is
a fallacy, we can go back to flipping coins. Toss a coin until it comes up
heads three times in a row. (This will take less time than you might imagine.)
What is the probability that it will come up heads a fourth time? Put crudely,
some people think that the probability of it coming up heads again must be
very small, because it is unlikely that a fair coin will come up heads four
times in a row, so a tails is needed to even things out. That is wrong. The
chances of getting heads on any given toss is the same, regardless of what
happened on previous tosses. Previous results cannot affect the probabilities
on this new toss.
STRANGE THINGS HAPPEN
Another common mistake is to ignore improbable events. It is very unlikely
that a fair coin will come up heads nineteen times in a row, so you might
think it could never happen. You would be wrong. Of course, if you sat flip-
ping a single coin, you might spend a very long time before you hit a se-
quence of nineteen consecutive heads, but there is a way of getting this
result (with a little help from your friends) in a single afternoon. You start
out with $6,000 worth of pennies and put them in a large truck. (Actually,
the truck need not be very large.) Dump the coins out and then pick up all
the coins that come up heads. Put them back in the truck and repeat the pro-
cedure nineteen times, always returning only those coins that come up
heads to the truck. With tolerably good luck, on the nineteenth dump of the
279
Some Fallac ies of Probab il i ty
coins, you will get at least one coin that comes up heads again. Any such
coin will have come up heads nineteen times in a row.
What is the point of this silly exercise? It is intended to show that we often
attribute abilities or the lack of abilities to people when, in fact, their perfor-
mances may be statistically insignificant. When people invest with stockbro-
kers, they tend to shift to a new broker when they lose money. When they hit
on a broker who earns them money, they stay and praise this broker’s abili-
ties. In fact, some financial advisers seem to be better than others—they have
a long history of sound financial advice—but the financial community is, in
many ways, like the truckload of pennies we have just examined. There are
many brokers giving all sorts of different advice and, by chance alone, some
of them are bound to give good advice. Furthermore, some of them are
bound to have runs of success, just as some of the pennies dumped from the
truck will have long strings of coming up heads. Thus, in some cases, what
appears to be brilliance in predicting stock prices may be nothing more than
a run of statistically expected good luck.
The gambling casinos of the world are like the truck full of pennies as
well. With roulette wheels spinning in a great many places over a great deal
of time, startlingly long runs are bound to occur. For example, in 1918, black
came up twenty-six consecutive times on a roulette wheel in Monte Carlo.
The odds against this are staggering. But before we can decide what to make
of this event, we would have to judge it in the context of the vast number of
times that the game of roulette has been played.
Students familiar with computer programming should not find it difficult
to write a program that will simulate a Monte Carlo roulette wheel and keep
track long runs of black and long runs of red. On a Monte Carlo wheel, the
odds coming up black are 18/37. The same odds hold for coming up red. Write
a program; run it for a day; then report the longest runs.
Honors Exercise
HEURISTICS
In daily life, we often have to make decisions quickly without full informa-
tion. To deal with this overload of decisions, we commonly employ what
cognitive psychologists call heuristics. Technically, a heuristic is a general
strategy for solving a problem or coming to a decision. For example, a good
heuristic for solving geometry problems is to start with the conclusion you
are trying to reach and then work backward.
Recent research in cognitive psychology has shown, first, that human be-
ings rely very heavily on heuristics and, second, that we often have too
much confidence in them. The result is that our probability judgments often
go very wrong, and sometimes our thinking gets utterly mixed up. In this
280
CHAPTER 11 ■ Chances
regard, two heuristics are particularly instructive: the representativeness
heuristic and the availability heuristic.
THE REPRESENTATIVENESS HEURISTIC. A simple example illustrates how
errors can arise from the representativeness heuristic. Imagine that you are
randomly dealt five-card hands from a standard deck. Which of the follow-
ing two hands is more likely to come up?
HAND #1 HAND #2
Three of clubs Ace of spades
Seven of diamonds Ace of hearts
Nine of diamonds Ace of clubs
Queen of hearts Ace of diamonds
King of spades King of spades
A surprisingly large number of people will automatically say that the sec-
ond hand is much less likely than the first. Actually, if you think about it for
a bit, it should be obvious that any two specific hands have exactly the same
likelihood of being dealt in a fair game. Here people get confused because
the first hand is unimpressive; and, because unimpressive hands come up
all the time, it strikes us as a representative hand. In many card games, how-
ever, the second hand is very impressive—something worth talking about—
and thus looks unrepresentative. Our reliance on representativeness blinds
us to a simple and obvious point about probabilities: Any specific hand is as
likely to occur as any other.
Linda is thirty-one years old, single, outspoken, and very bright. As a student,
she majored in philosophy, was deeply concerned with issues of discrimina-
tion and social justice, and also participated in antinuclear demonstrations.
Rank the following statements with respect to the probability that they are also
true of Linda, then explain your rankings:
Linda is a teacher in elementary school.
Linda works in a bookstore and takes yoga classes.
Linda is active in the feminist movement.
Linda is a psychiatric social worker.
Linda is a bank teller.
Linda is an insurance salesperson.
Linda is a bank teller and is active in the feminist movement.1
Discussion Question
THE AVAILABILITY HEURISTIC. Because sampling and taking surveys is
costly, we often do it imaginatively, that is, in our heads. If you ask a base-
ball fan which team has the better batting average, Detroit or San Diego, that
person might just remember, might go look it up, or might think about each
281
Some Fallac ies of Probab il i ty
The point of examining these heuristics and noting the errors that they pro-
duce is not to suggest that we should cease relying on them. First, there is a
good chance that this would be psychologically impossible, because the use of
such heuristics seems to be built into our psychological makeup. Second, over a
wide range of standard cases, these heuristics give quick and largely accurate
estimates. Difficulties typically arise in using these heuristics when the situation
is nonstandard—that is, when the situation is complex or out of the ordinary.
To avoid such mistakes when making important judgments about proba-
bilities, we need to ask, “Is the situation sufficiently standard to allow the
use of heuristics?” Because this is a mouthful, we might simply ask, “Is this
the sort of thing that people can figure out in their heads?” When the answer
to that question is “No,” as it often is, then we need to turn to more formal
procedures for determining probabilities.
In a remarkable study,3 Thomas Gilovich, Robert Vallone, and Amos Tversky
found a striking instance of people’s tendency to treat things as statistically
significant when they are not. In professional basketball, certain players have
the reputation of being streak shooters. Streak shooters seem to score points in
batches, then go cold and are not able to buy a basket. Stated more precisely,
in streak shooting, “the performance of a player during a particular period is
significantly better than expected on the basis of the player’s overall record.”
To test whether streak shooting really exists, the authors made detailed study
of a year’s shooting record for the players on the Philadelphia 76ers. This team
Discussion Question
In four pages of a novel (about 2,000 words), how many words would you
expect to find that have the form _ _ _ _ _n_ (seven-letter words with “n” in
the sixth place)? Write down your answer. Now, how many words would
you expect to find that have the form _ _ _ _ing (seven-letter words that end
with “ing”)? Explain your answers.2
Discussion Question
team and try to decide which has the most good batters. The latter approach,
needless to say, would be a risky business, but many baseball fans have re-
markable knowledge of the batting averages of top hitters. Even with this
knowledge, however, it is easy to go wrong. The players that naturally come
to mind are the stars on each team. They are more available to our memory,
and we are likely to make our judgment on the basis of them alone. Yet such
a sample can easily be biased because all the batters contribute to the team
average, not just the stars. The fact that the weak batters on one team are
much better than the weak batters on the other can swing the balance.
(continued)
282
CHAPTER 11 ■ Chances
THE LANGUAGE OF PROBABILITY
The first step in figuring out probabilities is to adopt a more precise way of
talking. Our common language includes various ways of expressing proba-
bilities. Some of the guarding terms discussed in Chapter 3 provide exam-
ples of informal ways of expressing probability commitments. Thus,
someone might say that it is unlikely that the New England Patriots will win
the Super Bowl this year without saying how unlikely it is. We can also spec-
ify various degrees of probability. Looking out the window, we might say
that there is a fifty-fifty chance of rain. More vividly, someone might have
remarked that Ralph Nader does not have a snowball’s chance in hell of ever
winning a presidential election. In each case, the speaker is indicating the
relative strength of the evidence for the occurrence or nonoccurrence of
some event. To say that there is a fifty-fifty chance that it will rain indicates
that we hold that the evidence is equally strong that it will rain rather than
not rain. The metaphor in the third statement indicates that the person who
uttered it believed that the probability of Nader winning a presidential elec-
tion is essentially nonexistent.
We can make our probability claims more precise by using numbers.
Sometimes we use percentages. For example, the weather bureau might say
that there is a 75 percent chance of snow tomorrow. This can naturally be
changed to a fraction: The probability is 3/4 that it will snow tomorrow.
Finally, this fraction can be changed to a decimal expression: There is a 0.75
probability that it will snow tomorrow.
The probability scale has two end points: the absolute certainty that the
event will occur and the absolute certainty that it will not occur. Because you
cannot do better than absolute certainty, a probability can neither rise above
100 percent nor drop below 0 percent (neither above 1, nor below 0). (This
should sound fairly obvious, but it is possible to become confused when
combining percentages and fractions, as when Yogi Berra was supposed to
have said that success is one-third talent and 75 percent hard work.) Of
course, what we normally call probability claims usually fall between these
two endpoints. For this reason, it sounds somewhat peculiar to say that
there is a 100 percent chance of rain and just plain weird to say the chance of
rain is 1 out of 1. Even so, these peculiar ways of speaking cause no proce-
dural difficulties and rarely come up in practice.
included Andrew Toney, noted around the league as being streak shooter. The
authors found no evidence for streak shooting, not even for Andrew Toney.
How would you go about deciding whether streak shooting exists or not? If,
as Gilovich, Vallone, and Tversky have argued, belief phenomenon is a “cog-
nitive illusion,” why do so many people, including most professional athletes,
believe that it does exist?
283
A Pr ior i Probab il i ty
A PRIORI PROBABILITY
When people make probability claims, we have a right to ask why they assign
the probability they do. In Chapter 8, we saw how statistical procedures can be
used for establishing probability claims. Here we will examine the so-called a
priori approach to probabilities. A simple example will bring out the differences
between these two approaches. We might wonder what the probability is of
drawing an ace from a standard deck of fifty-two cards. Using the procedure
discussed in Chapter 8, we could make a great many random draws from the
deck (replacing the card each time) and then form a statistical generalization
concerning the results. We would discover that an ace tends to come up
roughly one-thirteenth of the time. From this we could draw the conclusion
that the chance of drawing an ace is one in thirteen.
But we do not have to go to all this trouble. We can assume that each of the
fifty-two cards has an equal chance of being selected. Given this assumption,
an obvious line of reasoning runs as follows: There are four aces in a standard
fifty-two-card deck, so the probability of selecting one randomly is four in
fifty-two. That reduces to one chance in thirteen. Here the set of favorable
outcomes is a subset of the total number of equally likely outcomes, and to
compute the probability that the favorable outcome will occur, we merely
divide the number of favorable outcomes by the total number of possible
outcomes. This fraction gives us the probability that the event will occur on a
random draw. Since all outcomes here are equally likely,
Probability of drawing an ace = = =
Notice that in coming to our conclusion that there is one chance in thirteen
of randomly drawing an ace from a fifty-two-card deck, we used only math-
ematical reasoning. This illustrates the a priori approach to probabilities. It is
called the a priori approach because we arrive at the result simply by reason-
ing about the circumstances.
In calculating the probability of drawing an ace from a fifty-two-card
deck, we took the ratio of favorable equally likely outcomes to total equally
likely outcomes. Generally, then, the probability of a hypothesis h, symbol-
ized “Pr(h),” when all outcomes are equally likely, is expressed as follows:
Pr(h) =
We can illustrate this principle with a slightly more complicated example.
What is the probability of throwing an eight on the cast of two dice? The
following table lists all of the equally likely ways in which two dice can turn
up on a single cast. Notice that five of the thirty-six possible outcomes produce
an eight. Hence, the probability of throwing an eight is 5/36.
favorable outcomes
total outcomes
1
13
4
52
number of aces
total number of cards
284
CHAPTER 11 ■ Chances
Using the above chart, answer the following questions about the total on
throw of two dice:
1. What is the probability of throwing a five?
2. Which number has the highest probability of being thrown? What is its
probability?
3. What is the probability of throwing an eleven?
4. What is the probability of throwing either a seven or an eleven?
5. Which is more likely: throwing either a five or an eight?
6. Which is more likely: throwing a five or an eight, or throwing a two or a
seven?
7. What is the probability of throwing a ten or above?
8. What is the probability of throwing an even number?
9. What is the probability of throwing an odd number?
10. What is the probability of throwing a value from four to six?
11. What is the probability of throwing either a two or a twelve?
12. What is the probability of throwing a value from two to twelve?
Exercise I
285
Some Rules of Probab il i ty
SOME RULES OF PROBABILITY
Suppose you have determined the probability that certain simple events will
occur; how do you go about applying this information to combinations of
events? This is a complex question and one that can be touched on only
lightly in this text. There are, however, some simple rules of probability that
are worth knowing because they can guide us in making choices when out-
comes are uncertain.
By convention, events are assigned probabilities between 0 and 1 (inclusive).
An event is going to either occur or not occur; that, at least, is certain (that is, it
has a probability of 1). From this it is easy to see how to calculate the probabil-
ity that the event will not occur given the probability that it will occur: We sim-
ply subtract the probability that it will occur from 1. This is our first rule:
RULE 1: NEGATION. The probability that an event will not occur is 1 minus
the probability that it will occur. Symbolically:
Pr(not h) = 1 Pr(h)
For example, the probability of drawing an ace from a standard deck is one
in thirteen, so the probability of not drawing an ace is twelve in thirteen.
This makes sense because there are forty-eight out of fifty-two ways of not
drawing an ace, and this reduces to twelve chances in thirteen.
RULE 2: CONJUNCTION WITH INDEPENDENCE. Given two independent events,
the probability of their both occurring is the product of their individ-
ual probabilities. Symbolically (where h1 and h2 are independent):
Pr(h1 & h2) = Pr(h1) × Pr(h2)
Here the word “independent” needs explanation. Suppose you randomly
draw a card from the deck, then put it back, shuffle, and draw again. In this
case, the outcome of the first draw provides no information about the out-
come of the second draw, so it is independent of it. What is the probability of
drawing two aces in a row using this system? Using Rule 2, we see that the
answer is 1/13 × 1/13, or 1 chance in 169.
The situation is different if we do not replace the card after the first draw.
Rule 2 does not apply to this case because the two events are no longer inde-
pendent. The chances of getting an ace on the first draw are still one in thir-
teen, but if an ace is drawn and not returned to the pack, then there is one
less ace in the deck, so the chances of drawing an ace on the next draw are
reduced to three in fifty-one. Thus, the probability of drawing two consecu-
tive aces without returning the first draw to the deck is 4/52 × 3/51, or 1 in
221, which is considerably lower than 1 in 169.
If we want to extend Rule 2 to cover cases in which the events are not in-
dependent, then we will have to speak of the probability of one event occur-
ring given that another has occurred. The probability that h2 will occur given
that h1 has occurred is called the conditional probability of h2 on h1 and is usu-
ally symbolized thus: Pr(h2|h1). This probability is calculated by considering
286
CHAPTER 11 ■ Chances
only those cases where h1 is true and then dividing the number of cases
within that group where h2 is also true by the total number of cases in that
group. Symbolically:
Pr(h2|h1) = �
Using this notion of conditional probability, Rule 2 can be modified as fol-
lows to deal with cases in which events need not be independent:
RULE 2G: CONJUNCTION IN GENERAL. Given two events, the probability of
their both occurring is the probability of the first occurring times
the probability of the second occurring, given that the first has
occurred. Symbolically:
Pr(h1 & h2) � Pr(h1) × Pr(h2|h1)
Notice that, in the event that h1 and h2 are independent, the probability of h2
is not related to the occurrence of h1, so the probability of h2 on h1 is simply
the probability of h2. Thus, Rule 2 is simply a special case of the more gen-
eral Rule 2G.
We can extend these rules to cover more than two events. For example, with
Rule 2, however many events we might consider, provided that they are inde-
pendent of each other, the probability of all of them occurring is the product of
each one of them occurring. For example, the chances of flipping a coin and
having it come up heads is one chance in two. What are the chances of flipping
a coin eight times and having it come up heads every time? The answer is:
1/2 × 1/2 × 1/2 × 1/2 × 1/2 × 1/2 × 1/2 × 1/2
which equals 1 chance in 256.
Our next rule allows us to answer questions of the following kind: What
are the chances of either an eight or a two coming up on a single throw of the
dice? Going back to the chart, we saw that we could answer this question by
counting the number of ways in which a two can come up (which is one)
and adding this to the number of ways in which an eight can come up
(which is five). We could then conclude that the chances of one or the other
of them coming up is six chances in thirty-six, or 1/6. The principle involved
in this calculation can be stated as follows:
RULE 3: DISJUNCTION WITH EXCLUSIVITY. The probability that at least one of
two mutually exclusive events will occur is the sum of the proba-
bilities that each of them will occur. Symbolically (where h1 and
h2 are mutually exclusive):
Pr(h1 or h2) � Pr(h1) + Pr(h2)
To say that events are mutually exclusive means that they cannot both occur.
You cannot, for example, get both a ten and an eight on the same cast of two
dice. You might, however, throw neither one of them.
outcomes where h1 and h2���
total outcomes where h1
favorable outcomes where h1����
total outcomes where h1
287
Some Rules of Probab il i ty
When events are not mutually exclusive, the rule for calculating disjunc-
tive probabilities becomes more complicated. Suppose, for example, that ex-
actly half the class is female and exactly half the class is over nineteen and
the age distribution is the same for females and males. What is the probabil-
ity that a randomly selected student will be either a female or over nineteen?
If we simply add the probabilities (1/2 + 1/2 � 1), we would get the result
that we are certain to pick someone who is either female or over nineteen.
But that answer is wrong, because a quarter of the class is male and not over
nineteen, and one of them might have been randomly selected. The correct
answer is that the chances are 3/4 of randomly selecting someone who is ei-
ther female or over nineteen.
We can see that this is the correct answer by examining the following table:
Over Nineteen Not over Nineteen
Female 25% 25%
Male 25% 25%
It is easy to see that in 75 percent of the cases, a randomly selected student
will be either female or over nineteen. The table also shows what went
wrong with our initial calculation. The top row shows that 50 percent of the
students are female. The left column shows that 50 percent of the students
are over nineteen. But we cannot simply add these figures to get the proba-
bility of a randomly selected student being either female or over nineteen.
Why not? Because that would double-count the females over nineteen. We
would count them once in the top row and then again in the left column. To
compensate for such double-counting, we need to subtract the students who
are both female and over nineteen. The upper left figure shows that this is
25%. So the correct way to calculate the answer is 50% + 50% – 25% = 75%.
This pattern is reflected in the general rule governing the calculation of
disjunctive probabilities:
RULE 3G: DISJUNCTION IN GENERAL. The probability that at least one of two
events will occur is the sum of the probabilities that each of
them will occur, minus the probability that they will both occur.
Symbolically:
Pr(h1 or h2) = Pr(h1) + Pr(h2) – Pr(h1 & h2)
If h1 and h2 are mutually exclusive, then Pr(h1 & h2) = 0, and Rule 3G reduces
to Rule 3. Thus, as with Rules 2 and 2G, Rule 3 is simply a special case of the
more general Rule 3G.
Before stating Rule 4, we can think about a particular example. What is the
probability of tossing heads at least once in eight tosses of a coin? Here it
is tempting to reason in the following way: There is a 50 percent chance of
getting heads on the first toss and a 50 percent chance of getting heads on the
second toss, so after two tosses it is already certain that we will toss heads
at least once, and thus after eight tosses there should be a 400 percent chance.
288
CHAPTER 11 ■ Chances
In other words, you cannot miss. There are two good reasons for thinking
that this argument is fishy. First, probability can never exceed 100 percent.
Second, there must be some chance, however small, that we could toss a coin
eight times and not have it come up heads.
The best way to look at this question is to restate it so that the first two rules
can be used. Instead of asking what the probability is that heads will come up
at least once, we can ask what the probability is that heads will not come up at
least once. To say that heads will not come up even once is equivalent to say-
ing that tails will come up eight times in a row. By Rule 2 we know how to
compute that probability: It is 1/2 multiplied by itself eight times, and that, as
we saw, is 1/256. Finally, by Rule 1 we know that the probability that this will
not happen (that heads will come up at least once) is 1 – (1/256). In other
words, the probability of tossing heads at least once in eight tosses is 255/256.
That comes close to a certainty, but it is not quite a certainty.
We can generalize these results as follows:
RULE 4: SERIES WITH INDEPENDENCE. The probability that an event will
occur at least once in a series of independent trials is 1 minus the
probability that it will not occur in that number of trials. Symbol-
ically (where n is the number of independent trials):
Pr(h at least once in n trials) = 1 – Pr(not h)n
Strictly speaking, Rule 4 is unnecessary, since it can be derived from Rules 1
and 2, but it is important to know because it blocks a common misunder-
standing about probabilities: People often think that something is a sure
thing when it is not.
Another common confusion is between permutations and combinations.
A permutation is a set of items whose order is specified. A combination is a
set of items whose order is not specified. Imagine, for example, that three
cards—the jack, queen, and king of spades—are facedown in front of you. If
you pick two of these cards in turn, there are three possible combinations:
jack and queen, jack and king, and queen and king. In contrast, there are six
possible permutations: jack then queen, queen then jack, jack then king, king
then jack, queen then king, and king then queen.
Rule 2 is used to calculate probabilities of permutations—of conjunctions
of events in a particular order. For example, if you flip a fair coin twice, what
is the probability of its coming up heads and tails in that order (that is, heads
on the first flip and tails on the second flip)? Since the flips are independent,
Rule 2 tells us that the answer is 1/2 × 1/2 = 1/4. This answer is easily con-
firmed by counting the possible permutations (heads then heads, heads then
tails, tails then heads, tails then tails). Only one of these four permutations
(heads then tails) is a favorable outcome.
We need to calculate probabilities of combinations in a different way. For
example, if you flip a fair coin twice, what is the probability of its landing
heads and tails in any order? There are two ways for this to happen. The coin
could come up either heads then tails or tails then heads. These alternatives
289
Some Rules of Probab il i ty
are mutually exclusive, so the probability of this disjunction by Rule 3 is
1/4 + 1/4 = 1/2. This is confirmed by counting two possibilities (heads then
tails, tails then heads) out of four (heads then heads, heads then tails, tails
then heads, tails then tails). Another way to calculate this probability is to
realize that the first flip doesn’t matter. Whatever you get on the first flip
(heads or tails), you need the opposite on the second flip. You are certain to
get either heads or tails on the first flip, so this probability is 1. Then, regard-
less of what happens on the first flip, the probability of getting the opposite
on the second flip is 1/2. These results are independent, so the probability
of their conjunction by Rule 2 is the product 1 × 1/2 = 1/2.
We can also use our rules to calculate probabilities of combinations with-
out independence. Rule 2G tells us that the probability of drawing an ace,
not putting this card back in the deck, and then drawing a king is 4/52 ×
4/51 = 16/2,652. But what is the probability of drawing an ace and a king in
any order? It is the probability of drawing either an ace or a king and then
drawing the other one given that you drew the first one. That probability by
Rule 2G is 8/52 × 4/51 = 32/2,652. The difference between this result and
the previous one, where the order was specified, shows why we need to de-
termine whether we are dealing with permutations or combinations.
Use the rules of probability to calculate these probabilities:
1. What is the probability of rolling a five on one throw of a fair six-sided die?
2. What is the probability of not rolling a five on one throw of a fair six-sided
die?
3. If you roll a five on your first throw of a fair six-sided die, what is proba-
bility of rolling another five on a second throw of that die?
4. If you roll two fair six-sided dice one time, what are the chances that both
of the dice will come up a five?
5. If you roll two fair six-sided dice one time, what are the chances that one or
the other (or both) of the dice will come up a five?
6. If you roll two fair six-sided dice one time, what are the chances that one
and only one of the dice will come up a five?
7. If you roll two fair six-sided dice one time, what are the chances that at
least one of the dice will come up a five?
8. If you roll two fair six-sided dice one time, what are the chances that at
least one of the dice will not come up a five?
9. If you roll six fair six-sided dice one time, what are the chances that at least
one of the dice will come up a five?
10. If you roll six fair six-sided dice one time, what are the chances that at least
one of the dice will not come up a five?
Exercise II
290
CHAPTER 11 ■ Chances
Compute the probability of making the following draws from a standard fifty-
two-card deck:
1. Drawing either a seven or a five on a single draw.
2. Drawing neither a seven nor a five on a single draw.
3. Drawing a seven and then, without returning the first card to the deck,
drawing a five on the next draw.
4. Same as 3, but the first card is returned to the deck and the deck is
shuffled after the first draw.
5. Drawing at least one spade in a series of three consecutive draws, when
the card drawn is not returned to the deck.
6. Drawing at least one spade in a series of four consecutive draws, when
the card drawn is not returned to the deck.
7. Same as 6, but the card is returned to the deck after each draw and the
deck is reshuffled.
8. Drawing a heart and a diamond in that order in two consecutive draws,
when the first card is returned to the deck and the deck is reshuffled
the first draw.
9. Drawing a heart and a diamond in any order in two consecutive draws,
when the first card is returned to the deck and the deck is reshuffled
the first draw.
10. Drawing a heart and a diamond in any order in two consecutive draws,
when the first card is not returned to the deck after the first draw.
Exercise III
Suppose there are two little lotteries in town, each of which sells exactly
100 tickets.
1. If each lottery has only one winning ticket, and you buy two tickets to the
same lottery, what is the probability that you will have a winning ticket?
2. If each lottery has only one winning ticket, and you buy one ticket to each
of the two lotteries, what is the probability that you will have at least
one winning ticket?
3. If each lottery has only one winning ticket, and you buy one ticket to each
of the two lotteries, what is the probability that you will have two winning
tickets?
4. If each lottery has two winning tickets, and you buy one ticket to each of
the two lotteries, what is the probability that you will have at least one
winning ticket?
5. If each lottery has two winning tickets, and you buy two tickets to the same
lottery, what is the probability that you will have two winning tickets?
6. If each lottery has two winning tickets, and you buy two tickets to the same
lottery, what is the probability that you will have at least one winning ticket?
Exercise IV
291
Bayes ’ s Theorem
1. You are presented with two bags, one containing two ham sandwiches
and the other containing a ham sandwich and a cheese sandwich. You
reach in one bag and draw out a ham sandwich. What is the probability
that the other sandwich in the bag is also a ham sandwich?
2. You are presented with three bags: two contain a chicken-fat sandwich
and one contains a cheese sandwich. You are asked to guess which bag
contains the cheese sandwich. You do so, and the bag you selected is set
aside. (You obviously have one chance in three of guessing correctly.)
From the two remaining bags, one containing a chicken-fat sandwich is
then removed. You are now given the opportunity to switch your
selection to the remaining bag. Will such a switch increase, decrease, or
leave unaffected your chances of correctly ending up with the bag with
the cheese sandwich in it?
Exercise V
In a fair lottery with a million tickets, the probability that ticket #1 will lose is
0.999999, the probability that ticket #2 will lose is also 0.999999, and so on for
each ticket. On this basis, you might seem justified in believing each premise
in the following argument:
Ticket #1 will lose.
Ticket #2 will lose.
Ticket #3 will lose.
[and so on, until:]
Ticket #1,000,000 will lose.
There are only 1,000,000 tickets.
Every ticket will lose.
This conclusion must be false, since the lottery is fair. However, the conclusion
follows from the premises, and each of the premises seems justified. Does this
argument justify us in believing its conclusion? Why or why not?
Discussion Question
BAYES’S THEOREM
Although dice and cards provide nice, simple models for learning how to
calculate probabilities, real life is usually more complicated. One particu-
larly interesting and important form of problem arises often in medicine.
Suppose that Wendy tests positive for colon cancer. The treatment for colon
cancer is painful and dangerous, so, before subjecting Wendy to that treat-
ment, her doctor wants to determine how likely it is that Wendy really has
292
CHAPTER 11 ■ Chances
colon cancer. After all, no test is perfect. Regarding the test that was used on
Wendy, previous studies have revealed the following probabilities:
The probability that a person in the general population has colon cancer
is 0.3 percent (or 0.003).
If a person has colon cancer, then the probability that the test is positive
is 90 percent (or 0.9).
If a person does not have colon cancer, then the probability that the test is
positive is 3 percent (or 0.03).
On these assumptions, what is the probability that Wendy actually has colon
cancer, given that she tested positive? Most people guess that this probabil-
ity is fairly high. Even most trained physicians would say that Wendy prob-
ably has colon cancer.4
What is the correct answer? To calculate the probability that a person
who tests positive actually has colon cancer, we need to divide the number
of favorable outcomes by the number of total outcomes. The favorable out-
comes include everyone who tests positive and really has colon cancer.
This outcome is not “favorable” to Wendy, so we will describe this group as
true positives. The total outcomes include everyone who tests positive.
This includes the true positives plus the false positives, which are those who
test positive but do not have colon cancer. Given the stipulated probabili-
ties, in a normal population of 100,000 people, there will be 270 true posi-
tives (100,000 × 0.003 × 0.9) and 2,991 false positives [(100,000 – 300) × 0.03].
Thus, the probability that Wendy has colon cancer is about 270/(270 +
2,991). That is only about 8.3 percent, when most people estimate above 50
percent!
Why do people, including doctors, overestimate these probabilities so
badly? Part of the answer seems to be that they focus on the rate of true
positives (90 percent) and forget that, because there are so many people
without colon cancer (99.7 percent of the total population), even a small
rate of false positives (3 percent) will yield a large number of false positives
(2,991) that swamps the much smaller number of true positives (270).
(When the question about probability was reformulated in terms of the
number of people in each group, most doctors come up with the correct an-
swer.) For whatever reason, people have a strong tendency to make mis-
takes in cases like these, so we need to be careful, especially when so much
is at stake.
One way to calculate probabilities like these uses a famous theorem that
was first presented by an English clergyman named Thomas Bayes
(1702–1761). A simple proof of this theorem applies the laws of probability
from the preceding section. We want to figure out Pr(h|e), that is, the proba-
bility of the hypothesis h (e.g., Wendy has colon cancer), given the evidence
e (e.g., Wendy tested positive for colon cancer). To get there, we start from
Rule 2G:
1. Pr(e & h) � Pr(e) × Pr(h |e)
293
Bayes ’ s Theorem
Dividing both sides by Pr(e) gives us:
2. Pr(h |e) � �
Pr
P
(e
r(
&
e)
h)
�
If two formulas are logically equivalent, they must have the same probabil-
ity. We can establish by truth tables (as in Chapter 6) that “e” is logically
equivalent to “(e&h) (e&~h).“ Consequently, we may replace “e” in the
denominator of item 2 with “(e&h) (e&~h)” to get:
3. Pr(h |e) �
Since “e&h” and “e&~h” are mutually exclusive, we can apply Rule 3 to the
denominator of item 3 to get:
4. Pr(h |e) �
Finally, we apply Rule 2G to item 4 and get:
BT: Pr(h|e) �
This is a simplified version of Bayes’s theorem.
This theorem enables us to calculate the desired probability in our origi-
nal example:
h = the patient has colon cancer
e = the patient tests positive for colon cancer
Pr(h) = 0.003
Pr(~h) = 1 – Pr(h) = 0.997
Pr(e|h) = 0.9
Pr(e|~h) = 0.03
If we substitute these values into Bayes’s theorem, we get:
Pr(h|e) � � about 0.083
In this way, we can calculate the conditional probability of the hypothesis
given the evidence from its reverse, that is, from the conditional probability
of the evidence given the hypothesis. That is what makes Bayes’s theorem
so useful.
Many people find a different method more intuitive. The first step is
to set up a table. The two factors to be related are (1) whether the patient
has colon cancer and (2) whether the patient tests positive for colon cancer.
0.003 × 0.9
[0.003 × 0.9] + [0.997 × 0.03]
Pr(h) × Pr(e|h)
[Pr(h) × Pr(e|h)] + [Pr(~h) × Pr(e|~h)]
Pr(e & h)
���
Pr(e & h) + Pr(e & �h)
Pr(e & h)
���
Pr[(e & h) (e & �h)]
Next we need to enter a population size in the lower right box. The prob-
abilities will not be affected by the population size, but it is cleaner to pick a
population that is large enough to get whole numbers when the population
is multiplied by the given probabilities. To determine the right size popula-
tion, add the number of places to the right of the decimal point in the two
most specific probabilities, then pick a population of 10 to the power of that
sum. In our example, the most specific probabilities are 0.003 and 0.03, and
3 + 2 = 5, so we can enter 105:
294
CHAPTER 11 ■ Chances
To chart all possible combinations of these two factors, we need a table
like this:
Colon Cancer Not Colon Cancer Total
Test Positive
Do Not Test Positive
Total
Colon Cancer Not Colon Cancer Total
Test Positive
Do Not Test Positive
Total 100,000
This population size represents the total number of people who are tested.
We have no information about the ones who are not tested, so they cannot
figure into our calculations.
The bottom row can now be filled in by dividing the total population into
those who have colon cancer and those who do not have colon cancer. Just
multiply the population size by the probability of colon cancer in the general
population [Pr(h)] to get a number for the second box on the bottom row. This
figure represents the total number of people with colon cancer in this popula-
tion. Then subtract that product from the population size and put the remain-
der in the remaining box. This represents the total number of people without
colon cancer in this population. Since these two groups exhaust the popula-
tion, they must add up to the total population size. In our case, we were given
that the probability that a person in the general population has colon cancer is
0.003. On this basis, we can fill in the bottom row of the table:
Colon Cancer Not Colon Cancer Total
Test Positive
Do Not Test Positive
Total 300 99,700 100,000
295
Bayes ’ s Theorem
Next, fill out the second column by dividing the total number of people
with colon cancer into those who test positive and those who do not test
positive. These numbers can be calculated with the given conditional proba-
bility of testing positive, given colon cancer [Pr(e|h)]. In our example, if a
person has colon cancer, the probability that the test is positive is 0.9. Thus,
270 (=0.9 × 300) of the people in the colon cancer column will test positive
and the rest (300 – 270 = 30) will not, so we get these figures:
Colon Cancer Not Colon Cancer Total
Test Positive 270
Do Not Test Positive 30
Total 300 99,700 100,000
Similarly, we can fill out the third column by dividing the total number of
people without colon cancer into those who test positive and those who do
not test positive. Here we use the conditional probability of a positive test,
given that a person does not have colon cancer [Pr(e|~h)]. This probability
was given as 0.03, and 0.03 × 99,700 = 2,991. This number means that, out of
a normal population of 99,700 without colon cancer, 2,991 will test positive.
Since the figures in this column must add up to a total of 99,700, the remain-
ing figure is 99,700 – 2,991 = 96,709:
Colon Cancer Not Colon Cancer Total
Test Positive 270 2,991
Do Not Test Positive 30 96,709
Total 300 99,700 100,000
Colon Cancer Not Colon Cancer Total
Test Positive 270 2,991 3,261
Do Not Test Positive 30 96,709 96,739
Total 300 99,700 100,000
Finally, we can fill out the fourth column by calculating total numbers of
people who test positive or do not test positive. Simply add across the rows:
Check your calculations by adding the right column: 3,261 + 96,739 = 100,000.
Now that our population is divided up, the solution is staring you in the
face. This table shows us that, in a normal population of 100,000 tested
people distributed according to the given probabilities, a total of 3,261 will
296
CHAPTER 11 ■ Chances
test positive. Out of those, 270 will have colon cancer. Thus, the probabil-
ity that the patient has colon cancer, given that this patient tested positive,
is 270/3,261, which is about 0.083 or 8.3 percent, just as before.
You can also read off other conditional probabilities. If you want to know
the conditional probability of not having colon cancer, given that your test did
not come out positive, then you need to look at the row for those who do not
test positive. The figure at the right of this row tells you that a total of 96,739
out of the total population do not test positive. The column under “Not Colon
Cancer” then tells you that 96,709 of these do not have colon cancer. Thus, the
conditional probability of not having colon cancer given your test did not
come out positive is 96,709/96,739 or about 0.9997. That means that, if you test
negative, the odds are extremely high that you do not have colon cancer.
Tables like these work by dividing the population into groups. We
already learned some names for these groups:
Hypothesis (h) Not Hypothesis (~h)
Evidence (e) True Positives False positives
Not Evidence (~e) False Negatives True Negatives
Population
False positives are sometimes also called false alarms, and false negatives are
sometimes called misses. A little more terminology is also common:
Pr(h) = base rate or prevalence or prior probability
Pr(h|e) = solution or posterior probability
Pr(e|h) = sensitivity of the test
Pr(~e|~h) = specificity of the test
1 – Pr(e|h) = 1 – sensitivity = false negative rate
1 – Pr(~e|~h) = 1 – specificity = false positive rate
You don’t need to use these terms in order to calculate the probabilities, but
it is useful to learn them so that you will be able to understand people who
discuss these issues.
One of the most important lessons of Bayes’s theorem is that the base rate
has big effects. To see how much it matters, let’s recalculate the solution
[Pr(h|e)] in our colon cancer example for different values of the base rate
[Pr(h)] using the same test with the same sensitivity (Pr(e|h) = 0.9) and
specificity [Pr(~e|~h) = 0.97]:
If Pr(h) = 0.003, then Pr(h|e) = 0.083
If Pr(h) = 0.03, then Pr(h|e) = 0.48
If Pr(h) = 0.3, then Pr(h|e) = 0.93
297
Bayes ’ s Theorem
These calculations show that a positive test result for a given test means a
lot more when the base rate is high than when it is low. Thus, if doctors use
the specified test as a screening test in the general population, and if the rate of
colon cancer in that general population is only 0.003, then a positive test result
by itself does not show that the patient has cancer. In contrast, if doctors in-
stead use the specified test as a diagnostic test only for people with certain
symptoms, and if the rate of colon cancer among people with those symptoms
is 0.3, then a positive test result does show that the patient probably has can-
cer, though the test still might be mistaken. Bayes’s theorem, thus, reveals the
right ways and the wrong ways to use and interpret such tests.
Notice also what happens to the probabilities when additional tests are per-
formed. In our original example, one positive test result raises the probability
of cancer from the base rate of 0.003 to our solution of 0.083. Now suppose that
the doctor orders an additional independent test, and the result is again posi-
tive. To apply Bayes’s theorem at this point, we can take the probability after
the original positive test result (0.083) as the prior probability or base rate in
calculating the probability after the second positive test result. This method
makes sense because we are now interested not in the general population but
only in the subpopulation that already tested positive on the first test. The so-
lution after two tests [Pr(h|e)], where “e” is now two independent positive test
results in a row, is 0.731. Next, if the doctor orders a third independent test,
and if the result is positive yet again, then Pr(h|e) increases to 0.988. Bayes’s
theorem, thus, reveals the technical rationale behind the commonsense prac-
tice of ordering additional tests. Problems arise only when doctors put too
much faith in a single positive test result without doing any additional tests.
Construct tables to confirm these calculations of Pr(h|e) for base rates of 0.03
and 0.3.
Exercise VI
Construct tables to confirm the above calculations of probabilities after a sec-
ond and third positive test result.
Exercise VII
1. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the probability that a person in the general population has colon cancer
only 0.1 percent (or 0.001)?
Exercise VIII
(continued)
298
CHAPTER 11 ■ Chances
2. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the probability that a person in the general population has colon cancer
1 percent (0.01)?
3. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the conditional probability that the test is positive, given that the patient
has colon cancer, is only 50 percent (or 0.5)?
4. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the conditional probability that the test is positive, given that the patient
has colon cancer, is 99 percent (or 0.99)?
5. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the conditional probability that the test is positive, given that the patient
does not have colon cancer, is 1 percent (or 0.01)?
6. What would Wendy’s chances of having colon cancer be if the other
probabilities remained the same as in the original example, except that
the conditional probability that the test is positive, given that the patient
does not have colon cancer, is 10 percent (0.1)?
7. Chris tested positive for cocaine once in a random screening test. This
test has a sensitivity and specificity of 95 percent, and 20 percent of the
students in Chris’s school use cocaine. What is the probability that Chris
really did use cocaine?
8. As in problem 7, 20 percent of the students in Chris’s school use cocaine,
but this time Chris tests positive for cocaine on two independent tests,
both of which have a sensitivity and specificity of 95 percent. Now what is
the probability that Chris really did use cocaine?
9. In your neighborhood, 20 percent of the houses have high levels of radon
gas in their basements, so you ask an expert to test your basement. An in-
expensive test comes out positive in 80 percent of the basements that actu-
ally have high levels of radon, but it also comes out positive in 10 percent
of the basements that do not have high levels of radon. If this inexpensive
test comes out positive in your basement, what is the probability that
there is a high level of radon gas in your basement?
10. A more expensive test for radon is also more accurate. It comes out posi-
tive in 99 percent of the basements that actually have high levels of radon.
It also tests positive in 2 percent of the basements that do not high levels
of radon. As in problem 7, 20 percent of the houses in your neighborhood
have radon in their basement. If the expensive test comes out positive in
your basement, what is the probability that there is a high level of radon
gas in your basement?
11. Late last night a car ran into your neighbor and drove away. In your town,
there are 500 cars, and 2 percent of them are Porsches. The only eyewitness
to the hit-and-run says the car that hit your neighbor was a Porsche. Tested
299
Bayes ’ s Theorem
under similar conditions, the eyewitness mistakenly classifies cars of other
makes as Porsches 10 percent of the time, and correctly classifies Porsches
as such 80 percent of the time. What are the chances that the car that hit
your neighbor really was a Porsche?
12. Late last night a dog bit your neighbor. In your town, there are 400 dogs,
95 percent of them are black Labrador retrievers, and the rest are pit bulls.
The only eyewitness to the event, a veteran dog breeder, says that the dog
who bit your neighbor was a pit bull. Tested under similar low-light con-
ditions, the eyewitness mistakenly classifies black Labs as pit bulls only 2
percent of the time, and correctly classifies pit bulls as pit bulls 90 percent
of the time. What are the chances that dog who bit your neighbor really
was a pit bull?
13. In a certain school, the probability that a student reads the assigned pages
before a lecture is 80 percent (or 0.8). If a student does the assigned read-
ing in advance, then the probability that the student will understand the
lecture is 90 percent (or 0.9). If a student does not do the assigned reading
in advance, then the probability that the student will understand the lec-
ture is 10 percent (or 0.1). What is the probability that a student did the
reading in advance, given that she did understand the lecture? What is the
probability that a student did not do the reading in advance, given that
she did not understand the lecture?
14. In a different school, the probability that a student reads the assigned pages
before a lecture is 60 percent (or 0.6). If a student does the assigned reading
in advance, then the probability that, when asked, the student will tell the
professor that he did the reading is 100 percent (or 1.0). If a student does not
do the assigned reading in advance, then the probability that, when asked,
the student will tell the professor that he did the reading is 70 percent (or
0.7). What is the probability that a student did the reading in advance, given
that, when asked, he told the professor that he did the reading? What is the
probability that a student did not do the reading in advance, given that,
when asked, he told the professor that he did not do the reading?
Should Sally Clark be found guilty of murder? Why or why not? The follow-
ing account of her case was extracted from http://pass.maths.org.uk/
issue21/features/clark (10/22/04):
Five years ago, a young couple from Cheshire suffered one of the most devas-
tating losses imaginable—their baby Christopher died in his sleep, aged 11
weeks. Doctors, neighbours, all were sympathetic, and the death was certified
as natural causes—there was evidence of a respiratory infection, and no sign
of any failure of care.
Discussion Question
(continued)
http://pass.maths.org.uk/issue21/features/clark
http://pass.maths.org.uk/issue21/features/clark
300
CHAPTER 11 ■ Chances
But just a year later, in what must have felt like a horribly familiar night-
mare, the Clarks’ second child Harry died, aged 8 weeks. This time, there was
no sympathy from the professionals. Four weeks after Harry’s death the cou-
ple were arrested, and eventually Sally Clark was charged with murdering
both children. She was tried and convicted in 1999 and is now almost three
years into a life sentence.
The forensic evidence was slim to nonexistent—certainly neither case would
have stood up alone. . . . So how come Sally Clark is serving life in prison?
Simply put, because [of] a seemingly authoritative statement by pediatrician
Sir Roy Meadow . . . that the chance of two children in the same (affluent non-
smoking) family both dying a cot death was 1 in 73 million. . . .
INDEPENDENCE
This statistic of 1 in 73 million came from the Confidential Enquiry for Still-
births and Deaths in Infancy (CESDI), an authoritative and detailed study of
deaths of babies in five regions of England between 1993 and 1996. There it is
estimated that the chances of a randomly chosen baby dying a cot death are
1 in 1,303. If the child is from an affluent, nonsmoking family, with the mother
over 26, the odds fall to around 1 in 8,500. The authors go on to say that if
there is no link between cot deaths of siblings (because we have eliminated
the biggest known and possibly shared factors influencing cot death rates)
then we would be able to estimate the chances of two children from such
a family both suffering a cot death by squaring 1/8,500—giving 1 chance in
73 million.
So far so good. But are cot deaths in the same family really independent?
The website for the Foundation for the Study of Infant Death (FSID) says
baldly that “second cot deaths in the same family are very rare.“ This is no
help, because so are first-sibling cot deaths—what we need to know is, how
comparatively rare? Does having one child die a cot death increase the
chances that you will have another do so?
Ray Hill . . . estimates that siblings of children who die of cot death are be-
tween 10 and 22 times more likely than average to die the same way. Using the
figure of 1 in 1,303 for the chance of a first cot death, we see that the chances of
a second cot death in the same family are somewhere between 1 in 60 and 1 in
130. There isn’t enough data to be more precise, or to take familial factors into
account, but it seems reasonable to use a ballpark figure of 1 in 100. Multiply-
ing 1/1,303 by 1/100 gives an estimate for the incidence of double cot death of
around 1 in 130,000. . . .
BAYES’S THEOREM
. . . Very possibly, as you’re reading this, you are . . . thinking “okay, so
the odds aren’t as extreme as 1 in 73 million, but they’re still astronomi-
cally high. There’s not that much difference between odds of 1 in 73 million
301
Bayes ’ s Theorem
and 1 in 100,000, so Sally Clark must still be guilty.” If so, you‘re commit-
ting the “Prosecutor’s Fallacy.“
Simply put, this is the incorrect belief that the chance of a rare event hap-
pening is the same as the chance that the defendant is innocent. Even with the
more accurate figure of 1 in 100,000 for the chance that a randomly chosen pair
of siblings will both die of cot death, this is not the chance that Sally Clark is
innocent. It is the chance that an arbitrary family will lose two children in cot
deaths. . . .
In mathematical language, we need to find the conditional probabilities of
the various possible causes of death, given the fact that the children died. If H
is some hypothesis (for example, that both of Sally Clark’s children died of
natural causes) and D is some data (that both children are dead), we want to
find the probability of the hypothesis given the data, which is written
P(H|D). Let’s write A for the alternate hypothesis—that Sally Clark mur-
dered both her children. We will discount all other possibilities, for example
that someone else murdered both children, or that Sally Clark murdered only
one of them.
P(H|D) �
This is not as complicated as it looks. We already have an estimate for P(H)
of 1/100,000. . . . Trivially, both P(D|H) and P(D|A) are equal to 1. These
numbers are the probabilities that two of the children are dead, given that
that two of the children have died of natural causes, or been murdered,
respectively.
A completely accurate version of Bayes’s Theorem would take into account
all sorts of factors—for example, the fact that social services had not been
involved with the Clark family, their income level, and so on—but there isn’t a
sufficient amount of data available to do this. However, if we are only looking
to analyse the case against Sally Clark, it is sufficient . . . to make reasonable
estimates and show that these lead to reasonable doubt.
P(A) is the most difficult figure to estimate. It is the probability that a
randomly chosen pair of siblings will both be murdered. Statistics on such
double murders are pretty much nonexistent, because child murders are
so rare (far, far more rare than cot deaths) and because in most cases,
someone known to have murdered once is not free to murder again. So
we fall back on the Home Office statistic that fewer than 30 children are
known to be murdered by their mother each year in England and Wales.
Since 650,000 are born each year, and murders of pairs of siblings are clearly
rarer than single murders, we should use a figure much smaller than
30/650,000 = 0.000046. We will put a number ten times as small here—
0.0000046—which is almost certainly overestimating the incidence rate of
double murder.
Now we get a rough and ready estimate of Sally Clark’s innocence:
P(H|D) � � >
2
3
0.00001
0.00001 + 0.0000046
P(H)
P(H) + P(A)
P(D|H)P(H)
P(D|H)P(H) + P(D|A)P(A)
302
CHAPTER 11 ■ Chances
I must warn the reader that this figure isn’t intended to be in any way an accu-
rate estimate of the likelihood that Sally Clark is innocent. It is only meant to
show that, with some reasonable estimates of the likelihoods of relevant
events, the scales come down on the side of her innocence rather than her
guilt. The only way to disagree with this analysis is to assume that literally
hundreds, or thousands, of mothers murder their children undetected every
year. The Campaign to Free Sally Clark say that nearly fifty families have con-
tacted them to say that they have suffered double cot deaths. Are we to believe
that the majority of these couples are murderers, confident or mad enough to
draw attention to themselves in this way? . . .
Plus readers may be interested to know that Sally Clark’s case has been referred to
the Appeal Court for a second time. Evidence, not made available to her original de-
fence team, has recently come to light that at the time of Harry’s death, he was suffer-
ing from a bacterial blood infection known to cause sudden infant death.
NOTES
1 Amos Tversky and Daniel Kahneman, “Extensional Versus Intuitive Reasoning: The Conjunc-
tion Fallacy in Probability Judgment.” Psychological Review 90 (1983): 297.
2 Ibid.
3 Thomas Gilovich, Robert Vallone, and Amos Tversky, “The Hot Hand in Basketball: The
Misperception of Random Sequences,” Cognitive Psychology 17 (1985): 295–314. The quotation is
from pages 295–6.
4 See Gerd Gigerenzer, Calculated Risk: How to Know When Numbers Deceive You (New York:
Simon & Schuster, 2003).
CHOICES
Probabilities are used not only when we determine what to believe but also when we
choose what to do. Although we sometimes assume that we know how our actions
will turn out, we often have to make decisions in the face of risk, when we do not
know what the outcomes of our options will be, but we do know the probabilities
of those outcomes. To help us assess reasoning about choices involving risk, this
chapter will explain the notions of expected monetary value and expected overall
value. Our most difficult choices arise, however, when we do not know even the
probabilities of various outcomes. Such decisions under ignorance or uncertainty
pose special problems, for which a number of rules have been proposed. Although
these rules are useful in many situations, their limitations will also be noted.
EXPECTED MONETARY VALUE
It is obvious that having some sense of probable outcomes is important for
running our lives. If we hear that there is a 95 percent chance of rain, this
usually provides a good enough reason to call off a picnic. But the exact re-
lationship between probabilities and decisions is complex and often misun-
derstood.
The best way to illustrate these misunderstandings is to look at lotteries
in which the numbers are fixed and clear. A $1 bet in a lottery might make
you as much as $10 million. That sounds good. Why not take a shot at $10
million for only a dollar? Of course, there is not much chance of winning the
lottery—say, only 1 chance in 20 million—and that sounds bad. Why throw
$1 away on nothing? So we are pulled in two directions. What we want to
know is just how good the bet is. Is it, for example, better or worse than a
wager in some other lottery? To answer questions of this kind, we need to
introduce the notion of expected monetary value.
The idea of expected monetary value takes into account three features
that determine whether a bet is financially good or bad: the probability
of winning, the net amount you gain if you win, and the net amount you
lose if you lose. Suppose that on a $1 ticket there is 1 chance in 20 million
of winning the New York State Lottery, and you will get $10 million from
the state if you do. First, it is worth noticing that, if the state pays you
12
303
304
CHAPTER 12 ■ Cho ices
$10 million, your net gain on your $1 ticket is only $9,999,999. The state,
after all, still has your original $1. So the net gain equals the payoff minus
the cost of betting. This is not something that those who win huge lotteries
worry about, but taking into account the cost of betting becomes important
when this cost becomes high relative to the size of the payoff. There is noth-
ing complicated about the net amount that you lose when you lose on a
$1 ticket: It is $1.1
We can now compute the expected monetary value or financial worth of
a bet in the following way:
Expected monetary value =
(the probability of winning times the net gain in money of winning) minus
(the probability of losing times the net loss in money of losing)
In our example, a person who buys a $1 ticket in the lottery has 1 chance in
20 million of a net gain of $9,999,999 and 19,999,999 chances in 20 million of
a net loss of a dollar. So the expected monetary value of this wager equals:
(1/20,000,000 × $9,999,999) (19,999,999/20,000,000 × $1)
That comes out to $0.50.
What does this mean? One way of looking at it is as follows: If you could
somehow buy up all the lottery tickets and thus ensure that you would win,
your $20 million investment would net you $10 million, or $0.50 on the
dollar—certainly a bad investment. Another way of looking at the situation
is as follows: If you invested a great deal of money in the lottery over many
millions of years, you could expect to win eventually, but, in the long run,
you would be losing fifty cents on every ticket you bought. One last way of
looking at the situation is this: You go down to your local drugstore and buy
a blank lottery ticket for $0.50. Since it is blank, you have no chance of win-
ning, with the result that you lose $0.50 every time you bet. Although almost
no one looks at the matter in this way, this is, in effect, what you are doing
over the long run when you buy lottery tickets.
We are now in a position to distinguish favorable and unfavorable ex-
pected monetary values. The expected monetary value is favorable when it
is greater than zero. Changing our example, suppose the chances of hitting a
$20 million payoff on a $1 bet are 1 in 10 million. In this case, the state still
has the $1 you paid for the ticket, so your gain is actually $19,999,999. The
expected monetary value is calculated as follows:
(1/10,000,000 × $19,999,999) (9,999,999/10,000,000 × $1)
That comes to $1. Financially, this is a good bet; for in the long run you will
gain $1 for every $1 you bet in such a lottery.
The rule, then, has three parts: (1) If the expected monetary value of the
bet is more than zero, then the expected monetary value is favorable. (2) If
the expected monetary value of the bet is less than zero, then the expected
monetary value is unfavorable. (3) If the expected monetary value of the bet
is zero, then the bet is neutral—a waste of time as far as money is concerned.
305
Expected Monetary Value
Compute the probability and the expected monetary value for the following
bets. Each time, you lay down $1 to bet that a certain kind of card will appear
from a standard fifty-two-card deck. If you win, you collect the amount indi-
cated, so your net gain is $1 less. If you lose, of course, you lose your $1.
Example: Draw a seven of spades. Win: $26.
Probability of winning = 1/52
Expected value: [1/52 × $(26 1)] – (51/52 × $1) = $0.50
1. Draw a seven of spades or a seven of clubs. Win: $26.
2. Draw a seven of any suit. Win: $26.
3. Draw a face card (jack, queen, or king). Win: $4.
4. Do not draw a face card (jack, queen, or king). Win: $2.
5. On two consecutive draws without returning the first card to the deck,
draw a seven of spades and then a seven of clubs. Win: $1,989.
6. Same as in problem 3, but this time the card is returned to the deck and
the deck is shuffled before the second draw. Win: $1,989.
7. On two consecutive draws without returning the first card to the deck,
do not draw a club. Win: $1.78.
8. Same as in problem 7, but this time the card is returned to the deck and
the deck is shuffled before the second draw. Win: $1.78.
9. On four consecutive draws without returning any cards to the deck, a
seven of spades, then a seven of clubs, then a seven of hearts, and then
seven of diamonds. Win: $1,000,001.
10. On four consecutive draws without returning any cards to the deck, draw
four sevens in any order. Win: $1,000,001.
Exercise I
Fogelin’s Palace in Border, Nevada, offers the following unusual bet. If you
win, you make a 50 percent profit on your bet; if you lose, you take a 40 per-
cent loss. That is, if you bet $1 and win, then you get back $1.50; if you bet $1
and lose, you get back $0.60. The chances of winning are fifty-fifty. This sounds
like a marvelous opportunity, but there is one hitch: To play, you must let your
bet ride with its winnings, or losses, for four plays. For example, with $100,
a four-bet sequence might look like this:
Win Win Lose Win
Total $150 $225 $135 $202.50
At the end of this sequence, you can pick up $202.50, and thus make a $102.50
profit. It seems that Fogelin’s Palace is a good place to gamble, but consider
Exercise II
(continued)
306
CHAPTER 12 ■ Cho ices
EXPECTED OVERALL VALUE
Given that lotteries usually have an extremely unfavorable expected mone-
tary value, why do millions of people invest billions of dollars in them each
year? Part of the answer is that some people are stupid, superstitious, or
both. People will sometimes reason, “Somebody has to win; why not me?”
They can also convince themselves that their lucky day has come. But that is
not the whole story, for most people who put down money on lottery tickets
realize that the bet is a bad bet, but think that it is worth doing anyway. Peo-
ple fantasize about what they will do with the money if they win, and fan-
tasies are fun. Furthermore, if the bet is only $1, and the person making the
bet is not desperately poor, losing is not going to hurt much. Even if the ex-
pected monetary value on the lottery ticket is the loss of fifty cents, this
might strike someone as a reasonable price for the fun of thinking about
winning. (After all, you accept a sure loss of $8 every time you pay $8 to see
a movie.) So a bet that is bad from a purely monetary point of view might be
acceptable when other factors are considered.
The reverse situation can also arise: A bet may be unreasonable, even
though it has a positive expected monetary value. Suppose, for example,
that you are allowed to participate in a lottery in which a $1 ticket gives you
1 chance in 10 million of getting a payoff of $20 million. Here, as noted
above, the expected monetary value of a $1 bet is a profit of $1, so from the
point of view of expected monetary value, it is a good bet. This makes it
sound reasonable to bet in this lottery, and a small bet probably is reason-
able. But under these circumstances, would it be reasonable for you to sell
everything you owned to buy lottery tickets? The answer to this is almost
the following argument on the other side. Because the chances of winning are
fifty-fifty, you will, on the average, win half the time. But notice what happens
in such a case:
Win Lose Lose Win
Total $150 $90 $54 $81
So, even though you have won half the time, you have come out $19 behind.
Surprisingly, it does not matter what order the wins and losses come in; if
two are wins and two are losses, you come out behind. (You can check this.)
So, because you are only going to win roughly half the time, and when you
win half the time you actually lose money, it now seems to be a bad idea to
gamble at Fogelin’s Palace.
What should you do: gamble at Fogelin’s Palace or not? Why?
307
Expected Overall Value
certainly no, for, even though the expected monetary value is positive, the
odds of winning are still low, and the loss of your total resources would be
personally catastrophic.
When we examine the effects that success or failure will have on a particu-
lar person relative to his or her own needs, resources, preferences, and so on,
we are then examining what we shall call the expected overall value or expected
utility of a choice. Considerations of this kind often force us to make adjust-
ments in weighing the significance of costs and payoffs. In the examples we
just examined, the immediate catastrophic consequences of a loss outweigh
the long-term gains one can expect from participating in the lottery.
Another factor that typically affects the expected overall value of a bet is
the phenomenon known as the diminishing marginal value or diminishing mar-
ginal utility of a payoff as it gets larger. Suppose someone offers to pay a debt
by buying you a hamburger. Provided that the debt matches the cost of a
hamburger and you feel like having one, you might go along with this. But
suppose this person offers to pay off a debt ten times larger by buying you
ten hamburgers? The chances are that you will reject the offer, for even
though ten hamburgers cost ten times as much as one hamburger, they are
not worth ten times as much to you. At some point you will get stuffed and
not want any more. After one or two hamburgers, the marginal value of one
more hamburger becomes pretty low. The notion of marginal value applies
to money as well. If you are starving, $10 will mean a lot to you. You might
be willing to work hard to get it. If you are wealthy, $10 more or less makes
little difference; losing $10 might only be an annoyance.
Because of this phenomenon of diminishing marginal value, betting on
lotteries is an even worse bet than most people suppose. A lottery with a
payoff of $20 million sounds attractive, but it does not seem to be twenty
times more attractive than a payoff of $1 million. So even if the expected
monetary value of your $1 bet in a lottery is the loss of $0.50, the actual value
to you is really something less than this, and so the bet is even worse than it
seemed at first.
In general, then, when payoffs are large, the expected overall value of the
payoff to someone is reduced because of the effects of diminishing marginal
value. But not always. It is possible to think of exotic cases in which expected
overall value increases with the size of the payoff. Suppose a witch told you
that she would turn you into a toad if you did not give her $10 million by
tomorrow. You believe her, because you know for a fact that she has turned
others into toads when they did not pay up. You have only $1 to your name,
but you are given the opportunity to participate in the first lottery described
above, where a $1 ticket gives you 1 chance in 20 million of hitting a $10 mil-
lion payoff. We saw that the expected monetary value of that wager was an
unfavorable negative $0.50. But now consider the overall value of $1 to you if
you are turned into a toad. Toads have no use for money, so to you, as a toad,
the value of the dollar would drop to nothing. Thus, unless some other, more
attractive alternatives are available, it would be reasonable to buy a lottery
ticket, despite the unfavorable expected monetary value of the wager.
308
CHAPTER 12 ■ Cho ices
1. Though the situation is somewhat far-fetched, suppose you are going to
the drugstore to buy medicine for a friend who will die without it. You
have only $10—exactly what the medicine costs. Outside the drugstore a
young man is playing three-card monte, a simple game in which the
dealer shows you three cards, turns them over, shifts them briefly from
hand to hand, and then lays them out, face down, on the top of a box. You
are supposed to identify a particular card (usually the ace of spades); and,
if you do, you are paid even money. You yourself are a magician and
know the sleight-of-hand trick that fools most people, and you are sure
that you can guess the card correctly nine times out of ten. First, what is
the expected monetary value of a bet of $10? In this context, would it be
reasonable to make this bet? Why or why not?
2. Provide an example of your own where a bet can be reasonable even
though the expected monetary value is unfavorable. Then provide
another example where the bet is unreasonable even though the expected
monetary value is favorable. Explain what makes these bets reasonable or
unreasonable.
Exercise III
Consider the following game: You flip a coin continuously until you get tails
once. If you get no heads (tails on the first flip), then you are paid nothing. If
you get one heads (tails on the second flip), then you are paid $2. If you get
two heads (tails on the third flip), then you are paid $4. If you get three heads,
then you are paid $8. And so on. The general rule is that for any number n,
if you get n heads, then you are paid $2n. What is the expected monetary value
of this game? What would you pay to play this game? Why that amount rather
than more or less?
Discussion Question
DECISIONS UNDER IGNORANCE
So far we have discussed choices where the outcomes of the various options
are not certain, but we know their probabilities. Decisions of this kind are
called decisions under risk. In other cases, however, we do not know the prob-
abilities of various outcomes. Decisions of this kind are called decisions under
ignorance (or, sometimes, decisions under uncertainty). If we do not have any
idea where the probabilities of various outcomes lie, the ignorance is com-
plete. If we know that these probabilities lie within some general range, the
ignorance is partial.
As an example of partial ignorance, suppose that, just after graduating
from college, you are offered three jobs. First, the Exe Company offers you a
309
Dec is ions Under Ignorance
salary of $20,000. Exe is well-established and secure. The next offer comes
from the Wye Company. Here the salary is $30,000, but Wye is a new com-
pany, so it is less secure. You think that this new company will probably do
well, but you don’t know how likely it is to last or for how long. Wye might
go bankrupt, and then you will be left without a job. The final offer comes
from the Zee Company, which is as stable as Exe and offers you a salary of
$40,000 per year. These offers are summarized in the following table:
Wye does not go bankrupt Wye goes bankrupt
Take job at Exe $20,000 $20,000
Take job at Wye $30,000 $0
Take job at Zee $40,000 $40,000
Let’s assume that other factors (such as benefits, vacations, location, interest,
working conditions, bonuses, raises, and promotions) are all equally desir-
able in the three jobs. Which job should you take?
The answer is clear: Take the job from the Zee Company. This decision is
easy because you end up better off regardless of whether or not Wye goes
bankrupt, so it doesn’t matter how likely Wye’s bankruptcy is. Everyone
agrees that you should choose any option that is best whatever happens.
This is called the rule of dominance.
The problem with the rule of dominance is that it can’t help you make
choices when no option is better regardless of what happens. Suppose you
discover that the letter from the Zee Company is a forgery—part of a cruel
joke by your roommate. Now your only options are Exe and Wye. The job
with Wye will be better if Wye does not go bankrupt, but the job with Exe
will be better if Wye does go bankrupt. Neither job is better no matter what
happens, so the rule of dominance no longer applies.
To help you choose between Exe and Wye, you might look for a rational
way to assign probabilities despite your ignorance of which assignments are
correct. One approach of this kind uses the rule of insufficient reason: When
you have no reason to think that any outcome is more likely than any other,
assume that the outcomes are equally probable. This assumption enables us
to calculate expected monetary value or utility, as in the preceding sections,
and then we can choose the option with the highest expected utility. In our
example, this rule of insufficient reason favors the job at Exe, because your
expected income in that job is $20,000, whereas your expected income in the
job at Wye is only $15,000 (= 0.5 × $30,000), assuming that the Wye company
has as much chance of going bankrupt as of staying in business.
The problem with the rule of insufficient reason is that it may seem arbi-
trary to assume that unknown probabilities are equal. Often we suspect that
the probabilities of various outcomes are not equal, even while we do not
know what the probabilities are. Moreover, the rule of insufficient reason
yields different results when the options are described differently. We can
distinguish four possibilities: Wye goes bankrupt, Wye stays the same size,
Wye increases in size, and Wye decreases in size but stays in business. If we
310
CHAPTER 12 ■ Cho ices
do not have any reason to see any of these outcomes as more likely than any
other, then the rule of insufficient reason tells us to assign them equal proba-
bilities. On that assumption, and if you will keep your job as long as Wye
stays in business, then you have only one chance in four of losing your job;
so your expected income in the job at Wye is now $22,500 (= 3⁄4 × $30,000).
Thus, if we stick with the rule of insufficient reason, the expected value of
the job at Wye and whether you should take that job seem to depend on how
the options are divided. That seems crazy in this case.
Another approach tries to work without any assumptions about probabil-
ity in cases of ignorance. Within this approach, several rules might be
adopted. One possibility is the maximax rule, which tells you to choose the op-
tion whose best outcome is better than the best outcome of any other option.
If you follow the maximax rule, then you will accept the job with the Wye
Company, because the best outcome of that job is a salary of $30,000 when
this new company does not go bankrupt, and this is better than any outcome
with the Exe Company. Optimists and risk takers will favor this rule.
Other people are more pessimistic and tend to avoid risks. They will favor
a rule more like the maximin rule, which says to choose the option whose worst
outcome is better than the worst outcome of any other option. If you follow
the maximin rule, you will accept the job with the Exe Company, because the
worst outcome in that job is a steady salary of $20,000, whereas the worst out-
come is unemployment if you accept the job with the Wye Company.
Each of these rules works by focusing exclusively on part of your infor-
mation and disregarding other things that you know. The maximax rule con-
siders only the best outcomes for each option—the best-case scenario. The
maximin rule pays attention to only the worst outcome for each option—the
worst-case scenario. Because they ignore other outcomes, the maximax rule
strikes many people as too risky (since it does not consider how much you
could lose by taking a chance), and the maximin rule strikes many people as
too conservative (since it does not consider how much you could have
gained if you had taken a small risk).
Another problem is that the maximax and maximin rules do not take
probabilities into account at all. This makes sense when you know nothing
about the probabilities. But when some (even if limited) information about
probabilities is available, then it seems better to use as much information as
you have. Suppose, for example, that each of two options might lead to dis-
aster, and you do not know how likely a disaster is after either option, but
you do know that one option is more likely to lead to disaster than another.
In such situations, some decision theorists argue that you should choose the
option that minimizes the chance that any disaster will occur. This is called
the disaster avoidance rule.
To illustrate this rule, consider a different kind of case:
A forty-year-old man is diagnosed as having a rare disease and consults the
world’s leading expert on the disease. He is informed that the disease is almost
certainly not fatal but often causes serious paralysis that leaves its victims
311
Dec is ions Under Ignorance
bedridden for life. (In other cases it has no lasting effects.) The disease is so rare
that the expert can offer only a vague estimate of the probability of paralysis: 20
to 60 percent. There is an experimental drug that, if administered now, would al-
most certainly cure the disease. However, it kills a significant but not accurately
known percentage of those who take it. The expert guesses that the probability of
the drug being fatal is less than 20 percent, and the patient thus assumes that he
is definitely less likely to die if he takes the drug than he is to be paralyzed if he
lets the disease run its course. The patient would regard bedridden life as prefer-
able to death, but he considers both outcomes as totally disastrous compared to
continuing his life in good health. Should he take the drug?2
Since the worst outcome is death, and this outcome will not occur unless
he takes the drug, the maximin rule would tell him not to take the drug. In
contrast, the disaster avoidance rule would tell him to take the drug, because
both death and paralysis are disasters and taking the drug minimizes his
chances that any disaster will occur. Thus, although the disaster avoidance
rule opposes risk taking, it does so in a different way than the maximin rule.
We are left, then, with a plethora of rules: dominance, insufficient reason,
maximax, maximin, and disaster avoidance. Other rules have been proposed
as well. With all of these rules in the offing, it is natural to ask which is cor-
rect. Unfortunately, there is no consensus on this issue. Each rule applies and
seems plausible in some cases but not in others. Many people conclude that
each rule is appropriate to different kinds of situations. It is still not clear,
however, which rule should govern decisions in which circumstances. The
important problem of decision under ignorance remains unsolved.
1. In the game of ignorance, you draw one card from a deck, but you do not
know how many cards or which kinds of cards are in the deck. It might be
a normal deck or it might contain only diamonds or only aces of spades or
any other combination of cards. It costs nothing to play. If you bet that the
card you draw will be a spade, and it is a spade, then you win $100. If you
bet that the card you draw will not be a spade, and it is not a spade, then
you win $90. You may make only one bet. Which bet would you make if
you followed the maximax rule? The maximin rule? The disaster
avoidance rule? The rule of insufficient reason? Which rule seems most
plausible this case? Which bet should you make? Why?
2. In which circumstances do you think it is appropriate to use the
dominance rule? The rule of insufficient reason? The maximax rule? The
maximin rule? The disaster avoidance rule? Why?
3. Suppose that you may choose either of two envelopes. You know that one
envelope contains twice as much money as the other, but you do not know
the amount of money in either envelope. You choose an envelope, open it,
and see that it contains $100. Now you know that the other envelope must
Discussion Questions
(continued)
312
CHAPTER 12 ■ Cho ices
contain either $50 or $200. At this point, you are given a choice: You may
exchange your envelope for the other envelope. Should you switch
envelopes, according to the rule of insufficient reason? Is this result
plausible? Why or why not?
4. The following article raises a problem in dealing with infinity. How would
you answer it?
PLAYING GAMES WITH ETERNITY:
THE DEVIL’S OFFER3
by Edward J. Gracely
Suppose Ms C dies and goes to hell, or to a place that seems like hell. The devil
approaches and offers to play a game of chance. If she wins, she can go to
heaven. If she loses, she will stay in hell forever; there is no second chance to
play the game. If Ms C plays today, she has a 1/2 chance of winning. Tomorrow
the probability will be 2/3. Then 3/4, 4/5, 5/6, etc., with no end to the series.
Thus, every passing day increases her chances of winning. At what point
should she play the game?
The answer is not obvious: after any given number of days spent waiting,
it will still be possible to improve her chances by waiting yet another day. And
any increase in the probability of winning a game with infinite stakes has an
infinite utility. For example, if she waits a year, her probability of winning the
game would be approximately 0.997268; if she waits one more day, the proba-
bility would increase to 0.997275, a difference of only 0.000007. Yet even
0.000007 multiplied by infinity is infinite.
On the other hand, it seems reasonable to suppose the cost of delaying for
a day to be finite—a day’s more suffering in hell. So the infinite expected ben-
efit from a delay will always exceed the cost.
This logic might suggest that Ms C should wait forever, but clearly such a
strategy would be self-defeating: why should she stay forever in a place in or-
der to increase her chances of leaving it? So the question remains: what should
Ms C do?
5. Pascal is famous for invoking infinite utilities to argue for belief in God.
Explain and evaluate his argument in the following passage.
THE WAGER4
by Blaise Pascal
. . . Let us weigh the gain and the loss in wagering that God is. Let us estimate
these two chances. If you gain, you gain all; if you lose, you lose nothing.
313
Dec is ions Under Ignorance
Wager, then, without hesitation that He is.—“That is very fine. Yes, I must
wager; but I may perhaps wager too much.“—Let us see. Since there is an
equal risk of gain and of loss, if you had only to gain two lives, instead of one,
you might still wager. But if there were three lives to gain, you would have to
play (since you are under the necessity of playing), and you would be impru-
dent, when you are forced to play, not to chance your life to gain three at a
game where there is an equal risk of loss and gain. But there is an eternity of
life and happiness. And this being so, if there were an infinity of chances, of
which one only would be for you, you would still be right in wagering one to
win two, and you would act stupidly, being obliged to play, by refusing to
stake one life against three at a game in which out of an infinity of chances
there is one for you, if there were an infinity of an infinitely happy life to gain.
But there is here an infinity of an infinitely happy life to gain, a chance of gain
against a finite number of chances of loss, and what you stake is finite. It is all
divided; wherever the infinite is and there is not an infinity of chances of loss
against that of gain, there is no time to hesitate, you must give all. And thus,
when one is forced to play, he must renounce reason to preserve his life, rather
than risk it for infinite gain, as likely to happen as the loss of nothingness. . . .
The end of this discourse.—Now, what harm will befall you in taking this
side? You will be faithful, honest, humble, grateful, generous, a sincere friend,
truthful. Certainly you will not have those poisonous pleasures, glory and lux-
ury; but will you not have others? I will tell you that you will thereby gain in this
life, and that, at each step you take on this road, you will see so great certainty of
gain, so much nothingness in what you risk, that you will at last recognize that
you have wagered for something certain and infinite, for which you have given
nothing.
6. In the following passage, James Cargile responds to Pascal’s argument. Ex-
plain Cargile’s point. How could Pascal best respond? Is this response ade-
quate? Why or why not?
Either (a) there is a god who will send only religious people to heaven or
(b) there is not. To be religious is to wager for (a). To fail to be religious is to wa-
ger for (b). We can’t settle the question whether (a) or (b) is the case, at least not at
present. But (a) is clearly vastly better than (b). With (a), infinite bliss is guaran-
teed, while with (b) we are still in the miserable human condition of facing death
with no assurance as to what lies beyond. So (a) is clearly the best wager. . . .
This argument just presented is formally similar to the following: Either
(a) there is a god who will send you to heaven only if you commit a painful rit-
ual suicide within an hour of first reading this, or (b) there is not. We cannot
settle the question whether (a) or (b) is the case or it is at least not settled yet.
But (a) is vastly preferable to (b), since in situation (a) infinite bliss is guaran-
teed, while in (b) we are left in the miserable human condition. So we should
wager for (a) by performing ritual suicide.
It might be objected that we can be sure that there is not a god who will
send us to heaven only if we commit suicide, but we can’t be sure that there is
not a god who will send us to heaven only if we are religious. However, a
sceptic would demand proof of this. . . .5
314
CHAPTER 12 ■ Cho ices
NOTES
1 If the lottery gave a consolation prize of a shiny new quarter to all losers, their net loss would
be only seventy-five cents. Since most lotteries do not give consolation prizes, the net loss
equals the cost of playing such lotteries.
2 Gregory Kavka, “Deterrence, Utility, and Rational Choice,” reprinted in Moral Paradoxes of
Nuclear Deterrence (New York: Cambridge University Press, 1987), 65–66. Kavka uses this
medical example to argue for his disaster avoidance rule and, by analogy, to defend the ration-
ality of nuclear deterrence.
3 From Analysis 48 (1988): 113.
4 From Pensées and the Provincial Letters, trans. W. F. Trotter (New York: Modern Library, 1941).
5 From “Pascal’s Wager,” Philosophy 41 (1966): 254.
Fallacies
When inferences are defective, they are called fallacious. When defective styles of
reasoning are repeated over and over, because people often get fooled by them, then
we have an argumentative fallacy that is worth flagging with a name. The number
and variety of argumentative fallacies are limited only by the imagination. Conse-
quently, there is little point in trying to construct a complete list of fallacies. What
is crucial is to get a feel for the most common and most seductive kinds of fallacy.
Once this is done, we should be able to recognize many other kinds as well. The goal
of Part IV is to develop that skill.
IV
315
This page intentionally left blank
FALLACIES OF VAGUENESS
This chapter examines one of the main ways in which arguments can be defective or
fallacious because language is not used clearly enough for the context. This kind of
unclarity is vagueness. Vagueness occurs when, in a given context, a term is used in
a way that allows too many cases in which it is unclear whether or not the term applies.
Vagueness underlies several common fallacies, including three kinds of slippery-slope
arguments.
USES OF UNCLARITY
In a good argument, a person states a conclusion clearly and then, with equal
clarity, gives reasons for this conclusion. The arguments of everyday life of-
ten fall short of this standard. Usually, unclear language is a sign of unclear
thought. There are times, however, when people are intentionally unclear.
They might use unclarity for poetic effect or to leave details to be decided
later. But often their goal is to confuse others. This is called obfuscation.
Before we look at the various ways in which language can be unclear, a
word of caution is needed: There is no such thing as absolute clarity.
Whether something is clear or not depends on the context in which it occurs.
A botanist does not use common vocabulary in describing and classifying
plants. At the same time, it would usually be foolish for a person to use
botanical terms in describing the appearance of her backyard. Aristotle said
that it is the mark of an educated person not to expect more rigor than the
subject matter will allow. Because clarity and rigor depend on context, it
takes judgment and good sense to pitch an argument at the right level.
13
317
Image not available due to copyright restrictions
318
CHAPTER 13 ■ Fallac ies of Vagueness
VAGUENESS
Perhaps the most common form of unclarity is vagueness. It arises when a
concept applies along a continuum or a series of very small changes. The
standard example is baldness. A person with a full head of hair is not bald.
A person without a hair on his head is bald. In between, however, is a range
of cases in which we cannot say definitely whether the person is bald or not.
These are called borderline cases. Here we say something less definite, such
as that this person is “going bald.”
Our inability to apply the concept of baldness in a borderline case is not
due to ignorance of the number of hairs on the person’s head. It will not help
to count the number of hairs there. Even if we knew the exact number, we
would still not be able to say whether the person was bald or not. The same
is true of most adjectives that concern properties admitting of degrees—for
example, “rich,” “healthy,” “tall,” “wise,” and “ruthless.”
For the most part, imprecision—the lack of sharply defined limits—
causes little difficulty. In fact, this is a useful feature of our language, for
suppose we did have to count the number of grains of salt between our fin-
gers to determine whether or not we hold a pinch of salt. It would take a long
time to follow a simple recipe that calls for a pinch of salt.
Yet difficulties can arise when borderline cases themselves are at issue.
Suppose that a state passes a law forbidding all actions that offend a large
number of people. There will be many cases that clearly fall under this law
and many cases that clearly do not fall under it. There will also be many
cases in which it will not be clear whether or not they fall under this law.
Laws are sometimes declared unconstitutional for this very reason. Here we
shall say that the law is vague. In calling the law vague, we are criticizing it.
We are not simply noticing the existence of borderline cases, for there will
usually be borderline cases no matter how careful we are. Instead, we are
saying that there are too many borderline cases for this context. More pre-
cisely, we shall say that an expression in a given context is used vaguely if it
leaves open too wide a range of borderline cases for the successful and legit-
imate use of that expression in that context.
Vagueness thus depends on context. To further illustrate this context
dependence, consider the expression “light football player.” There are, of
course, borderline cases between those football players who are light and
those who are not light. But on these grounds alone, we would not say that the
expression is vague. It is usually a perfectly serviceable expression, and we
can indicate borderline cases by saying such things as “Jones is a bit light for a
football player.” Suppose, however, that Ohio State and Cal Tech wish to have
a game between their light football players. It is obvious that the previous un-
derstanding of what counts as being light is too vague for this new context. At
Ohio State, anyone under 210 pounds is considered light. At Cal Tech, anyone
over 150 pounds is considered heavy. What is needed, then, is a ruling, such
as that anyone under 175 pounds will be considered a lightweight. This exam-
ple illustrates a common problem and its solution. A term that works perfectly
319
V A G U E N E S S
well in one area becomes vague when applied in some other (usually more
specialized) area. This vagueness can then be removed by adopting more
precise rules in the problematic area. Vagueness is resolved by definition.
For each of the following terms, give one case to which the term clearly ap-
plies, one case to which the term clearly does not apply, and one borderline
case. Then try to explain why the borderline case is a borderline case.
Example: In the northern hemisphere, “summer month” clearly applies to
July; clearly does not apply to January; and June is a borderline case, because
the summer solstice is June 21, and schools usually continue into June, but
June, July, and August are, nonetheless, often described as the summer
months.
1. large animal
2. populous state
3. long book
4. old professor
5. popular singer
6. powerful person
7. difficult subject
8. late meeting
9. arriving late to a meeting
EXERCISE I
Each of the following sentences contains words or expressions that are poten-
tially vague. Describe a context in which this vagueness might make a difference,
and explain what difference it makes. Then reduce this vagueness by replacing
the italicized expression with one that is more precise.
Example: Harold has a bad reputation.
Context: If Harold applies for a job as a bank security guard, then some
but not all kinds of bad reputation are relevant. A reputation for doing bad
construction work is irrelevant, but a reputation for dishonesty is relevant.
Replacement: Harold is a known thief.
1. Ross has a large income.
2. Cocaine is a dangerous drug.
3. Ruth is a clever woman.
4. Andre is a terrific tennis player.
5. Mark is not doing too well (after his operation).
EXERCISE II
(continued)
320
CHAPTER 13 ■ Fallac ies of Vagueness
HEAPS
The existence of borderline cases is essential to various styles of reasoning
that have been identified and used since ancient times. One such argument
was called the argument from the heap or the sorites argument (from the Greek
word “soros,” which means “heap”). The classic example was intended to
show that it is impossible to produce a heap of sand by adding one grain at
a time. As a variation on this, we will show that no one can become rich. The
argument can be formulated as a long series like this:
(1) Someone with only one cent is not rich.
(2) If someone with only one cent is not rich, then someone with
only two cents is not rich.
(3) Someone with only two cents is not rich.
(4) If someone with only two cents is not rich, then someone with only
three cents is not rich.
(5) Someone with only three cents is not rich.
(6) If someone with only three cents is not rich, then someone with
only four cents is not rich.
(7) Someone with only four cents is not rich.
[and so on, until:]
(199,999,999,999) Someone with only 100,000,000,000 cents is not rich.
The problem, of course, is that someone with 100,000,000,000 cents is rich. If
someone denies this, we can keep on going. Or we can just sum up the
whole argument like this:
(1*) Someone with only one cent is not rich.
(2*) For any number, n, if someone with only n cents is not rich, then
someone with n + 1 cents is not rich.
(3*) Someone with any number of cents is not rich.
Premise (2*) is, of course, just a generalization of premises (2), (4), (6), and so on.
Despite its plausibility, everyone should agree that there is something
wrong with these arguments. If we hand over enough pennies to Peter,
6. Shaq’s a big fellow.
7. Dan’s grades are low.
8. Walter can’t see well.
9. The earthquake was a disaster.
10. The news was wonderful.
321
h E A P S
previously poor Peter will become the richest person in the world. Another
sign of a problem is that a parallel argument runs in the other direction:
Someone with 100 billion cents is rich. For any number, n, if someone with n
cents is rich, then someone with n – 1 cents is also rich. Therefore, someone
with no cents at all is rich. This is absurd (since we are not talking about how
rich one’s life can be as long as one has friends).
We can see that these arguments turn on borderline cases in the following
way: The argument would fail if we removed borderline cases by laying
down a ruling (maybe for tax purposes) that anyone with a million dollars
or more is rich and anyone with less than this is not rich. A person with
$999,999.99 would then pass from not being rich to being rich when given a
single penny, so premise (2*) would be false at that point under this ruling.
Of course, we do not usually use the word “rich” with this much precision.
We see some people as clearly rich and others as clearly not rich, but in be-
tween there is a fuzzy area where we are not prepared to say that people ei-
ther are or are not rich. In this fuzzy area, as well as in the clear areas, a
penny one way or the other will make no difference.
That is how the argument works, but exactly where does it go wrong?
This question is not easy to answer and remains a subject of vigorous de-
bate. Here is one way to view the problem: Consider a person who is
80 pounds overweight, where we would all agree that that person would
pass from being fat to not being fat by losing over 100 pounds. If he or she
lost an ounce a day for five years, this would be equivalent to losing just
over 114 pounds. An argument from the heap denies that this person would
ever cease to be fat. (So what is the point of dieting?) Anyone who accepted
that conclusion, or (3*), would seem to claim that a series of insignificant
changes cannot be equivalent to a significant change. Surely, this is wrong.
Here we might be met with the reply that every change must occur at some
particular time (and place), but there would be no particular day on which
this person would pass from being fat to not being fat. The problem with
this reply is that, with concepts like this, changes seem to occur gradually
over long stretches of time without occurring at any single moment. Any-
way, however or whenever it occurs, the change does occur. Some people
do cease to be fat if they lose enough weight.
This tells us that conclusions of arguments from the heap, such as (3*), are
false, so these arguments cannot be sound. Almost everyone agrees to that
much. Moreover, if an appropriate starting point is chosen, then premises
like (1) and (1*) will also be accepted as true by almost everyone. So the
main debate focuses on premise (2*) and on whether the argument is valid.
Some philosophers reject premise (2*) and claim that there is a precise point
at which a person becomes rich, even though we don’t know where that
point is. Others try to avoid any sharp cutoff point by developing some kind
of alternative logic. Still others just admit that the premises seem true, and
the argument seems valid, but the conclusion seems false, so the argument
creates a paradox to which they have no solution. These views become com-
plicated and technical, so we will not discuss them here. Suffice it to say that
almost everyone agrees that conclusions like (3*) and (199,999,999,999) are
322
CHAPTER 13 ■ Fallac ies of Vagueness
false, so arguments from the heap are unsound for one reason or another.
That is why such arguments are labeled fallacies.
Paris Hilton
Nonliving Living
Salt Crystal Virus
We all agree that a salt crystal is not alive. Yet a salt crystal is very similar to
other more complex crystals, and these crystals are similar to certain viruses.
Where exactly do you think arguments from the heap go astray?
DISCUSSION QUESTION
SLIPPERY SLOPES
Near cousins to arguments from the heap are slippery-slope arguments, but
they reach different conclusions. Whereas heap arguments conclude that
nothing has a certain property, such as baldness, a slippery-slope argu-
ment could be trotted out to try to show that there is no real or defensible
or significant or important difference between being bald and not being
bald. The claim is not that no change occurs because the person who loses
all his hair is still not bald, as in an argument from the heap. Instead,
the slippery-slope argument claims that we should not classify people as
either bald or not bald, because there is no significant difference between
these classifications.
Whether a difference is significant depends on a variety of factors. In
particular, what is significant for one purpose might not be significant for
other purposes. Different concerns then yield different kinds of slippery-
slope arguments. We will discuss three kinds, beginning with conceptual
slippery-slope arguments.
CONCEPTUAL SLIPPERY-SLOPE ARGUMENTS
Conceptual slippery-slope arguments try to show that things at opposite
ends of a continuum do not differ in any way that would be important
enough to justify drawing a distinction in one’s concepts or theories. As an
example, consider the difference between living and nonliving things.
323
S L I P P E R Y S L O P E S
Still, a virus is on the borderline between living and nonliving things. A
virus does not take nourishment and does not reproduce itself. Instead, a
virus invades the reproductive mechanisms of cells, and these cells then pro-
duce the virus. As viruses become more complex, the differences between
them and “higher” life forms become less obvious. Through a series of such
small transitions, we finally reach a creature who is obviously alive: Paris
Hilton. So far, we have merely described a series of gradual transitions along
a continuum. We get a conceptual slippery-slope argument when we draw
the following conclusion from these facts: There is no significant difference
between living things and nonliving things.
To avoid this conclusion, we need to figure out where the argument goes
wrong. Such arguments often seem to depend on the following principles:
1. We should not draw a distinction between things that are not
significantly different.
2. If A is not significantly different from B, and B is not significantly
different from C, then A is not significantly different from C.
This first principle is interesting, complicated, and at least generally true. We
shall examine it more closely in a moment. The second principle is obviously
false. As already noted, a series of insignificant differences can add up to a
significant difference. Senator Everett Dirksen put the point well when he
said, “A billion dollars here and a billion dollars there can add up to some
real money.” To the extent that conceptual slippery-slope arguments depend
on this questionable assumption, they provide no more support for their
conclusions than do arguments from the heap.
Unlike arguments from the heap, however, conceptual slippery-slope
arguments do often lead people to accept their conclusions. Slippery-slope
arguments have been used to deny the difference between sanity and insan-
ity (some people are just a little weirder than others) and between amateur
and professional athletics (some athletes just get paid a little more or more
directly than other athletes). When many small differences make a big
difference, such conceptual slippery-slope arguments are fallacious.
This fallacy is seductive, because it is often hard to tell when many small
differences do make a big difference. Here is a recent controversial example:
Some humans have very dark skin. Others have very pale skin. As members
of these different groups marry, their children’s skin can have any interme-
diate shade of color. This smooth spectrum leads some people to deny that
any differences among races will be important to developed theories in biol-
ogy. Their argument seems to be that the wealth of intermediate cases will
make it difficult or impossible to formulate precise and exceptionless laws
that apply to one racial group but not to others, so differences among races
will play no important role in sciences that seek such laws. Critics respond
that some scientific laws about races still might hold without exception even
if skin color and other features do vary in tiny increments.
Whichever side one takes, this controversy shows that, even if there is a
smooth spectrum between endpoints, this continuity is not enough by itself
324
CHAPTER 13 ■ Fallac ies of Vagueness
Whenever we find one thing passing over into its opposite through a gradual
series of borderline cases, we can construct (a) an argument from the heap and
(b) a conceptual slippery-slope argument by using the following method: Find
some increase that will not be large enough to carry us outside the borderline
area. Then use the patterns of argument given above. Applying this method,
formulate arguments for the following claims. Then explain what is wrong
with these arguments.
1. a. There are no heaps.
b. There is no difference between a heap and a single grain of sand.
2. a. Nobody is tall.
b. There is no difference between being tall and being short.
3. a. Books do not exist.
b. There is no difference between a book and a pamphlet.
4. a. Heat is not real.
b. There is no difference between being hot and being cold.
5. a. Taxes are never high.
b. There is no difference between high taxes and low taxes.
6. a. Science is an illusion.
b. There is no difference between science and faith.
EXERCISE III
1. Do you think that differences among races have any role in developed
theories in biology or sociology or any other science? Why or why not?
2. If animals evolve gradually from one species to another, does that show
that there is no significant difference in biology between any species
(say, horses and dogs)? Why or why not? Does it show that there is no
important difference in moral theory between the rights of humans and
the rights of animals in other species? Why or why not?
DISCUSSION QUESTIONS
to show that there are no scientifically significant differences among races.
That conclusion would need to be supported by more than just a conceptual
slippery-slope argument. To show that certain concepts are useless for the
purposes of a certain theory, one would need to add more information, par-
ticularly about the purposes of that theory and its laws. That is what deter-
mines which differences are important in that particular area. Conceptual
slippery-slope arguments might work in conjunction with such additional
premises, but they cannot work alone.
325
S L I P P E R Y S L O P E S
FAIRNESS SLIPPERY-SLOPE ARGUMENTS
When borderline cases form a continuum, if someone classifies a case at one
end of the continuum, an opponent often challenges this classification by
asking, “Where do you draw the line?” Sometimes this challenge is out of
place. If I claim that Babe Ruth was a superstar, I will not be refuted if I can-
not draw a sharp dividing line between athletes who are superstars and
those who are not. There are some difficult borderline cases, but Babe Ruth
is not one of them. Nor will we be impressed if someone tells us that the dif-
ference between Babe Ruth and the thousands of players who never made it
to the major leagues is “just a matter of degree.” What is usually wrong with
this phrase is the emphasis on the word “just,” which suggests that dif-
ferences of degree do not count. Of course, it is a matter of degree, but the
difference in degree is so great that it should be marked by a special word.
There are other occasions when a challenge to drawing a line is appropri-
ate. For example, most schools and universities have grading systems that
draw a fundamental distinction between passing grades and failing grades.
Of course, a person who barely passes a course does not perform very dif-
ferently from one who barely fails a course, yet they are treated very differ-
ently. Students who barely pass a course get credit for it; those who barely
fail it do not. This, in turn, can lead to serious consequences in an academic
career and even beyond. It is entirely reasonable to ask for a justification of a
procedure that treats cases that are so similar in such strikingly different
ways. We are not being tender-hearted; we are raising an issue of fairness or
justice. It seems unfair to treat very similar cases in strikingly different ways.
The point is not that there is no difference between passing and failing.
That is why this argument is not a conceptual slippery-slope argument. The
claim, instead, is that the differences that do exist (as little as one point out of
a hundred on a test) do not make it fair to treat people so differently (credit
versus no credit for the course). This unfairness does not follow merely from
the scores forming a continuum, but the continuum does put pressure on us
to show why small differences in scores do justify big differences in treatment.
Questions about the fairness of drawing a line often arise in the law. For
example, given reasonable cause, the police generally do not have to obtain
a warrant to search a motor vehicle, for the obvious reason that the vehicle
might be driven away while the police go to a judge to obtain a warrant. On
the other hand, with few exceptions, the police may not search a person’s
home without a search warrant. In the case of California v. Carney,1 the U.S.
Supreme Court had to rule on whether the police needed a warrant to search
for marijuana in an “oversized van, fully mobile,” parked in a downtown
parking lot in San Diego. Because the van was a fully mobile vehicle, it seemed
to fall under the first principle; but because it also served as its owner’s
home, it seemed to fall under the second. The difficulty, as the Court saw,
was that there is a gray area between those things that clearly are motor vehi-
cles and not homes (for example, motorcycles) and those things that clearly
are homes and not motor vehicles (for example, apartments). Chief Justice
326
CHAPTER 13 ■ Fallac ies of Vagueness
Warren Burger wondered about a mobile home in a trailer park hooked up
to utility lines with its wheels removed. Justice Sandra Day O’Connor asked
whether a tent, because it is highly mobile, could also be searched without a
warrant. As the discussion continued, houseboats (with or without motors
or oars), covered wagons, and finally a house being moved from one place to
another on a trailer truck came under examination. In the end, our highest
court decided that the van in question was a vehicle and could be searched
without first obtaining a warrant to do so. The court did not fully explain
why it is fair to allow warrantless searches—and to send people to jail as a
result—in cases of vans used as homes but not in other very similar cases.
Questions about where to draw a line often have even more important im-
plications than in the case just examined. Consider the death penalty. Most
societies have reserved the death penalty for those crimes they consider the
most serious. But where should we draw the line between crimes punishable
by death and crimes not punishable by death? Should the death penalty be
given to murderers of prison guards? To rapists? To drug dealers? To drunk
drivers who cause death? Wherever we draw the line, it seems to be an un-
avoidable consequence of the death penalty that similar cases will be treated in
radically different ways. A defender of the death penalty can argue that it is not
unfair to draw a line because, once the line is drawn, the public will have fair
warning about which crimes are subject to the death penalty and which are
not. It will then be up to each person to decide whether to risk his or her life by
crossing this line. It remains a matter of debate, however, whether the law can
be administered in a predictable way that makes this argument plausible.
The finality of death raises a profoundly difficult problem in another area,
too: the legalization of abortion. There are some people who think abortion
is never justified and ought to be declared totally illegal. There are others
who think abortion does not need any justification at all and should be com-
pletely legalized. Between these extremes, there are many people who be-
lieve abortion is justified in certain circumstances but not in others (such as
when abortion is the only way to save the life of the mother but not when it
prevents only lesser harms to the mother). There are also those who think
abortion should be allowed for a certain number of months of pregnancy,
but not thereafter. People holding these middle positions face the problem
of deciding where to draw a line, and this makes them subject to criticisms
from holders of either extreme position.
This problem admits of no easy solution. Because every line we draw will
seem arbitrary to some extent, a person who holds a middle position needs
to argue that it is better to draw some line—even a somewhat arbitrary one—
than to draw no line at all. The recognition that some line is needed, and
why, can often help us locate the real issues. This is the first step toward a
reasonable position.
Of course, this still does not tell us where to draw the line. A separate argu-
ment is needed to show that the line should be drawn at one point, or in one
area, rather than another. In the law, such arguments often appeal to value judg-
ments about the effects of drawing the line at one place rather than another.
327
S L I P P E R Y S L O P E S
For example, it is more efficient to draw a line where it is easy to detect, and
drawing the line at one place will provide greater protection for some values
or some people than will drawing it at another place. Different values often
favor drawing different lines, and sometimes such arguments are not avail-
able at all. Thus, in the end, it will be difficult to solve many of these pro-
found and important problems.
Is it unfair for teachers to fail students who get one point out of a hundred less
than others students who pass? Why or why not? Would an alternative grading
system be fairer?
DISCUSSION QUESTION
CAUSAL SLIPPERY-SLOPE ARGUMENTS
Another common kind of argument is also often described as a slippery-
slope argument. In these arguments, the claim is made that, once a certain
kind of event occurs, other similar events will also occur, and this will lead
Image not available due to copyright restrictions
328
CHAPTER 13 ■ Fallac ies of Vagueness
eventually to disaster. The most famous (or infamous) argument of this kind
was used by the U.S. government to justify its intervention in Vietnam in the
1960s. It was claimed that, if the communists took over Vietnam, they would
then take over Cambodia, the rest of Asia, and other continents, until they
ruled the whole world. This was called the domino theory, since the fall of
one country would make neighboring countries fall as well. Arguments of
this kind are sometimes called domino arguments. Such arguments claim that
one event, which might not seem bad by itself, would lead to other, more hor-
rible events, so such arguments can also be called parades of horrors.
Causal slippery slopes can also slide into good results. After all, someone
who wants communists to take over the world might use the above domino ar-
gument to show why the United States should not intervene in Vietnam. Such
optimistic slippery-slope arguments are, however, much less common than pa-
rades of horrors, so we will limit our discussion to the pessimistic versions.
These arguments resemble other slippery-slope arguments in that they de-
pend on a series of small changes. The domino argument does not, however,
claim that there is no difference between the first step and later steps—be-
tween Vietnam going communist and the rest of Asia going communist. Nor
is there supposed to be anything unfair about letting Vietnam go communist
without letting other countries also go communist. The point of a parade of
horrors is that certain events will cause horrible effects because of their simi-
larity or proximity to other events. Since the crucial claim is about causes and
effects, these arguments will be called causal slippery-slope arguments.
We saw another example in Chapter 4. While arguing against an increase
in the clerk hire allowance, Kyl says,
The amount of increase does not appear large. I trust, however, there is no one
among us who would suggest that the addition of a clerk would not entail
allowances for another desk, another typewriter, more materials, and it is not
beyond the realm of possibility that the next step would then be a request for
additional office space, and ultimately new buildings.
Although this argument is heavily guarded, the basic claim is that increasing
the clerk hire allowance is likely to lead to much larger expenditures that will
break the budget. The argument can be represented more formally this way:
(1) If the clerk hire allowance is increased, other expenditures will also
probably be increased.
(2) These other increases would be horrible.
(3) The clerk hire allowance should not be increased.
Opponents can respond in several ways. One response is to deny that the
supposedly horrible effects really are so horrible. One might argue, for exam-
ple, that additional office space and new buildings would be useful. This re-
sponse is often foreclosed by describing the effects in especially horrible terms.
A second possible response would be to deny that increasing the clerk
hire allowance really would have the horrible effects that are claimed in the
first premise. One might argue, for example, that the old offices already
have adequate room for additional clerks.
329
S L I P P E R Y S L O P E S
Often the best response is a combination of these. One can admit that certain
claimed effects would be horrible, but deny that these horrible effects really are
likely. Then one can acknowledge that some more minor problems will ensue,
but argue that these costs are outweighed by the benefits of the program.
To determine which, if any, of these responses is adequate, one must look
closely at each particular argument and ask the following questions:
Are any of the claimed effects really very bad?
Are any of these effects really very likely?
Do these dangers outweigh all the benefits of what is being criticized?
If the answers to all these questions are “Yes,” then the causal slippery-slope
argument is strong. But if any of these questions receives a negative answer,
then the causal slippery-slope argument is questionable on that basis.
Classify each of the following arguments as either (H) an argument from the
heap, (C) a conceptual slippery-slope argument, (F) a fairness slippery-slope
argument, or (S) a causal slippery-slope argument. Explain why you classify
each example as you do.
1. We have to take a stand against sex education in junior high schools. If we
allow sex education in the eighth grade, then the seventh graders will
want it, and then the sixth graders, and pretty soon we will be teaching
sex education to our little kindergartners.
2. People are found not guilty by reason of insanity when they cannot avoid
breaking the law. But people who are brought up in certain deprived
social circumstances are not much more able than the insane to avoid
breaking the law. So it would be unjust to find them guilty.
3. People are called mentally ill when they do very strange things, but many
so-called eccentrics do things that are just as strange. So there is no real
difference between insanity and eccentricity.
4. If you try to smoke one cigarette a day, you will end up smoking two and
then three and four and five, and so on, until you smoke two packs every
day. So don’t try even one.
5. A human egg one minute after fertilization is not very different from what
it is one minute later, or one minute after that, and so on. Thus, there is re-
ally no difference between just-fertilized eggs and adult humans.
6. Since no moment in the continuum of development between an egg and a
baby is especially significant, it is not fair to grant a right to life to a baby
unless one grants the same right to every fertilized egg.
7. If we let doctors kill dying patients who are in great pain, then they will
kill other patients who are in less pain and patients who are only slightly
disabled. Eventually, they will kill anyone who is not wanted by society.
EXERCISE IV
330
CHAPTER 13 ■ Fallac ies of Vagueness
Explain the reasons, if any, for drawing a definite line in each of the following
cases. Then further explain how this line can be drawn, if at all, in a reasonable
way.
1. Minimum (or maximum?) age to drive a car
2. Minimum age to vote
3. Minimum age to enter (or be drafted into) the military
4. Minimum age to drink alcoholic beverages
5. Minimum age for election to the U.S. presidency
6. Maximum age before retirement becomes mandatory
EXERCISE V
Determine whether each of the following arguments provides adequate
support, or any support, for its conclusion. Explain why.
1. I shouldn’t get a speeding ticket for going fifty-six miles per hour, because
my driving did not all of a sudden get more dangerous when I passed the
speed limit of fifty-five.
2. No student should ever be allowed to ask a question during a lecture,
because once one student asks a question, then another one wants to ask a
question, and pretty soon the teacher doesn’t have any time left to lecture.
3. Pornography shouldn’t be illegal, because you can’t draw a line between
pornography and erotic art.
4. Marijuana should be legal, because it is no more dangerous than alcohol
or nicotine.
5. Marijuana should be illegal, because people who try marijuana are likely
to go on to try hashish, and then snorting cocaine, and then freebasing
cocaine or shooting heroin.
6. The government should not put any new restrictions on free trade,
because once they place some restrictions, they will place more and more
until foreign trade is so limited that our own economy will suffer.
7. Governments should never bargain with any terrorist. Once they do, they
will have to bargain with every other terrorist who comes along.
8. If assault weapons are banned, Congress will ban handguns next, and
then rifles. Eventually, hunters will not be able to hunt, and law-abiding
citizens will have no way to defend themselves against criminals.
EXERCISE VI
331
Sl ippery Slopes
1. Explain and evaluate the following argument against restrictions on hate
speech:
To attempt to craft free speech exceptions only for racist speech would create a sig-
nificant risk of a slide down the proverbial “slippery slope.” . . . Censorial conse-
quences could result from many proposed or adopted university policies, including
the Stanford code, which sanctions speech intended to “insult or stigmatize” on the
basis of race or other prohibited grounds. For example, certain feminists suggest that
all heterosexual sex is rape because heterosexual men are aggressors who operate in
a cultural climate of pervasive sexism and violence against women. Aren’t these fem-
inists insulting or stigmatizing heterosexual men on the basis of their sex and sexual
orientation? And how about a Holocaust survivor who blames all (“Aryan”) Ger-
mans for their collaboration during World War II? Doesn’t this insinuation insult and
stigmatize on the basis of national and ethnic origin? And surely we can think of nu-
merous other examples that would have to give us pause.2
2. Explain and evaluate the following response to critics of college
restrictions on hate speech:
[Defenders of such restrictions] will ask whether an educational institution does not
have the power . . . to enact reasonable regulations aimed at assuring equal person-
hood on campus. If one characterizes the issue this way, . . . a different set of slopes
will look slippery. If we do not intervene to protect equality here, what will the next
outrage be?3
3. When John Stewart interviewed William Bennett (former Secretary of
Education under President Ronald Reagan) about gay marriage, both of
them used slippery slopes and responded to each other’s slippery slopes
in the following exchange. What kinds of slippery slopes did they use?
Was either argument better than the other? Was either response better than
the other? Why or why not?
BENNETT: The question is: How do you define marriage? Where do you draw the
line? What do you say to the polygamist?
STEWART: You don’t say anything to the polygamist. That is a choice, to get three
or four wives. That is not a biological condition that “I gots to get laid by differ-
ent women that I am married to.” That’s a choice. Being gay is part of the human
condition. There’s a huge difference.
BENNETT: Well, some people regard their human condition as having three
women. Look, the polygamists are all over this.
STEWART: Then let’s go slippery slope the other way. If government says I can
define marriage as between a man and a woman, what says they can’t define it
between people of different income levels, or they can decide whether or not you
are a suitable husband for a particular woman?
BENNETT: Because gender matters in marriage, it has mattered to every human
society, it matters in every religion . . .
STEWART: Race matters in every society as well. Isn’t progress understanding?
4. What, if anything, is shown when slippery-slope arguments can be used
on both sides of an issue?
DISCUSSION QUESTIONS
332
CHAPTER 13 ■ Fallac ies of Vagueness
NOTES
1 1471 U.S. 386 (1984). This case was reported by Linda Greenhouse, “Of Tents with Wheels and
Houses with Oars,” New York Times, May 15, 1985.
2 Nadine Strossen, “Regulating Racist Speech on Campus: A Modest Proposal?” Duke Law Jour-
nal (1990): 537–38. When she wrote this, Strossen was on the National Board of Directors of the
American Civil Liberties Union.
3 Richard Delgado, “Campus Antiracism Rules: Constitutional Narratives in Collision,” North-
western University Law Review 85 (1991): 346.
Fallacies OF Ambiguity
This chapter examines fallacies that arise from a second kind of unclarity: ambiguity.
Ambiguity occurs when it is unclear which meaning of a term is intended in a given
context. Ambiguity leads to the fallacy of equivocation, which will be defined and
illustrated. The chapter closes with a discussion of different kinds of definitions that
can be useful in avoiding or responding to fallacies of clarity.
AMBIGUITY
The idea of vagueness is based on a common feature of words in our language:
Many of them leave open a range of borderline cases. The notion of ambiguity
is also based on a common feature of our language: Words often have a num-
ber of different meanings. For example, the New Merriam-Webster Pocket Dictio-
nary has the following entry under the word “cardinal”:
cardinal adj. 1: of basic importance; chief, main, primary,
2: of cardinal red color.
n. 1: an ecclesiastical official of the Roman Catholic Church
ranking next below the pope,
2: a bright red,
3: any of several American finches of which the male is
bright red.
In the plural, “the Cardinals” is the name of an athletic team that inhabits
St. Louis; “cardinal” also describes the numbers used in simple counting.
It is not likely that people would get confused about these very different
meanings of the word “cardinal,” but we might imagine a priest, a bird-
watcher, and a baseball fan all hearing the remark “The cardinals are in
town.” The priest would prepare for a solemn occasion, the bird-watcher
would get out binoculars, and the baseball fan would head for the stadium.
In this context, the remark might be criticized as ambiguous. More precisely,
we shall say that an expression in a given context is used ambiguously if and
only if it is misleading or potentially misleading because it is hard to tell
which of a number of possible meanings is intended in that context.
14
333
334
CHAPTER 14 ■ Fallac ies of Amb igu ity
Using this definition, the word “bank” is not used ambiguously in the fol-
lowing sentence:
Joan deposited $500 in the bank and got a receipt.
Some writers, however, call an expression ambiguous simply if it admits of
more than one interpretation, without adding that it is not possible to tell
which meaning is intended. With this definition, the above sentence is am-
biguous because it could mean that Joan placed $500 in a riverbank, and
someone, for whatever reason, gave her a receipt for doing so. On this second
definition of ambiguity, virtually every expression is ambiguous, because vir-
tually every expression admits of more than one interpretation. On our first
definition, only uses of expressions that are misleading or potentially mislead-
ing will be called ambiguous. In what follows, we will use the word “ambigu-
ous” in accordance with the first definition. Ambiguity then depends on the
context, because whether something is misleading also depends on context.
In everyday life, context usually settles which of a variety of meanings is
appropriate. Yet sometimes genuine misunderstandings do arise. An Ameri-
can and a European discussing “football” may have different games in mind.
The European is talking about what Americans call “soccer”; the American is
talking about what Europeans call “American football.” It is characteristic of
the ambiguous use of a term that when it comes to light, we are likely to say
something like, “Oh, you mean that kind of cardinal!” or “Oh, you were talk-
ing about American football!” This kind of misunderstanding can cause trou-
ble. When it does, if we want to criticize the expression that creates the
problem, we call it ambiguous.
Thus, “ambiguous” is both dependent on context and a term of criticism
in much the same ways as “vague.” But these kinds of unclarity differ in
other ways. In a context where the use of a word is ambiguous, it is not clear
which of two meanings to attach to a word. In a context where the use of a
word is vague, we cannot attach any precise meaning to the use of a word.
So far we have talked about the ambiguity of individual terms or words.
This is called semantic ambiguity. But sometimes we do not know which in-
terpretation to give to a phrase or a sentence because its grammar or syntax
admits of more than one interpretation. This is called syntactic ambiguity or
amphiboly. Thus, if we talk about the conquest of the Persians, we might be refer-
ring either to the Persians’ conquering someone or to someone’s conquering
Image not available due to copyright restrictions
335
Amb igu ity
the Persians. Sometimes the grammar of a sentence leaves open a great
many possible interpretations. For example, consider the following sentence
(from Paul Benacerraf):
Only sons marry only daughters.
One thing this might mean is that a person who is a male only child will
marry a person who is a female only child. Again, it might mean that sons
are the only persons who marry daughters and do not marry anyone else.
Other interpretations are possible as well.
The process of rewriting a sentence so that one of its possible meanings be-
comes clear is called disambiguating the sentence. One way of disambiguating a
sentence is to rewrite it as a whole, spelling things out in detail. That is how we
disambiguated the sentence “Only sons marry only daughters.” Another proce-
dure is to continue the sentence in a way that supplies a context that forces one
interpretation over others. Consider the sentence “Mary had a little lamb.” No-
tice how the meaning changes completely under the following continuations:
1. Mary had a little lamb; it followed her to school.
2. Mary had a little lamb and then some broccoli.
Just in passing, it is not altogether obvious how we should describe the am-
biguity in the sentence “Mary had a little lamb.” The most obvious sugges-
tion is that the word “had” is ambiguous, meaning “owned” on the first
reading and “ate” on the second reading. Notice, however, that this also
forces alternative readings for the expression “a little lamb.” Presumably, it
was a small, whole, live lamb that followed Mary to school, whereas it
would have been a small amount of cooked lamb that she ate. So if we try to
locate the ambiguity in particular words, we must say that not only the
word “had” but also the word “lamb” are being used ambiguously. This is a
reasonable approach, but another is available. In everyday speech, we often
leave things out. Thus, instead of saying “Mary had a little portion of meat de-
rived from a lamb to eat,” we simply say “Mary had a little lamb,” dropping
out the italicized words on the assumption that they will be understood. In
most contexts, such deletions cause no misunderstanding. But sometimes
deletions are misunderstood, and this can produce ambiguity.
Show that each of the following sentences admits of at least two interpretations
by (1) rewriting the sentence as a whole in two different ways and (2) expanding
the sentence in two different ways to clarify the context:
Example: Kenneth let us down.
Rewriting: Kenneth lowered us.
Kenneth disappointed us.
Expanding: Kenneth let us down with a rope.
Kenneth let us down just when we needed him.
EXERCISE I
(continued)
336
CHAPTER 14 ■ Fallac ies of Amb igu ity
1. Barry Bonds (the baseball player) was safe at home.
2. I don’t know what state Meredith is in.
3. Where did you get bitten?
4. The president sent her congratulations.
5. Visiting professors can be boring.
6. Wendy ran a marathon.
7. The meaning of the term “altering” is changing.
8. I don’t want to get too close to him.
9. I often have my friends for dinner.
10. Slow Children Playing. (on a street sign.)
11. Save Soap and Waste Paper. (on a sign during World War II.)
12. In his will, he left $1,000 to his two sons, Jim and John.
13. There is some explanation for everything.
14. She is an Asian historian.
15. Nobody may be in the lounge this evening.
16. Nobody came to the concert at 8 PM.
Follow the same instructions for the following actual newspaper headlines, many
of which come from Columbia Journalism Review, editors, Squad Helps Dog Bite
Victim and Other Flubs from the Nation’s Press (Garden City, NY: Doubleday, 1980).
1. Milk Drinkers Turn to Powder
2. Anti-busing Rider Killed by Senate
3. Mrs. Gandhi Stoned in Rally in India
4. College Graduates Blind Senior Citizen
5. Jumping Bean Prices Affect the Poor
6. Tuna Biting off Washington Coast
7. Time for Football and Meatball Stew
8. Police Kill Man with Ax
9. Squad Helps Dog Bite Victim
10. Child Teaching Expert to Speak
11. Prostitutes Appeal to Pope
12. Legalized Outhouses Aired by Legislature
13. Police Can’t Stop Gambling
14. Judge Permits Club to Continue Sex Bar
15. Greeks Fine Hookers
16. Survivor of Siamese Twins Joins Parents
EXERCISE II
337
Equ ivocat ion
EQUIVOCATION
Ambiguity can cause a variety of problems for arguments. Often it produces
hilarious or embarrassing side effects, and it is hard to get your arguments
taken seriously if your listeners are giggling over an unintended double en-
tendre in which one of the double meanings has risqué connotations.
Ambiguity can also generate bad arguments that involve the fallacy of
equivocation. An argument is said to commit this fallacy when it uses the
same expression in different senses in different parts of the argument, and
this ruins the argument. Here is a silly example (from Carl Wolf):
Six is an odd number of legs for a horse.
Odd numbers cannot be divided by two.
Six cannot be divided by two.
17. Caribbean Islands Drift to the Left
18. Teenage Prostitution Problem Is Mounting
19. Miners Refuse to Work After Death
20. Police Begin Campaign to Run Down Jaywalkers
21. Red Tape Holds Up New Bridges
22. Juvenile Court to Try Shooting Defendant
23. Kids Make Nutritious Snacks
24. Study of Obesity Looks for Larger Test Group
25. Hospitals Sued by Seven Foot Doctors
26. Local High School Dropouts Cut in Half
27. Iraqi Head Seeks Arms
28. Drunk Gets Nine Months in Violin Case
29. Teacher Strikes Idle Kids
30. British Left Waffles on Falkland Islands
31. Stolen Painting Found by Tree
32. New Vaccine May Contain Rabies
Poetry, songs, and jokes often intentionally exploit multiple meanings for effect.
Find examples in poems, songs, and jokes that you like. Are these examples of
ambiguity on the above definition? Why or why not?
EXERCISE III
338
CHAPTER 14 ■ Fallac ies of Amb igu ity
Clearly, “odd” means “unusual” in the first premise, but it means “not
even” in the second premise. Consequently, both premises are true, even
though the conclusion is false, so the argument is not valid.
Let’s consider another, more serious, example. In Utilitarianism (1861),
John Stuart Mill claims to “prove” that “happiness is a good” with the fol-
lowing argument:
The only proof capable of being given that an object is visible is that people actu-
ally see it. The only proof that a sound is audible is that people hear it. In like
manner the sole evidence it is possible to produce that anything is desirable is
that people actually desire it. . . . [E]ach person, so far as he believes it to be
attainable, desires his own happiness. This, however, being a fact, we have not
only all the proof which the case admits of, but all which it is possible to require,
that happiness is a good.
Mill has sometimes been charged with committing a transparent fallacy in
this passage. Specifically, the following argument is attributed to him:
(1) If something is desired, then it is desirable.
(2) If it is desirable, then it is good.
(3) If something is desired, then it is good.
Mill never presents his argument in this form, and it may be uncharitable to
attribute it to him. Still, whether it is Mill’s way of arguing or not, it provides
a good specimen of a fallacy of equivocation.
The objection to this argument is that the word “desirable” is used in dif-
ferent senses in the two premises. Specifically, in the first premise, it is used to
mean “capable of being desired,” whereas in the second premise, it is used to
mean “worthy of being desired.” If so, the argument really amounts to this:
(1*) If something is desired, then it is capable of being desired.
(2*) If something is worthy of being desired, then it is good.
(3) If something is desired, then it is good.
This argument is clearly not valid. To make the charge of equivocation stick,
however, it has to be shown that the argument is not valid when the mean-
ing of the word “desirable” is used in the same sense in the two premises.
This produces two cases to be examined:
(1*) If something is desired, then it is capable of being desired.
(2**) If something is capable of being desired, then it is good.
(3) If something is desired, then it is good.
We now have a valid argument, but the second premise is not true, for some-
times people are capable of desiring things that are not good. The second
way of restoring validity takes the following form:
(1**) If something is desired, then it is worthy of being desired.
(2*) If something is worthy of being desired, then it is good.
(3) If something is desired, then it is good.
339
Equ ivocat ion
Again, we have a valid argument, but this time the first premise is false,
since sometimes people do desire things that they should not desire. Thus,
in both cases, altering the premises to produce a valid argument produces a
false premise, so the argument cannot be sound.
This is a pattern that emerges when dealing with arguments that involve
the fallacy of equivocation. When the premises are interpreted in a way that
produces a valid argument, then at least one of the premises is false. When
the premises are interpreted in a way that makes them true, then the argu-
ment is not valid. Here, then, is the strategy for dealing with arguments that
may involve a fallacy of equivocation:
1. Distinguish the possible meanings of the potentially ambiguous
expressions in the argument.
2. For each possible meaning, restate the argument so that each expression
clearly has the same meaning in all of the premises and the conclusion.
3. Evaluate the resulting arguments separately.
If the argument fails whenever each term has a consistent meaning through-
out the argument, then the argument is guilty of equivocation.
Each of the following arguments trades on an ambiguity. For each, locate the
ambiguity by showing that one or more of the statements can be interpreted in
different ways.
1. We shouldn’t hire Peter, because our company has a policy against hiring
drug users, and I saw Peter take aspirin, which is a drug.
2. Man is the only rational animal, and no woman is a man, so women are
not rational.
3. My doctor has been practicing medicine for thirty years, and practice
makes perfect, so my doctor must be nearly perfect.
4. Our cereal is all natural, for there is obviously nothing supernatural about it.
5. Ice cream is never all natural, since it never appears in nature without
human intervention.
6. I have a right to spend all my money on lottery tickets. Therefore, when
I spend all my money on lottery tickets, I am doing the right thing.
7. You passed no one on the road; therefore, you walked faster than no one.
8. Everything must have some cause; therefore, something must be the cause
of everything.
9. The apostles were twelve. Matthew was an apostle. Hence, Matthew was
twelve. (attributed to Bertrand Russell)
10. If I have only one friend, then I cannot say that I have any number of
friends. So one is not any number. (from Timothy Duggan)
EXERCISE IV
(continued)
340
CHAPTER 14 ■ Fallac ies of Amb igu ity
11. “Our bread does have fiber, because it contains wood pulp.” (The Federal
Trade Commission actually ordered the Continental Baking Company to
indicate in their advertising that this is the kind of fiber in their Fresh
Horizons bread.)
12. Anyone who tries to violate a law, even if the attempt fails, should be pun-
ished. People who try to fly are trying to violate the law of gravity. So they
should be punished. (This argument is reported to have been used in an
actual legal case during the nineteenth century, but compare Stephen Col-
bert, “Physics is the ultimate Big Government interference—universal laws
meant to constrain us at every turn. . . . Hey, is it wrong that I sometimes
want to act without having to deal with an equal and opposite reaction?”1)
1. When a newspaper was criticized as a scandalous rumormonger, its editor
responded with the following argument (as paraphrased by Deni Elliot).
Does the editor’s argument commit the fallacy of equivocation?
It’s not wrong for newspapers to pass on rumors about sex scandals. Newspapers
have a duty to print stories that are in the public interest, and the public clearly
has a great interest in rumors about sex scandals, since, when newspapers print
such stories, their circulation increases, and they receive a large number of letters.
2. In the following passage, Tom Hill Jr. claims that a common argument
against affirmative action commits a fallacy of equivocation. Do you agree
that this argument equivocates? Why or why not?
Some think that the injustice of all affirmative action programs is obvious or eas-
ily demonstrated. [One argument] goes this way: “Affirmative action, by defini-
tion, gives preferential treatment to minorities and women. This is discrimination
in their favor and against non-minority males. All discrimination by public insti-
tutions is unjust, no matter whether it is the old kind or the newer ‘reverse dis-
crimination.’ So all affirmative action programs in public institutions are unjust.”
This deceptively simple argument, of course, trades on an ambiguity. In one
sense, to “discriminate” means to “make a distinction,” to pay attention to a differ-
ence. In this evaluatively neutral sense, of course, affirmative action programs do
discriminate. But public institutions must, and justifiably do, “discriminate” in this
sense, for example, between citizens and noncitizens, freshmen and seniors, the tal-
ented and the retarded, and those who pay their bills and those who do not.
Whether it is unjust to note and make use of a certain distinction in a given context
depends upon many factors: the nature of the institution, the relevant rights of the
parties involved, the purposes and effects of making that distinction, and so on.
All this would be obvious except for the fact that the word “discrimination” is
also used in a pejorative sense, meaning (roughly) “making use of a distinction in
an unjust or illegitimate way.” To discriminate in this sense is obviously wrong,
but now it remains an open question whether the use of gender and race distinc-
tions in affirmative action programs is really “discrimination” in this sense. The
simplistic argument uses the evaluatively neutral sense of “discrimination” to
show that affirmative action discriminates; it then shifts to the pejorative sense
DISCUSSION QUESTIONS
341
Equ ivocat ion
when it asserts that discrimination is always wrong. Although one may, in the end,
conclude that all public use of racial and gender distinctions is unjust, to do so re-
quires more of an argument than the simple one (just given) that merely exploits an
ambiguity of the word “discrimination.”2
3. Many people argue that homosexuality is immoral because it is unnatural.
In the following reading,3 Burton Leiser criticizes this argument for equiv-
ocating on five meanings of the term “natural.” Does the argument really
equivocate? Why or why not?
HOMOSEXUALITY AND NATURAL LAW
by Burton Leiser
When theologians and moralists speak of homosexuality, contraception, abor-
tion, and other forms of human behavior as being unnatural and say that for
that reason such behavior must be considered to be wrong, in what sense are
they using the word unnatural? Are they saying that homosexual behavior and
the use of contraceptives are [1] contrary to the scientific laws of nature, are they
saying that they are [2] artificial forms of behavior, or are they using the terms
natural and unnatural in some third sense?
They cannot mean that homosexual behavior (to stick to the subject
presently under discussion) violates the laws of nature in the first sense [in-
cluding, for example, Boyle’s law that the volume of a gas varies inversely
with the pressure that is applied to it], for . . . in that sense it is impossible to
violate the laws of nature. Those laws, being merely descriptive of what actu-
ally does happen, would have to include homosexual behavior if such behav-
ior does actually take place. . . .
If those who say that homosexual behavior is unnatural are using the term
unnatural in the second sense as artificial, it is difficult to understand their ob-
jection. That which is artificial is often far better than what is natural. . . .
[Moreover,] homosexual behavior can hardly be considered unnatural in this
sense. There is nothing artificial about such behavior. On the contrary, it is
quite natural, in this sense, to those who engage in it. And, even if it were not,
this is not in itself a ground for condemning it.
It would seem, then, that those who condemn homosexuality as an unnatural
form of behavior must mean something else by the word unnatural, something
not covered by either of the preceding definitions. A third possibility is this:
3. Anything uncommon or abnormal is unnatural. If this is what is meant by
those who condemn homosexuality on the ground that it is unnatural, it is
quite obvious that their condemnation cannot be accepted without further
argument. The fact that a given form of behavior is uncommon provides no
justification for condemning it. . . . Great artists, poets, musicians, and scien-
tists are uncommon in this sense; but clearly the world is better off for having
them, and it would be absurd to condemn them or their activities for their
failure to be common and normal. If homosexual behavior is wrong, then, it
must be for some reason other than its unnaturalness in this sense of the word.
(continued)
342
CHAPTER 14 ■ Fallac ies of Amb igu ity
4. Any use of an organ or an instrument that is contrary to its principal purpose or
function is unnatural. Every organ and every instrument—perhaps even every
creature—has a function to perform, one for which it is particularly designed.
Any use of those instruments and organs that is consonant with their purposes
is natural and proper, but any use that is inconsistent with their principal func-
tions is unnatural and improper, and to that extent evil or harmful. Human
teeth, for example, are admirably designed for their principal functions—biting
and chewing the kinds of food suitable for human consumption. But they are
not particularly well suited for prying the caps from beer bottles. If they are
used for that purpose, they are likely to crack or break under the strain. . . .
What are the sex organs peculiarly suited to do? . . . Our sexual organs are
uniquely adapted for procreation, but that is obviously not the only function for
which they are adapted. Human beings may—and do—use those organs for a
great many other purposes, and it is difficult to see why any one use should be
considered to be the only proper one. The sex organs seem to be particularly
well adapted to give their owners and others intense sensations of pleasure.
Unless one believes that pleasure itself is bad, there seems to be little reason to
believe that the use of the sex organs for the production of pleasure in oneself or
in others is evil. In view of the peculiar design of these organs, with their great
concentration of nerve endings, it would seem that they were designed (if they
were designed) with that very goal in mind, and that their use for such purposes
would be no more unnatural than their use for the purpose of procreation.
Nor should we overlook the fact that human sex organs may be and are used
to express, in the deepest and most intimate way open to man, the love of one
person for another. Even the most ardent opponents of “unfruitful” intercourse
admit that sex does serve this function. They have accordingly conceded that a
man and his wife may have intercourse even though she is pregnant, or past the
age of child bearing, or in the infertile period of her menstrual cycle. . . .
To sum up, then, the proposition that any use of an organ that is contrary to
its principal purpose or function is unnatural assumes that organs have a prin-
cipal purpose or function, but this may be denied on the ground that the pur-
pose or function of a given organ may vary according to the needs or desires
of its owner. It may be denied on the ground that a given organ may have
more than one principal purpose or function, and any attempt to call one use
or another the only natural one seems to be arbitrary, if not question-begging.
Also, the proposition suggests that what is unnatural is evil or depraved. This
goes beyond the pure description of things, and enters into the problem of the
evaluation of human behavior, which leads us to the fifth meaning of natural.
5. That which is natural is good, and whatever is unnatural is bad. . . . Clearly,
[people who say this] cannot have intended merely to reduce the word natural
to a synonym of good, right, and proper, and unnatural to a synonym of evil,
wrong, improper, corrupt, and depraved. If that were all they had intended to
do, . . . it would follow inevitably that whatever is good must be natural, and
vice versa, by definition. This is certainly not what the opponents of homosex-
uality have been saying when they claim that homosexuality, being unnatural,
is evil. For if it were, their claim would be quite empty. They would be saying
merely that homosexuality, being evil, is evil—a redundancy that could as
easily be reduced to the simpler assertion that homosexuality is evil. This
343
Def in it ions
DEFINITIONS
It is sometimes suggested that a great many disputes could be avoided if
people simply took the precaution of defining their terms. To some extent
this is true. People do sometimes seem to disagree just because they are us-
ing terms in different ways, even though they agree on the nonverbal issues.
Nonetheless, definitions will not solve all problems, and a mindless insis-
tence on definitions can turn a serious discussion into a semantic quibble. If
you insist on defining every term, you will never be satisfied, because every
definition will introduce new terms to be defined. Furthermore, definitions
themselves can be confusing or obfuscating as, for example, when an econo-
mist tells us:
I define “inflation” as too much money chasing too few goods.
Not only is this definition metaphorical and obscure, it also has a theory of
the causes of inflation built into it.
To use definitions correctly, we must realize that they come in various
forms and serve various purposes. There are at least five kinds of definitions
that need to be distinguished:
1. Lexical or dictionary definitions are the most common kind of definition.
We consult a dictionary when we are ignorant about the meaning of a word in
a particular language. If you do not happen to know what the words “jejune,”
“ketone,” or “Kreis” mean, then you can look these words up in an English, a
scientific, and a German dictionary, respectively.
Except for an occasional diagram, dictionaries explain the meaning of a
word by using other words that the reader presumably already under-
stands. These explanations often run in a circle, such as when the Oxford
American Dictionary defines “car” as “automobile” and “automobile” as
“car.” Circular definitions can still be useful, because if you know what one
assertion, however, is not an argument. . . . “Unnaturalness” and “wrongful-
ness” are not synonyms, then, but different concepts.
The problem with which we are wrestling is that we are unable to find a
meaning for unnatural that enables us to arrive at the conclusion that homosexu-
ality is unnatural or that if homosexuality is unnatural, it is therefore wrongful
behavior. We have examined [five] common meanings of natural and unnatural,
and have seen that none of them performs the task that it must perform if the
advocates of this argument are to prevail. Without some more satisfactory
explanation of the connection between the wrongfulness of homosexuality and
its alleged unnaturalness, the argument [that homosexuality is wrong because
it is unnatural] must be rejected.
344
CHAPTER 14 ■ Fallac ies of Amb igu ity
of the terms in the circle means, you can use that background knowledge
plus the definition to figure out what the other terms mean.
The goal of dictionary definitions is to supply us with factual information
about the standard meanings of words in a particular language. As dictionary
definitions are, in effect, factual claims about how people in general actually
use certain words, dictionary definitions can be either accurate or inaccurate.
The Oxford American Dictionary defines one meaning of “fan” as “a device
waved in the hand or operated mechanically to create a current of air.” This is,
strictly speaking, incorrect because a bellows also meets these conditions but is
not a fan. Dictionary definitions can be criticized or defended on the basis of a
speaker’s sense of the language or, more formally, by empirical surveys of
what speakers accept as appropriate or reject as inappropriate uses of the term.
2. Disambiguating definitions specify a sense in which a word or phrase is or
might be being used by a particular speaker on a particular occasion. (“When I
said that the banks were collapsing, I meant river banks, not financial institu-
tions.”) Disambiguating definitions can tell us which dictionary definition ac-
tually is intended in a particular context, or they can distinguish several
meanings that might be intended. They can also be used to remove syntactic
ambiguity or amphiboly. (“When I said that all of my friends are not students,
I meant that not all of them are students, not that none of them are students.”)
Whether the ambiguity is semantic or syntactic, the goal of a disam-
biguating definition is to capture what the speaker intended, so such defini-
tions can be justified by asking the speaker what he or she meant. This is a
different question than asking what a word means. Whereas dictionary defi-
nitions say what words mean or how they are used by most speakers of the
language, a disambiguating definition focuses on a particular speaker and
specifies which meaning that speaker intended on a particular occasion.
Such disambiguating definitions can be used in response to arguments that
seem to commit the fallacy of equivocation. A critic can use disambiguating
definitions to distinguish possible meanings and then ask, “Did you mean this
or that?” The person who gave the argument can answer by picking one of
these alternatives or by providing another disambiguating definition to spec-
ify what was meant. Speakers are sometimes not sure which meaning they
intended, and then the critic needs to show that the argument cannot work if
a single disambiguating definition is followed throughout. Whether one sides
with the arguer or the critic, arguments that use terms ambiguously cannot be
evaluated thoroughly without the help of disambiguating definitions.
3. Stipulative definitions are used to assign a meaning to a new (usually
technical) term or to assign a new or special meaning to a familiar term.
They have the following general form: “By such and such expression I (or
we) will mean so and so.” Thus, mathematicians introduced the new term
“googol” to stand for the number expressed by 1 followed by 100 zeroes.
Physicists use words like “charm,” “color,” and “strangeness” to stand for
certain features of subatomic particles. Stipulative definitions do not report
345
Def in it ions
what a word means; they give a new word a meaning or an old word a new
meaning.
Notice that if I say, “I stipulate that . . . ” I thereby stipulate that . . . ; so such
utterances are explicit performatives, and stipulation is a speech act. (See
Chapter 2.) This explains why stipulative definitions cannot be false, since no
performatives can be false. Stipulative definitions can, however, be criticized in
other ways. They can be vague or ambiguous. They can be useless or confus-
ing. Someone who stipulates a meaning for a term might go on to use the term
with a different meaning (just as people sometimes fail to keep their promises).
Still, stipulative definitions cannot be false by virtue of failing to correspond to
the real meaning of a word, because they give that meaning to that word.
4. Precising definitions are used to resolve vagueness. They are used to
draw a sharp (or sharper) boundary around the things to which a term
refers, when this collection has a fuzzy or indeterminate boundary in ordi-
nary usage. For example, it is not important for most purposes to decide
how big a population center must be in order to count as a city rather than
as a town. We can deal with the borderline cases by using such phrases as
“very small city” or “quite a large town.” It will not make much difference
which phrase we use on most occasions. Yet it is not hard to imagine a situa-
tion in which it might make a difference whether a center of population is a
city or not. As a city, it might be eligible for development funds that are not
available to towns. Here a precising definition—a definition that draws a
sharp boundary where none formerly existed—would be useful.
Precising definitions are, in effect, combinations of stipulative definitions
and dictionary definitions. Like stipulative definitions, they involve a choice.
One could define a city as any population center with more than 50,000 peo-
ple, or one could decide to decrease the minimum to 30,000 people. Precising
definitions are not completely arbitrary, however, because they usually
should conform to the generally accepted meaning of a term. It would be un-
reasonable to define a city as any population center with more than seven-
teen people. Dictionary definitions, thus, set limits to precising definitions.
Precising definitions are also not arbitrary in another way: There can be
good reasons to prefer one precising definition over another, when adopting
the preferred definition will have better effects than the alternative. If devel-
opment funds are to be distributed only to cities, then to define cities as hav-
ing more than 50,000 people will deny those funds to smaller population
centers with, say, 10,000 people. Consequently, we need some reason to re-
solve the vagueness of the term “city” in one way rather than another. In this
case, the choice might be based on the amount of funds available for devel-
opment. In a more dramatic example, a precising definition of “death” might
be used to resolve controversial issues about euthanasia—about what doc-
tors may or must do to patients who are near death—and then our choices
between possible precising definitions might be based on our deepest value
commitments. In any case, we need some argument to show that one precis-
ing definition is better than other alternatives.
346
CHAPTER 14 ■ Fallac ies of Amb igu ity
Such arguments often leave some leeway. Even if one can justify defining
cities as having a minimum of 50,000 people instead of 10,000, one’s reason
is not likely to justify a cutoff at 50,000 as opposed to 49,000. A different kind
of defense would be needed if someone used a slippery-slope argument to
show that it is unfair to provide development funds to one city with 50,000
people but to deny such funds to its neighbor with only 49,000 people.
Against this kind of charge, the only way to defend a precising definition
might be to show that some precising definition is needed, the cutoff should
lie inside a certain general area, one’s preferred definition does lie within
that area, and no alternative is any better. Such responses might also apply
to nearby alternatives, but they are still sometimes enough to support a pre-
cising definition. If responses like these are not available, then a precising
definition can be criticized as unjustified.
5. Systematic or theoretical definitions are introduced to give a systematic
order or structure to a subject matter. For example, in geometry, every term
must be either a primitive (undefined) term or a term defined by means of
these primitive terms. Thus, if we take points and distances as primitives, we
can define a straight line as the shortest distance between two points. Then,
assuming some more concepts, we can define a triangle as a closed figure
with exactly three straight lines as sides. By a series of such definitions, the
terms in geometry are placed in systematic relationships with one another.
In a similar way, we might try to represent family relationships using
only the primitive notions of parent, male, and female. We could then con-
struct definitions of the following kind:
“A is the brother of B.” = “A and B have the same parents and A is male.”
“A is B’s grandmother.” = “A is a parent of a parent of B and A is female.”4
Things become more complicated when we try to define such notions as
“second cousin once removed” or “stepfather.” Yet, by extending some basic
definitions from simple to more complicated cases, all family relationships
can be given a systematic presentation.
Formulating systematic definitions for family relationships is relatively
easy, but similar activities in science, mathematics, and other fields can de-
mand genius. It often takes deep insight into a subject to see which concepts
are genuinely fundamental and which are secondary and derivative. When
Sir Isaac Newton defined force in terms of mass and acceleration, he was not
simply stating how he proposed to use certain words; he was introducing a
fundamental conceptual relationship that improved our understanding of
the physical world.
Such theoretical definitions can be evaluated on the basis of whether they
really do help us formulate better theories and understand the world. Evalu-
ating theoretical definitions often requires a great deal of empirical investiga-
tion. When water was defined as H2O,5 this made it possible to formulate
more precise laws about how water interacted with other chemicals. Other
alternatives were available. Whereas molecules count as H2O, and hence as
347
Def in it ions
water, even if they contain unusual isotopes of hydrogen and oxygen,
chemists could define water so that it would have to contain only the most
common isotopes of hydrogen and oxygen. Why don’t they? Because they
discovered that differences among isotopes generally do not affect how mole-
cules of H2O react with other chemicals. As a result, the simplest and most
useful generalizations about the properties of water can be formulated in
terms of H2O without regard to certain isotopes of hydrogen and oxygen. This
illustrates one way in which choosing one theoretical definition over another
can lead to a better theory.
Definitions can play important roles in the presentation of arguments, but
demands for definitions can also hinder the progress of an argument. In the
middle of discussions people often ask for definitions or even state, usually
with an air of triumph, that e